\"intelligent\" telephone answering machine...

On 10/24/20 4:36 PM, Don Y wrote:
> How do all IVR systems \"recognize voices\" -- voice recognition!

Be mindful that \"Interactive Voice Response\" can mean multiple different
things.

I\'ve used IVRs that are simply \"interactive\" as in they \"respond\" to key
presses with a \"voice\" recording. Or \"interactively (a recorded) voice
responds (to your key presses)\".

I\'ve also used IVRs that actually apply voice recognition and try to
suss out words / phrases. I find these to be extremely annoying, if not
actually irritating.



--
Grant. . . .
unix || die
 
On 2020-10-24, Don Y <blockedofcourse@foo.invalid> wrote:
On 10/24/2020 2:54 PM, Jasen Betts wrote:
Someone needs to take a class action suit against the VOIP providers
against their publishing unverified CID data. Users should have to
prove control of the number before they are allowed to use it in CID

This is not trivial.

The 6,000 telephones at the local hospital would provide meaningless
information if uniquely identified. Instead, they\'d all want to
map to \"The Local Hospital\".

Someone working from home would want to be identified as an
agent of \"Company XYZ\" -- but only while on company business.

And, you\'d want the number used for RETURNED calls to reflect
the needs of that (legitimate) caller; the home worker wouldn\'t
want to be receiving calls back from \"customers\"

Seems failry trivial, substitute customer for user in my statement, in
this contenxt the terms are almost synonymous.

bottom line CID identities should not be usable without permission
from the party identified.

--
Jasen.
 
On 10/24/2020 6:33 PM, Jasen Betts wrote:
On 2020-10-24, Don Y <blockedofcourse@foo.invalid> wrote:
On 10/24/2020 2:54 PM, Jasen Betts wrote:
Someone needs to take a class action suit against the VOIP providers
against their publishing unverified CID data. Users should have to
prove control of the number before they are allowed to use it in CID

This is not trivial.

The 6,000 telephones at the local hospital would provide meaningless
information if uniquely identified. Instead, they\'d all want to
map to \"The Local Hospital\".

Someone working from home would want to be identified as an
agent of \"Company XYZ\" -- but only while on company business.

And, you\'d want the number used for RETURNED calls to reflect
the needs of that (legitimate) caller; the home worker wouldn\'t
want to be receiving calls back from \"customers\"

Seems failry trivial, substitute customer for user in my statement, in
this contenxt the terms are almost synonymous.

bottom line CID identities should not be usable without permission
from the party identified.

CID tries to be two *different* things: ANI and \"caller identification\".

The \"called party\" likely does NOT want to know that he\'s receiving
a call from some particular number but, rather, from some particular
*party*. And, he may not want to know the name of the \"individual\"
calling but, rather, the entity that is trying to contact him.
E.g., I don\'t want to know that Mary Beth at 555-1212 is calling me.
Rather, I want to know that The Billing Department at The Local
Hosital is calling. And, 555-1212 may not accept inbound calls so
having that number presented to me is useless (will we *require*
any number displayed as also being callable?)

If Mary Beth is now working from home, I sure don\'t want to see
her home phone number (nor does she want me to see it!) or
that ambiguous moniker. And, come 5:00PM, calls that she makes
should reveal *her* identity. etc.

When dealing with individuals, you want to know the identity of the
individual that is calling you. When dealing with organizations,
you want the identity of the organization. Who makes the decision
as to what information gets conveyed to the callee? If I see
CID indicating \"Mary Beth\", will I wonder if Mary Beth isn\'t,
perhaps, some friend I\'ve forgotten? Or, some employee of a
firm that I want contact with? If I see \"The Local Hospital\",
will I know that it\'s my buddy who works in the IT department,
there, calling? Or, maybe the billing department?

I got a call, yesterday, logged as \"Kathy XXXXX\" from some number
I didn\'t recognize. Turned out to be my buddy Pete using his
wife\'s phone (her maiden name unbeknownst to me). But, because I
(my box) recognized his voice, the call was gated through.

You can\'t solve this problem with a simplistic binary decision
mechanism based on ONE datum (CID).
 
On 10/24/2020 6:10 PM, Grant Taylor wrote:
On 10/24/20 4:36 PM, Don Y wrote:
How do all IVR systems \"recognize voices\" -- voice recognition!

Be mindful that \"Interactive Voice Response\" can mean multiple different things.

I\'ve used IVRs that are simply \"interactive\" as in they \"respond\" to key
presses with a \"voice\" recording. Or \"interactively (a recorded) voice
responds (to your key presses)\".

I\'ve also used IVRs that actually apply voice recognition and try to suss out
words / phrases. I find these to be extremely annoying, if not actually
irritating.

We routinely interact with systems that request \"a few words about what you
are calling about\" (I had to confirm receipt of a new credit card, yesterday,
without talking to anyone). They work well because they expect to operate in
reasonably limited vocabularies -- you\'re unlikely to ask for \"help purchasing
waterskis for your aardvark!\" When you think about it, it\'s a multiple choice
question with very many possible \"keys\" mapping to each \"choice\".

[Though I know many folks who regularly use unconstrained speech recognition
with a reasonably high degree of accuracy! And, that\'s not to mention Alexa,
Siri, etc.]

If you know ahead of time what the caller is likely to say (e.g., one of
several one-time passwords that you\'ve issued), then you can greatly
improve the recognition accuracy. An \"uncooperative\" caller just wastes
his (and your) time getting nothing for their efforts. (if that\'s your
goal, why not just constantly keep redialing the number all day long??)

[And, you can always fall back on, \"I\'m sorry, I didn\'t understand you.
Could you please repeat that?\"]
 
On Sunday, October 25, 2020 at 1:19:36 AM UTC-4, Don Y wrote:
On 10/24/2020 6:10 PM, Grant Taylor wrote:
On 10/24/20 4:36 PM, Don Y wrote:
How do all IVR systems \"recognize voices\" -- voice recognition!

Be mindful that \"Interactive Voice Response\" can mean multiple different things.

I\'ve used IVRs that are simply \"interactive\" as in they \"respond\" to key
presses with a \"voice\" recording. Or \"interactively (a recorded) voice
responds (to your key presses)\".

I\'ve also used IVRs that actually apply voice recognition and try to suss out
words / phrases. I find these to be extremely annoying, if not actually
irritating.

We routinely interact with systems that request \"a few words about what you
are calling about\" (I had to confirm receipt of a new credit card, yesterday,
without talking to anyone). They work well because they expect to operate in
reasonably limited vocabularies -- you\'re unlikely to ask for \"help purchasing
waterskis for your aardvark!\" When you think about it, it\'s a multiple choice
question with very many possible \"keys\" mapping to each \"choice\".

They work well because they aren\'t trying to figure out who you are. They know because of specific information you provide they can verify like your card number.

I typically am calling because an automated system can\'t help me so I press other buttons and repeat \"let me talk to a person\" many, many times until I reach a person... but by then I am pissed off and I let them know it. No one likes the stupid machines. Using them is a great way to irritate and lose customers.


[Though I know many folks who regularly use unconstrained speech recognition
with a reasonably high degree of accuracy! And, that\'s not to mention Alexa,
Siri, etc.]

Lol! They are totally bad. I ask Google to turn off airplane mode and sometimes it does it, other times it says, \"I can\'t do that, your phone is in airplane mode\". Or I ask it to snooze the alarm and it says, \"You can do that in the app\". So I reply with, \"Snooze the fucking alarm!\" and it does it. If I didn\'t enjoy the fact that swearing at the phone gets the job done I might just throw it out the window.


If you know ahead of time what the caller is likely to say (e.g., one of
several one-time passwords that you\'ve issued), then you can greatly
improve the recognition accuracy. An \"uncooperative\" caller just wastes
his (and your) time getting nothing for their efforts. (if that\'s your
goal, why not just constantly keep redialing the number all day long??)

Better yet is to do business with a company that employees people. Google has dropped phone support for their phones \"because of the pandemic\". They can\'t seem to figure out how to let people work from home. I guess I\'m going to switch to ATT now. That\'s the last straw for me, not that ATT is very good either.


[And, you can always fall back on, \"I\'m sorry, I didn\'t understand you.
Could you please repeat that?\"]

That\'s when I start figuring out the home mailing address of the company president so I can mail a bomb. Then I start working on the technical people who put in the voice systems.

--

Rick C.

---- Get 1,000 miles of free Supercharging
---- Tesla referral code - https://ts.la/richard11209
 
On 10/24/2020 5:36 PM, Ricketty C wrote:
On Saturday, October 24, 2020 at 6:37:01 PM UTC-4, Don Y wrote:

You either block specific numbers you recognize as spam, which is
what the NOMOROBO does, or you block EVERY CALL that isn\'t white
listed.

No, you allow whitelisted calls (to be further conditioned based on
identity, time-of-day, etc.) to be \"accepted\" (ring-through, routed
to personalized mailbox, etc.).

I didn\'t say anything different.

But you assume \"non-white\" calls are black. That\'s the mistake.

No, if you pay attention to what I write I am saying your system is fatally
flawed regardless because you are making unfounded assumptions. You are the
one trying to distinguish white/black/grey calls. This simply does not work
in a useful way for anyone other than yourself because you have designed the
system around your tastes and presumably are willing to live with the
limitations.

There are always white, grey and black calls. Assuming that you can make a
binary distinction (black vs white, reject vs. accept) based on something
as crude as a CID is naive. You don\'t know what a particular CID *means*,
even if it is unforgeable! (is the hospital calling to tell you about
your upcoming appointment? or, to request a payment? or, to see if you\'d
like to participate in their blood drive??)

[I have an \"account\" with the local university as I buy surplus equipment
from them. When they call, are they trying to inform me that I won a
particular lot at auction? Or, hoping I\'ll donate to their public television
station?? Both have happened. I\'m obviously far more interested in
the former call than the latter!]

Non-whitelisted calls you identify with other MORE RELIABLE means. In
our case, we do that by recognizing the voice of the caller (we don\'t
\"recognize\" any spammers :> )

How does your machine recognize voices?

How do all IVR systems \"recognize voices\" -- voice recognition!

How does your BANK *identify* you as the caller -- speaker
identification!

My bank identifies me by my proving intimate details of my account. I have
no idea what you are talking about now.

\"My voice is my password\" TmReg.

Companies and agencies are starting to realize that folks simply can\'t
keep up with the various artificial authentication mechanisms that they put in
place (PINs, secret questions, etc.) and are moving towards more biometric
identifiers. When I call my credit card company, they note the number from
which I\'m calling to identify me and my account. I am only forced to talk
with a human if I choose to call from some \"unregistered\" phone number.

[What if someone spoofs my CID when calling them?]

As there is no universal \"spammer voice\", there\'s no need to try to
identify the spammer. \"Voice (identity) not recognized\"

This seems really out in left field now.

If you mean YOU listen, I don\'t even need to do that. But I don\'t want
to be bothered with the calls. No ring, no messages, no monitoring the
call. That\'s what the $0.10 per call fee will do. Spammers and the
services that connect them to the phone network won\'t be able to afford
to make thousands of calls to bring in a few hundred dollars.

I don\'t listen, the voice attendant does. It decides who you are by
analyzing the feature-set that it derives (while listening to you) from
your words. It compares this to a database of feature-sets to form an
appraisal of WHO you are. It augments this with any supplemental
authenticators (CID) -- even if not authoritative -- as well as the
CONTENT of your speech and conditions of the call (time of day, etc.)

So you aren\'t going to explain what you are talking about??? Whatever.

There is a little box. The wires to the PSTN come in one end. An ethernet
connection comes in the other (bringing power with it). Inside the box is
an interface to the telco network. It allows the processor (in the box)
to detect ring cadences, go off-hook, \"listen\" to incoming audio as well
as generate outgoing audio.

It can be programmed to \"answer\" -- or not -- based on whatever criteria
I deem appropriate (I might want to block ALL incoming calls at certain times).

When it answers, it can be programmed to emit whatever I choose to say
(via synthesized voice). That \"voice prompt\" can be determined by time-of-day,
incoming CID, \"expectations\" (i.e., if I\'m *expecting* a call and, who from),
etc.

Presumably, the calling party will eventually \"make sounds\" which the box
will detect (and analyze and record).

If the caller \"talks over\" my outgoing message, then it is likely a DUMB
autodialer -- one that simply waits for the phone to be answered and
then begins its spiel. A real human -- especially one familiar with
how my phone works -- would patiently await the end of the OGM.
[assuming they don\'t know of shortcuts built into the answer process]

The caller will \"make sounds\" (typically \"speak\" -- but it could also
be a FAX machine, etc.) for some amount of time. If speech, the amount
of time before a pause reveals something about the caller (a telemarketer
or robodialer will typically not wait for the short speech burst that are
typical at the start of an interactive conversation).

Typical conversations have a formulaic beginning (think about them... from
friends/relatives/colleagues vs. businesses vs. spammers). It\'s unlikely
that the caller is going to utter \"alpha geronimo boondoggle 27\" -- so
you have a restricted vocabulary to typically process.

I can recognize likely names in those utterances -- especially because
I know which names are significant, to *me*! (if you start talking
about Art Linkletter, I\'m likely going to glaze over on you)

While listening to WHAT you are saying (speech recognition), I\'m also
listening to HOW you are saying it -- the cadence, fundamental frequency,
how you pronounce vowels (cuz I know what vowels are in each of the
RECOGNIZED WORDS that you\'ve uttered). This \"feature set\" is a key into
a database of feature sets compiled from speech samples obtained from
other calls WITH KNOWN OTHER PARTIES (regardless of whether I was caller
or callee). So, I can form an assessment of WHO you likely are
(speaker identification -- not to be confused with authentication!).

I can prod you by emitting additional utterances intended (in a live caller)
to elicit certain types of information in certain formulaic ways (NLP).
E.g., if I (\"I\" in all of this is The Box) say, \"Who are you looking to speak
with?\", I know I should expect to hear one of two names in response...
not \"The Man of The House\" or \"Your Mother/Father\" (yeah, I can answer the
phone with any voice characteristics I choose! I can actually synthesize
YOUR voice after I\'ve heard enough samples on which to build a model!)

Of course, legitimate callers who\'ve been through the routine know
what to expect and how to respond (i.e., like a normal human!). And,
each response gives me more data to refine my assessment of their
identity. If I ever need confirmation, I can utter, \"Who is calling,
please?\" and \"recognize\" the response in a dictionary of \"valid callers\".

[Each of these provides data that the AI uses to build confidence in
its assessment of the caller\'s identity AND how it should handle the
call FROM that caller.]

Once I\'m confident in the caller\'s identity, I can decide how to handle
their call in the current circumstances (\"Sorry, Don is not available\";
\"Don is on his way to your house, Bruce\"; \"What should I tell Don this
call concerns?\"; etc.)

Again, depending on caller\'s identity, time of day, mood, occupancy, etc.
the call may be \"announced\" to me (or my other half, if intended for her)
with the expectation that I\'ll respond \"put him/her through\". I can
alternately say \"take a message\" or \"tell him to call back later\" or...

Just like you\'d tell your secretary when she announces that \"Bob Smith
is on line 2\"

While I\'m talking with Bob Smith, the box is listening in and recording
the conversation. This frees me from the need to analyze the content
in real-time.

At the same time, it schedules an analysis task that will listen to
these TAGGED (caller identified) recordings. This begins with diarization
then analysis of the callee\'s speech to further refine the models that
are used to identify that caller -- as well as enhance the speech
recognition FOR that caller.

This mimics how a human user recognizes callers and \"learns\" about
them, over time (e.g., Pete always calls at 5:00PM).

Note that there is no explicit \"training\" event that I\'d have to
require callers to endure -- it\'s a \"free\" side-effect of the
implementation (by design!). And, the accumulated dataset allows
those KNOWN callers to interact with my system in other ways.

In particular, it lets very special individuals (e.g., me and mine) gain
access to subsystems that wouldn\'t be accessible to mere mortals,
remotely. So, I can give my neighbors access to the house (or parts
thereof) based on their voice and other 2FA authenticators. Again,
without having to \"train\" the system for that access.

[In my initial demo of the speaker identification system, I had
my doorbell query arriving visitors and welcomed them, by name,
based on my recognition of their individual voices from past
phone conversations.]

These factors are then used to decide how your call should be handled. A
business relation\'s call might be routed to voice mail \"after hours\"
(because I don\'t think they deserve to cut into my personal time). A
family member\'s call might prompt the attendant to contact me: \"Your
sister is on the phone. Do you want to speak with her, now?\" A neighbor\'s
call (or visit!) might warrant chasing me down -- even if I\'m in the back
yard or asleep. And, a call from my other half (assuming she\'s not home)
should go to great pains to find me IF SHE CLAIMS IT IS URGENT.

You keep going on about handling different calls differently without
explaining how you distinguish the callers. In particular you claim you are
not blocking calls based on CID but then say you do use CID along with
\"feature-sets\" which you fail to explain. Looking back for the term you
used so I could quote it I see your post has gotten longer and longer.

You want go-nogo to be based on a single criteria. I take a more
realistic approach and base it on many criteria. Speaker identification,
alone, requires an analysis of many non-binary criteria -- feature sets.

This is common in more advanced pattern recognition problems. How do you
differentiate a written 4 from a 9? Is there some magic threshold that
you can use to make a binary choice? No, you look at many aspects of
each glyph and come up with a probabilistic assessment of each \"set of
features\" observed in the writings vs. some prior knowledge of what
those features are likely to be for those particular glyphs.

How do you know that the caller is male vs. female? Yet, you know
that \"Bob Smith\" is definitely MALE -- so, if you think the current caller
is FEmale, it\'s likely NOT Bob Smith. That aspect of the observed feature
set conflicts with the STORED feature set for Bob Smith!

I\'m not interested in reading your book. If you want to explain simply I\'m
happy to listen, but if you just want to go on and on, that makes it very
clear to me that you have developed a very sophisticated non-solution to a
real problem.

You are obviously ignorant of the technologies involved and, thus, can\'t
wrap your head around what is currently possible (and already in use).
Google \"diarization\" (so you know WHY you need it), \"speaker recognition\",
\"speech recognition\", \"dynamic pattern recognition\", \"expert systems\",
\"speech synthesis\", etc.

Familiarize yourself with the technologies and the issues addressed by each.
Familiarize yourself with the \"open\" implementations for each of these (so
you understand how much is already in place). Actually BUILD some demo apps
so you can see what they can and can\'t reliably do. Be amazed!

Then, sit down and think about how you could make a similar system to do what
I\'ve described. I\'ve two colleagues that are making significant investments
in technology derived from my implementation.

If you can\'t, well... the world needs ditch diggers, too!

Technology can solve every problem. Sometimes laws and regulations are the
right solution. In this case a simple $0.10 fee per call is not only easy
to implement, the mechanics are ALREADY built into the system. It just has
to be turned on.

When skimming I saw you mention that cell towers have large computational

That\'s your problem -- you\'re skimming and not reading. You\'re not interested
in understanding an implementation, just looking to poke at (incorrectly
perceived) issues. Grep my message for \"cell tower\"... didn\'t find it, did
you? :> I said \"cell phone\". And most (many) cell phones have oodles of
computational power. I\'ve got a rescued phone lying on the desk in front of
me that has 2G of RAM, 16G FLASH and an 8 core 1.7GHz CPU. And, the \"DAA\"
is already present in it -- along with a network connection to a conceivably
unlimited amount of \"remote\" storage AND an interface to the callee\'s \"ear\"!

capacities... which is used by the cell calls, that\'s why the capacities are
there, because they are needed. They didn\'t just toss a bunch of processors
onto a board they don\'t need.

I\'m not interested in continuing the conversation if you can\'t get to the
point.

I still say you have no useful way (by \"useful\" I mean to the rest of the
world) to distinguish spammers from anyone else, because no matter what you
do, the spammers will adapt. That\'s the reason why there is so much email
spam. Do you really think email just needs a better filter?

I\'ve not received a piece of junk email in YEARS! Nor has my other half -- and
she\'s far less \"computer literate\" than I! You obviously have been doing
something wrong!
 
On Sunday, October 25, 2020 at 2:26:24 AM UTC-4, Don Y wrote:
On 10/24/2020 5:36 PM, Ricketty C wrote:
On Saturday, October 24, 2020 at 6:37:01 PM UTC-4, Don Y wrote:

You either block specific numbers you recognize as spam, which is
what the NOMOROBO does, or you block EVERY CALL that isn\'t white
listed.

No, you allow whitelisted calls (to be further conditioned based on
identity, time-of-day, etc.) to be \"accepted\" (ring-through, routed
to personalized mailbox, etc.).

I didn\'t say anything different.

But you assume \"non-white\" calls are black. That\'s the mistake.

No, if you pay attention to what I write I am saying your system is fatally
flawed regardless because you are making unfounded assumptions. You are the
one trying to distinguish white/black/grey calls. This simply does not work
in a useful way for anyone other than yourself because you have designed the
system around your tastes and presumably are willing to live with the
limitations.

There are always white, grey and black calls. Assuming that you can make a
binary distinction (black vs white, reject vs. accept) based on something
as crude as a CID is naive. You don\'t know what a particular CID *means*,
even if it is unforgeable! (is the hospital calling to tell you about
your upcoming appointment? or, to request a payment? or, to see if you\'d
like to participate in their blood drive??)

[I have an \"account\" with the local university as I buy surplus equipment
from them. When they call, are they trying to inform me that I won a
particular lot at auction? Or, hoping I\'ll donate to their public television
station?? Both have happened. I\'m obviously far more interested in
the former call than the latter!]

Non-whitelisted calls you identify with other MORE RELIABLE means. In
our case, we do that by recognizing the voice of the caller (we don\'t
\"recognize\" any spammers :> )

How does your machine recognize voices?

How do all IVR systems \"recognize voices\" -- voice recognition!

How does your BANK *identify* you as the caller -- speaker
identification!

My bank identifies me by my proving intimate details of my account. I have
no idea what you are talking about now.

\"My voice is my password\" TmReg.

Companies and agencies are starting to realize that folks simply can\'t
keep up with the various artificial authentication mechanisms that they put in
place (PINs, secret questions, etc.) and are moving towards more biometric
identifiers. When I call my credit card company, they note the number from
which I\'m calling to identify me and my account. I am only forced to talk
with a human if I choose to call from some \"unregistered\" phone number.

I don\'t know what credit cards you have, but I have to give them my card number and other identifying information that no one else has. They don\'t give a crap about my phone number.

I literally don\'t know what you are talking about with \"forced\" to talk to a person. There\'s virtually no reason to call on a phone if you aren\'t going to talk. I can much more easily and usefully get the info on my account that can be handled automatically simply by using the web site. I only use the phone when I have to talk to someone.


[What if someone spoofs my CID when calling them?]

As there is no universal \"spammer voice\", there\'s no need to try to
identify the spammer. \"Voice (identity) not recognized\"

This seems really out in left field now.

If you mean YOU listen, I don\'t even need to do that. But I don\'t want
to be bothered with the calls. No ring, no messages, no monitoring the
call. That\'s what the $0.10 per call fee will do. Spammers and the
services that connect them to the phone network won\'t be able to afford
to make thousands of calls to bring in a few hundred dollars.

I don\'t listen, the voice attendant does. It decides who you are by
analyzing the feature-set that it derives (while listening to you) from
your words. It compares this to a database of feature-sets to form an
appraisal of WHO you are. It augments this with any supplemental
authenticators (CID) -- even if not authoritative -- as well as the
CONTENT of your speech and conditions of the call (time of day, etc.)

So you aren\'t going to explain what you are talking about??? Whatever.

There is a little box. The wires to the PSTN come in one end. An ethernet
connection comes in the other (bringing power with it). Inside the box is
an interface to the telco network. It allows the processor (in the box)
to detect ring cadences, go off-hook, \"listen\" to incoming audio as well
as generate outgoing audio.

That\'s what I thought. An invention for the previous century. Some huge percentage of phone numbers are cell phones. So your \"invention\" is pretty much worth less. That\'s why no one has produced it before.


It can be programmed to \"answer\" -- or not -- based on whatever criteria
I deem appropriate (I might want to block ALL incoming calls at certain times).

When it answers, it can be programmed to emit whatever I choose to say
(via synthesized voice). That \"voice prompt\" can be determined by time-of-day,
incoming CID, \"expectations\" (i.e., if I\'m *expecting* a call and, who from),
etc.

Presumably, the calling party will eventually \"make sounds\" which the box
will detect (and analyze and record).

If the caller \"talks over\" my outgoing message, then it is likely a DUMB
autodialer -- one that simply waits for the phone to be answered and
then begins its spiel. A real human -- especially one familiar with
how my phone works -- would patiently await the end of the OGM.
[assuming they don\'t know of shortcuts built into the answer process]

The caller will \"make sounds\" (typically \"speak\" -- but it could also
be a FAX machine, etc.) for some amount of time. If speech, the amount
of time before a pause reveals something about the caller (a telemarketer
or robodialer will typically not wait for the short speech burst that are
typical at the start of an interactive conversation).

Typical conversations have a formulaic beginning (think about them... from
friends/relatives/colleagues vs. businesses vs. spammers). It\'s unlikely
that the caller is going to utter \"alpha geronimo boondoggle 27\" -- so
you have a restricted vocabulary to typically process.

I can recognize likely names in those utterances -- especially because
I know which names are significant, to *me*! (if you start talking
about Art Linkletter, I\'m likely going to glaze over on you)

While listening to WHAT you are saying (speech recognition), I\'m also
listening to HOW you are saying it -- the cadence, fundamental frequency,
how you pronounce vowels (cuz I know what vowels are in each of the
RECOGNIZED WORDS that you\'ve uttered). This \"feature set\" is a key into
a database of feature sets compiled from speech samples obtained from
other calls WITH KNOWN OTHER PARTIES (regardless of whether I was caller
or callee). So, I can form an assessment of WHO you likely are
(speaker identification -- not to be confused with authentication!).

I can prod you by emitting additional utterances intended (in a live caller)
to elicit certain types of information in certain formulaic ways (NLP).
E.g., if I (\"I\" in all of this is The Box) say, \"Who are you looking to speak
with?\", I know I should expect to hear one of two names in response...
not \"The Man of The House\" or \"Your Mother/Father\" (yeah, I can answer the
phone with any voice characteristics I choose! I can actually synthesize
YOUR voice after I\'ve heard enough samples on which to build a model!)

Of course, legitimate callers who\'ve been through the routine know
what to expect and how to respond (i.e., like a normal human!). And,
each response gives me more data to refine my assessment of their
identity. If I ever need confirmation, I can utter, \"Who is calling,
please?\" and \"recognize\" the response in a dictionary of \"valid callers\".

[Each of these provides data that the AI uses to build confidence in
its assessment of the caller\'s identity AND how it should handle the
call FROM that caller.]

Once I\'m confident in the caller\'s identity, I can decide how to handle
their call in the current circumstances (\"Sorry, Don is not available\";
\"Don is on his way to your house, Bruce\"; \"What should I tell Don this
call concerns?\"; etc.)

Again, depending on caller\'s identity, time of day, mood, occupancy, etc.
the call may be \"announced\" to me (or my other half, if intended for her)
with the expectation that I\'ll respond \"put him/her through\". I can
alternately say \"take a message\" or \"tell him to call back later\" or...

Just like you\'d tell your secretary when she announces that \"Bob Smith
is on line 2\"

While I\'m talking with Bob Smith, the box is listening in and recording
the conversation. This frees me from the need to analyze the content
in real-time.

At the same time, it schedules an analysis task that will listen to
these TAGGED (caller identified) recordings. This begins with diarization
then analysis of the callee\'s speech to further refine the models that
are used to identify that caller -- as well as enhance the speech
recognition FOR that caller.

This mimics how a human user recognizes callers and \"learns\" about
them, over time (e.g., Pete always calls at 5:00PM).

Note that there is no explicit \"training\" event that I\'d have to
require callers to endure -- it\'s a \"free\" side-effect of the
implementation (by design!). And, the accumulated dataset allows
those KNOWN callers to interact with my system in other ways.

In particular, it lets very special individuals (e.g., me and mine) gain
access to subsystems that wouldn\'t be accessible to mere mortals,
remotely. So, I can give my neighbors access to the house (or parts
thereof) based on their voice and other 2FA authenticators. Again,
without having to \"train\" the system for that access.

[In my initial demo of the speaker identification system, I had
my doorbell query arriving visitors and welcomed them, by name,
based on my recognition of their individual voices from past
phone conversations.]

Wow! So much noise and so little utility. You are designing a product for the Ozzie and Harriet generation.


These factors are then used to decide how your call should be handled. A
business relation\'s call might be routed to voice mail \"after hours\"
(because I don\'t think they deserve to cut into my personal time). A
family member\'s call might prompt the attendant to contact me: \"Your
sister is on the phone. Do you want to speak with her, now?\" A neighbor\'s
call (or visit!) might warrant chasing me down -- even if I\'m in the back
yard or asleep. And, a call from my other half (assuming she\'s not home)
should go to great pains to find me IF SHE CLAIMS IT IS URGENT.

You keep going on about handling different calls differently without
explaining how you distinguish the callers. In particular you claim you are
not blocking calls based on CID but then say you do use CID along with
\"feature-sets\" which you fail to explain. Looking back for the term you
used so I could quote it I see your post has gotten longer and longer.

You want go-nogo to be based on a single criteria.

You keep talking about what I want without having any understanding of what I am saying or what the real problem is. The problem is the huge number of spam phone calls. Your idea will do literally nothing for me because I can\'t connect your little box to my cell phone.

Why can\'t you understand that?


I take a more
realistic approach and base it on many criteria. Speaker identification,
alone, requires an analysis of many non-binary criteria -- feature sets.

There are calls I get that have nothing to do with any calls I\'ve received in the past. They should not go to voicemail. They should not be intercepted by a silly voice prompting system and routed to voice mail. There\'s nothing you box does that is useful to most people.


This is common in more advanced pattern recognition problems. How do you
differentiate a written 4 from a 9? Is there some magic threshold that
you can use to make a binary choice? No, you look at many aspects of
each glyph and come up with a probabilistic assessment of each \"set of
features\" observed in the writings vs. some prior knowledge of what
those features are likely to be for those particular glyphs.

How do you know that the caller is male vs. female? Yet, you know
that \"Bob Smith\" is definitely MALE -- so, if you think the current caller
is FEmale, it\'s likely NOT Bob Smith. That aspect of the observed feature
set conflicts with the STORED feature set for Bob Smith!

What do you do about Alex Montgomery or Pat Paulson? No, don\'t tell me. You\'ll just talk about something else that isn\'t workable.


I\'m not interested in reading your book. If you want to explain simply I\'m
happy to listen, but if you just want to go on and on, that makes it very
clear to me that you have developed a very sophisticated non-solution to a
real problem.

You are obviously ignorant of the technologies involved and, thus, can\'t
wrap your head around what is currently possible (and already in use).
Google \"diarization\" (so you know WHY you need it), \"speaker recognition\",
\"speech recognition\", \"dynamic pattern recognition\", \"expert systems\",
\"speech synthesis\", etc.

I did google it and I understand what it does. But you are building a self driving car without understanding it will have to travel roads that have not been mapped.


Familiarize yourself with the technologies and the issues addressed by each.
Familiarize yourself with the \"open\" implementations for each of these (so
you understand how much is already in place). Actually BUILD some demo apps
so you can see what they can and can\'t reliably do. Be amazed!

Then, sit down and think about how you could make a similar system to do what
I\'ve described. I\'ve two colleagues that are making significant investments
in technology derived from my implementation.

Your system does nothing useful for me. In addition it will ultimately be circumvented by the spammers. As I\'ve said several times already, spammers use the sort of technology you are talking about. They already are ahead of you and will very easily circumvent your system. Cat and mouse except there are thousands of them ready to spend lots of money to defeat your system.


> If you can\'t, well... the world needs ditch diggers, too!

That\'s the sort of job that will be open for you if you can\'t understand the limitations of your approach.


Technology can solve every problem. Sometimes laws and regulations are the
right solution. In this case a simple $0.10 fee per call is not only easy
to implement, the mechanics are ALREADY built into the system. It just has
to be turned on.

When skimming I saw you mention that cell towers have large computational

That\'s your problem -- you\'re skimming and not reading.

No, that\'s not the problem. Your long-windedness is the problem and skimming is the cure. Well, a partial cure.


You\'re not interested
in understanding an implementation, just looking to poke at (incorrectly
perceived) issues. Grep my message for \"cell tower\"... didn\'t find it, did
you? :> I said \"cell phone\". And most (many) cell phones have oodles of
computational power.

OMG! That\'s even worse. The phone has enough resources to do it\'s job and does not have tons of leftover computational power for such a job. CPU speeds aren\'t even the limitation. People can\'t keep their phones charged now. What you are talking about would require crunching for every call, often not keeping up when other tasks are running and eating up the battery charge. When I do some surfing trying to keep up with the news the battery goes very quickly. What you are talking about will slam the battery in short order.


I\'ve got a rescued phone lying on the desk in front of
me that has 2G of RAM, 16G FLASH and an 8 core 1.7GHz CPU. And, the \"DAA\"
is already present in it -- along with a network connection to a conceivably
unlimited amount of \"remote\" storage AND an interface to the callee\'s \"ear\"!

capacities... which is used by the cell calls, that\'s why the capacities are
there, because they are needed. They didn\'t just toss a bunch of processors
onto a board they don\'t need.

I\'m not interested in continuing the conversation if you can\'t get to the
point.

I still say you have no useful way (by \"useful\" I mean to the rest of the
world) to distinguish spammers from anyone else, because no matter what you
do, the spammers will adapt. That\'s the reason why there is so much email
spam. Do you really think email just needs a better filter?

I\'ve not received a piece of junk email in YEARS! Nor has my other half -- and
she\'s far less \"computer literate\" than I! You obviously have been doing
something wrong!

Yes, not asking them to mail it to me so I can compost it!

--

Rick C.

+++ Get 1,000 miles of free Supercharging
+++ Tesla referral code - https://ts.la/richard11209
 
On Sunday, October 25, 2020 at 4:15:49 AM UTC-4, Ricketty C wrote:
On Sunday, October 25, 2020 at 2:26:24 AM UTC-4, Don Y wrote:
On 10/24/2020 5:36 PM, Ricketty C wrote:
On Saturday, October 24, 2020 at 6:37:01 PM UTC-4, Don Y wrote:

You either block specific numbers you recognize as spam, which is
what the NOMOROBO does, or you block EVERY CALL that isn\'t white
listed.

No, you allow whitelisted calls (to be further conditioned based on
identity, time-of-day, etc.) to be \"accepted\" (ring-through, routed
to personalized mailbox, etc.).

I didn\'t say anything different.

But you assume \"non-white\" calls are black. That\'s the mistake.

No, if you pay attention to what I write I am saying your system is fatally
flawed regardless because you are making unfounded assumptions. You are the
one trying to distinguish white/black/grey calls. This simply does not work
in a useful way for anyone other than yourself because you have designed the
system around your tastes and presumably are willing to live with the
limitations.

There are always white, grey and black calls. Assuming that you can make a
binary distinction (black vs white, reject vs. accept) based on something
as crude as a CID is naive. You don\'t know what a particular CID *means*,
even if it is unforgeable! (is the hospital calling to tell you about
your upcoming appointment? or, to request a payment? or, to see if you\'d
like to participate in their blood drive??)

[I have an \"account\" with the local university as I buy surplus equipment
from them. When they call, are they trying to inform me that I won a
particular lot at auction? Or, hoping I\'ll donate to their public television
station?? Both have happened. I\'m obviously far more interested in
the former call than the latter!]

Non-whitelisted calls you identify with other MORE RELIABLE means. In
our case, we do that by recognizing the voice of the caller (we don\'t
\"recognize\" any spammers :> )

How does your machine recognize voices?

How do all IVR systems \"recognize voices\" -- voice recognition!

How does your BANK *identify* you as the caller -- speaker
identification!

My bank identifies me by my proving intimate details of my account. I have
no idea what you are talking about now.

\"My voice is my password\" TmReg.

Companies and agencies are starting to realize that folks simply can\'t
keep up with the various artificial authentication mechanisms that they put in
place (PINs, secret questions, etc.) and are moving towards more biometric
identifiers. When I call my credit card company, they note the number from
which I\'m calling to identify me and my account. I am only forced to talk
with a human if I choose to call from some \"unregistered\" phone number.
I don\'t know what credit cards you have, but I have to give them my card number and other identifying information that no one else has. They don\'t give a crap about my phone number.

I literally don\'t know what you are talking about with \"forced\" to talk to a person. There\'s virtually no reason to call on a phone if you aren\'t going to talk. I can much more easily and usefully get the info on my account that can be handled automatically simply by using the web site. I only use the phone when I have to talk to someone.
[What if someone spoofs my CID when calling them?]

As there is no universal \"spammer voice\", there\'s no need to try to
identify the spammer. \"Voice (identity) not recognized\"

This seems really out in left field now.

If you mean YOU listen, I don\'t even need to do that. But I don\'t want
to be bothered with the calls. No ring, no messages, no monitoring the
call. That\'s what the $0.10 per call fee will do. Spammers and the
services that connect them to the phone network won\'t be able to afford
to make thousands of calls to bring in a few hundred dollars.

I don\'t listen, the voice attendant does. It decides who you are by
analyzing the feature-set that it derives (while listening to you) from
your words. It compares this to a database of feature-sets to form an
appraisal of WHO you are. It augments this with any supplemental
authenticators (CID) -- even if not authoritative -- as well as the
CONTENT of your speech and conditions of the call (time of day, etc.)

So you aren\'t going to explain what you are talking about??? Whatever..

There is a little box. The wires to the PSTN come in one end. An ethernet
connection comes in the other (bringing power with it). Inside the box is
an interface to the telco network. It allows the processor (in the box)
to detect ring cadences, go off-hook, \"listen\" to incoming audio as well
as generate outgoing audio.
That\'s what I thought. An invention for the previous century. Some huge percentage of phone numbers are cell phones. So your \"invention\" is pretty much worth less. That\'s why no one has produced it before.
It can be programmed to \"answer\" -- or not -- based on whatever criteria
I deem appropriate (I might want to block ALL incoming calls at certain times).

When it answers, it can be programmed to emit whatever I choose to say
(via synthesized voice). That \"voice prompt\" can be determined by time-of-day,
incoming CID, \"expectations\" (i.e., if I\'m *expecting* a call and, who from),
etc.

Presumably, the calling party will eventually \"make sounds\" which the box
will detect (and analyze and record).

If the caller \"talks over\" my outgoing message, then it is likely a DUMB
autodialer -- one that simply waits for the phone to be answered and
then begins its spiel. A real human -- especially one familiar with
how my phone works -- would patiently await the end of the OGM.
[assuming they don\'t know of shortcuts built into the answer process]

The caller will \"make sounds\" (typically \"speak\" -- but it could also
be a FAX machine, etc.) for some amount of time. If speech, the amount
of time before a pause reveals something about the caller (a telemarketer
or robodialer will typically not wait for the short speech burst that are
typical at the start of an interactive conversation).

Typical conversations have a formulaic beginning (think about them... from
friends/relatives/colleagues vs. businesses vs. spammers). It\'s unlikely
that the caller is going to utter \"alpha geronimo boondoggle 27\" -- so
you have a restricted vocabulary to typically process.

I can recognize likely names in those utterances -- especially because
I know which names are significant, to *me*! (if you start talking
about Art Linkletter, I\'m likely going to glaze over on you)

While listening to WHAT you are saying (speech recognition), I\'m also
listening to HOW you are saying it -- the cadence, fundamental frequency,
how you pronounce vowels (cuz I know what vowels are in each of the
RECOGNIZED WORDS that you\'ve uttered). This \"feature set\" is a key into
a database of feature sets compiled from speech samples obtained from
other calls WITH KNOWN OTHER PARTIES (regardless of whether I was caller
or callee). So, I can form an assessment of WHO you likely are
(speaker identification -- not to be confused with authentication!).

I can prod you by emitting additional utterances intended (in a live caller)
to elicit certain types of information in certain formulaic ways (NLP).
E.g., if I (\"I\" in all of this is The Box) say, \"Who are you looking to speak
with?\", I know I should expect to hear one of two names in response...
not \"The Man of The House\" or \"Your Mother/Father\" (yeah, I can answer the
phone with any voice characteristics I choose! I can actually synthesize
YOUR voice after I\'ve heard enough samples on which to build a model!)

Of course, legitimate callers who\'ve been through the routine know
what to expect and how to respond (i.e., like a normal human!). And,
each response gives me more data to refine my assessment of their
identity. If I ever need confirmation, I can utter, \"Who is calling,
please?\" and \"recognize\" the response in a dictionary of \"valid callers\".

[Each of these provides data that the AI uses to build confidence in
its assessment of the caller\'s identity AND how it should handle the
call FROM that caller.]

Once I\'m confident in the caller\'s identity, I can decide how to handle
their call in the current circumstances (\"Sorry, Don is not available\";
\"Don is on his way to your house, Bruce\"; \"What should I tell Don this
call concerns?\"; etc.)

Again, depending on caller\'s identity, time of day, mood, occupancy, etc.
the call may be \"announced\" to me (or my other half, if intended for her)
with the expectation that I\'ll respond \"put him/her through\". I can
alternately say \"take a message\" or \"tell him to call back later\" or...

Just like you\'d tell your secretary when she announces that \"Bob Smith
is on line 2\"

While I\'m talking with Bob Smith, the box is listening in and recording
the conversation. This frees me from the need to analyze the content
in real-time.

At the same time, it schedules an analysis task that will listen to
these TAGGED (caller identified) recordings. This begins with diarization
then analysis of the callee\'s speech to further refine the models that
are used to identify that caller -- as well as enhance the speech
recognition FOR that caller.

This mimics how a human user recognizes callers and \"learns\" about
them, over time (e.g., Pete always calls at 5:00PM).

Note that there is no explicit \"training\" event that I\'d have to
require callers to endure -- it\'s a \"free\" side-effect of the
implementation (by design!). And, the accumulated dataset allows
those KNOWN callers to interact with my system in other ways.

In particular, it lets very special individuals (e.g., me and mine) gain
access to subsystems that wouldn\'t be accessible to mere mortals,
remotely. So, I can give my neighbors access to the house (or parts
thereof) based on their voice and other 2FA authenticators. Again,
without having to \"train\" the system for that access.

[In my initial demo of the speaker identification system, I had
my doorbell query arriving visitors and welcomed them, by name,
based on my recognition of their individual voices from past
phone conversations.]
Wow! So much noise and so little utility. You are designing a product for the Ozzie and Harriet generation.
These factors are then used to decide how your call should be handled. A
business relation\'s call might be routed to voice mail \"after hours\"
(because I don\'t think they deserve to cut into my personal time). A
family member\'s call might prompt the attendant to contact me: \"Your
sister is on the phone. Do you want to speak with her, now?\" A neighbor\'s
call (or visit!) might warrant chasing me down -- even if I\'m in the back
yard or asleep. And, a call from my other half (assuming she\'s not home)
should go to great pains to find me IF SHE CLAIMS IT IS URGENT.

You keep going on about handling different calls differently without
explaining how you distinguish the callers. In particular you claim you are
not blocking calls based on CID but then say you do use CID along with
\"feature-sets\" which you fail to explain. Looking back for the term you
used so I could quote it I see your post has gotten longer and longer..

You want go-nogo to be based on a single criteria.
You keep talking about what I want without having any understanding of what I am saying or what the real problem is. The problem is the huge number of spam phone calls. Your idea will do literally nothing for me because I can\'t connect your little box to my cell phone.

Why can\'t you understand that?
I take a more
realistic approach and base it on many criteria. Speaker identification,
alone, requires an analysis of many non-binary criteria -- feature sets..
There are calls I get that have nothing to do with any calls I\'ve received in the past. They should not go to voicemail. They should not be intercepted by a silly voice prompting system and routed to voice mail. There\'s nothing you box does that is useful to most people.
This is common in more advanced pattern recognition problems. How do you
differentiate a written 4 from a 9? Is there some magic threshold that
you can use to make a binary choice? No, you look at many aspects of
each glyph and come up with a probabilistic assessment of each \"set of
features\" observed in the writings vs. some prior knowledge of what
those features are likely to be for those particular glyphs.

How do you know that the caller is male vs. female? Yet, you know
that \"Bob Smith\" is definitely MALE -- so, if you think the current caller
is FEmale, it\'s likely NOT Bob Smith. That aspect of the observed feature
set conflicts with the STORED feature set for Bob Smith!
What do you do about Alex Montgomery or Pat Paulson? No, don\'t tell me. You\'ll just talk about something else that isn\'t workable.
I\'m not interested in reading your book. If you want to explain simply I\'m
happy to listen, but if you just want to go on and on, that makes it very
clear to me that you have developed a very sophisticated non-solution to a
real problem.

You are obviously ignorant of the technologies involved and, thus, can\'t
wrap your head around what is currently possible (and already in use).
Google \"diarization\" (so you know WHY you need it), \"speaker recognition\",
\"speech recognition\", \"dynamic pattern recognition\", \"expert systems\",
\"speech synthesis\", etc.
I did google it and I understand what it does. But you are building a self driving car without understanding it will have to travel roads that have not been mapped.
Familiarize yourself with the technologies and the issues addressed by each.
Familiarize yourself with the \"open\" implementations for each of these (so
you understand how much is already in place). Actually BUILD some demo apps
so you can see what they can and can\'t reliably do. Be amazed!

Then, sit down and think about how you could make a similar system to do what
I\'ve described. I\'ve two colleagues that are making significant investments
in technology derived from my implementation.
Your system does nothing useful for me. In addition it will ultimately be circumvented by the spammers. As I\'ve said several times already, spammers use the sort of technology you are talking about. They already are ahead of you and will very easily circumvent your system. Cat and mouse except there are thousands of them ready to spend lots of money to defeat your system..
If you can\'t, well... the world needs ditch diggers, too!
That\'s the sort of job that will be open for you if you can\'t understand the limitations of your approach.
Technology can solve every problem. Sometimes laws and regulations are the
right solution. In this case a simple $0.10 fee per call is not only easy
to implement, the mechanics are ALREADY built into the system. It just has
to be turned on.

When skimming I saw you mention that cell towers have large computational

That\'s your problem -- you\'re skimming and not reading.
No, that\'s not the problem. Your long-windedness is the problem and skimming is the cure. Well, a partial cure.
You\'re not interested
in understanding an implementation, just looking to poke at (incorrectly
perceived) issues. Grep my message for \"cell tower\"... didn\'t find it, did
you? :> I said \"cell phone\". And most (many) cell phones have oodles of
computational power.
OMG! That\'s even worse. The phone has enough resources to do it\'s job and does not have tons of leftover computational power for such a job. CPU speeds aren\'t even the limitation. People can\'t keep their phones charged now. What you are talking about would require crunching for every call, often not keeping up when other tasks are running and eating up the battery charge.. When I do some surfing trying to keep up with the news the battery goes very quickly. What you are talking about will slam the battery in short order.
I\'ve got a rescued phone lying on the desk in front of
me that has 2G of RAM, 16G FLASH and an 8 core 1.7GHz CPU. And, the \"DAA\"
is already present in it -- along with a network connection to a conceivably
unlimited amount of \"remote\" storage AND an interface to the callee\'s \"ear\"!

capacities... which is used by the cell calls, that\'s why the capacities are
there, because they are needed. They didn\'t just toss a bunch of processors
onto a board they don\'t need.

I\'m not interested in continuing the conversation if you can\'t get to the
point.

I still say you have no useful way (by \"useful\" I mean to the rest of the
world) to distinguish spammers from anyone else, because no matter what you
do, the spammers will adapt. That\'s the reason why there is so much email
spam. Do you really think email just needs a better filter?

I\'ve not received a piece of junk email in YEARS! Nor has my other half -- and
she\'s far less \"computer literate\" than I! You obviously have been doing
something wrong!
Yes, not asking them to mail it to me so I can compost it!
--

Rick C.

+++ Get 1,000 miles of free Supercharging
+++ Tesla referral code - https://ts.la/richard11209
 
On Thursday, October 22, 2020 at 5:26:14 PM UTC-4, Ricketty C wrote:
On Thursday, October 22, 2020 at 3:23:06 PM UTC-4, Robert Baer wrote:
Have need for such a beast; must recognize caller ID and determine if
call is from a town/city.

Wait 7 rings before pickup.

Emit loudest noise allowed by Ma Bell specs, primary frequency 3-4KC
for greatest effect; perhaps sawtooth for most grating.
Mixed with second tone about 14Hz above fundamental (understand 14Hz
is least pleasing frequency difference).

Simple-minded way of town/city determination: extract state
designation (seems 90 percent of these fake callers have state included)
and use look-up table for verification.

Second \"level\" would be: verify all capitalized letters, only one
\"word\" excluding state if any.

Anyone willing to take on such a project?

Thanks,
R. Baer
Not sure what you are talking about. Seems a very complicated way to force a dialing robot to listen to robot music. Actually, most of the robocalls I get are from out of state only because my phone number is from out of state. They even pick an exchange local to where my phone number was originally from. So how would this detect a spammer?

The only way to fix the spamming call problem is to force a per call charge of ALL calls like they do if you have the budget service from ma bell. Legislate that originating companies for all calls with US destinations, regardless of origin, pay $0.10 into a fund to fight phone spamming. The providers will make it unaffordable for spammers to robocall and 99% of spam calls will end at the low, low cost of just one thin dime for each of your calls.

--

Rick C.

- Get 1,000 miles of free Supercharging
- Tesla referral code - https://ts.la/richard11209
 
On Sunday, October 25, 2020 at 9:47:36 AM UTC-4, Charles Elias wrote:
On Thursday, October 22, 2020 at 5:26:14 PM UTC-4, Ricketty C wrote:
On Thursday, October 22, 2020 at 3:23:06 PM UTC-4, Robert Baer wrote:
Have need for such a beast; must recognize caller ID and determine if
call is from a town/city.

Wait 7 rings before pickup.

Emit loudest noise allowed by Ma Bell specs, primary frequency 3-4KC
for greatest effect; perhaps sawtooth for most grating.
Mixed with second tone about 14Hz above fundamental (understand 14Hz
is least pleasing frequency difference).

Simple-minded way of town/city determination: extract state
designation (seems 90 percent of these fake callers have state included)
and use look-up table for verification.

Second \"level\" would be: verify all capitalized letters, only one
\"word\" excluding state if any.

Anyone willing to take on such a project?

Thanks,
R. Baer
Not sure what you are talking about. Seems a very complicated way to force a dialing robot to listen to robot music. Actually, most of the robocalls I get are from out of state only because my phone number is from out of state. They even pick an exchange local to where my phone number was originally from. So how would this detect a spammer?

The only way to fix the spamming call problem is to force a per call charge of ALL calls like they do if you have the budget service from ma bell. Legislate that originating companies for all calls with US destinations, regardless of origin, pay $0.10 into a fund to fight phone spamming. The providers will make it unaffordable for spammers to robocall and 99% of spam calls will end at the low, low cost of just one thin dime for each of your calls.

--

Rick C.

- Get 1,000 miles of free Supercharging
- Tesla referral code - https://ts.la/richard112
 
On 10/22/2020 6:16 PM, Sylvia Else wrote:
On 23-Oct-20 6:22 am, Robert Baer wrote:
   Have need for such a beast; must recognize caller ID and determine
if call is from a town/city.

   Wait 7 rings before pickup.

   Emit loudest noise allowed by Ma Bell specs, primary frequency
3-4KC for greatest effect; perhaps sawtooth for most grating.
   Mixed with second tone about 14Hz above fundamental (understand
14Hz is least pleasing frequency difference).

I\'d go for retching sounds myself.

Sylvia.

I think I would use my modem to answer. Makes a nice warbling noise.
 
How does that help when the callers use randomly selected phone numbers and never the same one twice?

--

Rick C.

---+ Get 1,000 miles of free Supercharging
---+ Tesla referral code - https://ts.la/richard11209

On Sunday, October 25, 2020 at 9:47:36 AM UTC-4, Charles Elias wrote:
On Thursday, October 22, 2020 at 5:26:14 PM UTC-4, Ricketty C wrote:
On Thursday, October 22, 2020 at 3:23:06 PM UTC-4, Robert Baer wrote:
Have need for such a beast; must recognize caller ID and determine if
call is from a town/city.

Wait 7 rings before pickup.

Emit loudest noise allowed by Ma Bell specs, primary frequency 3-4KC
for greatest effect; perhaps sawtooth for most grating.
Mixed with second tone about 14Hz above fundamental (understand 14Hz
is least pleasing frequency difference).

Simple-minded way of town/city determination: extract state
designation (seems 90 percent of these fake callers have state included)
and use look-up table for verification.

Second \"level\" would be: verify all capitalized letters, only one
\"word\" excluding state if any.

Anyone willing to take on such a project?

Thanks,
R. Baer
Not sure what you are talking about. Seems a very complicated way to force a dialing robot to listen to robot music. Actually, most of the robocalls I get are from out of state only because my phone number is from out of state. They even pick an exchange local to where my phone number was originally from. So how would this detect a spammer?

The only way to fix the spamming call problem is to force a per call charge of ALL calls like they do if you have the budget service from ma bell. Legislate that originating companies for all calls with US destinations, regardless of origin, pay $0.10 into a fund to fight phone spamming. The providers will make it unaffordable for spammers to robocall and 99% of spam calls will end at the low, low cost of just one thin dime for each of your calls.

--

Rick C.

- Get 1,000 miles of free Supercharging
- Tesla referral code - https://ts.la/richard11209
 
On 2020-10-25, John S <Sophi.2@invalid.org> wrote:
On 10/22/2020 6:16 PM, Sylvia Else wrote:
On 23-Oct-20 6:22 am, Robert Baer wrote:
   Have need for such a beast; must recognize caller ID and determine
if call is from a town/city.

   Wait 7 rings before pickup.

   Emit loudest noise allowed by Ma Bell specs, primary frequency
3-4KC for greatest effect; perhaps sawtooth for most grating.
   Mixed with second tone about 14Hz above fundamental (understand
14Hz is least pleasing frequency difference).

I\'d go for retching sounds myself.

Sylvia.

I think I would use my modem to answer. Makes a nice warbling noise.

I used that tactic once before against a nuissance caller and actually
got a connection, it turned out that some prankster person had
announced the victims number as a BBS.


--
Jasen.
 
Don Y wrote:
On 10/23/2020 8:21 PM, Ricketty C wrote:
Caller ID tells you nothing about this callers.  Nothing.

Caller ID is useful for whitelisting calls as it\'s unlikely that
a spammer will know the phone number of your MD, dentist, friend(s),
etc.  Less useful for things like the local police department (but,
how often do THEY call you???) which could be faked.

Caller ID can tell you the caller is someone you WANT to talk to
but can\'t tell you the caller is someone you DON\'T want to talk to!
* Not true; READ what i wrote; over 90 percent are FAKE callers and have
a state code (WA, NY, TX, etc) after a city name.

So, *blocking* based on CID (unless you\'re really pissed off at a
friend/neighbor) is useless.
* READ again.

Note that you can also add heuristics to further bias your
assessment of an incoming call; e.g., if \"around dinner time\" it
is likely a spammer/pollster/etc.  We only allow calls through
\"after hours\" if they are from folks who deserve \"special access\"
to us (e.g., calling to ask for a ride home from a club or a
ride to the hospital, etc.).  I certainly don\'t expect to hear
from my MD *much* after dinner time (though a 6PM call is possible
if he\'s returning MY call).
 
Martin Brown wrote:
On 24/10/2020 03:51, Robert Baer wrote:
Martin Brown wrote:
On 22/10/2020 20:22, Robert Baer wrote:
   Have need for such a beast; must recognize caller ID and
determine if call is from a town/city.

   Wait 7 rings before pickup.

   Emit loudest noise allowed by Ma Bell specs, primary frequency
3-4KC for greatest effect; perhaps sawtooth for most grating.
   Mixed with second tone about 14Hz above fundamental (understand
14Hz is least pleasing frequency difference).

Why on earth would you want to do this?
Robotic dialers don\'t care what noises you make at them!

   Simple-minded way of town/city determination: extract state
designation (seems 90 percent of these fake callers have state
included) and use look-up table for verification.

There are plenty of existing products that will block fake calls
based on CLID or number withheld
* Please name some..

https://www.ebay.co.uk/i/233615867828
That is a start..especially if it can be programmed to block on basis
of caller id as i mentioned.
A blacklist is almost useless as majority of fake calls are new.
 
Ricketty C wrote:
On Saturday, October 24, 2020 at 2:08:42 AM UTC-4, Don Y wrote:
On 10/23/2020 8:21 PM, Ricketty C wrote:
Caller ID tells you nothing about this callers. Nothing.

Caller ID is useful for whitelisting calls as it\'s unlikely that
a spammer will know the phone number of your MD, dentist, friend(s),
etc. Less useful for things like the local police department (but,
how often do THEY call you???) which could be faked.

Caller ID can tell you the caller is someone you WANT to talk to
but can\'t tell you the caller is someone you DON\'T want to talk to!

So, *blocking* based on CID (unless you\'re really pissed off at a
friend/neighbor) is useless.

You either block specific numbers you recognize as spam, which is what the NOMOROBO does, or you block EVERY CALL that isn\'t white listed. Either way it is blocking based on CID and useless. My entering the number of every contact I\'ve ever had is not at all practical. I had to talk to police recently. There are many others who call me first and I don\'t always want to make them use voice mail. Then there is the issue of the robocalls leaving voice mail which I get all the time.

Caller ID is not a useful basis for blocking spam calls. My phone already has it through the phone company and it only flags a fraction of the calls.
* For what i see, it is extremely useful via a machine i specified.

Note that you can also add heuristics to further bias your
assessment of an incoming call; e.g., if \"around dinner time\" it
is likely a spammer/pollster/etc.
* Nope.

Bull. I get spam calls all day long.


We only allow calls through
\"after hours\" if they are from folks who deserve \"special access\"
to us (e.g., calling to ask for a ride home from a club or a
ride to the hospital, etc.). I certainly don\'t expect to hear
from my MD *much* after dinner time (though a 6PM call is possible
if he\'s returning MY call).

I\'m glad that you don\'t mind listening to your phone right 20 times a day from spammers. The NOMOROBO doesn\'t even work with most phone services, so it\'s really just a joke.
 
On 10/25/2020 1:15 AM, Ricketty C wrote:
Companies and agencies are starting to realize that folks simply can\'t
keep up with the various artificial authentication mechanisms that they
put in place (PINs, secret questions, etc.) and are moving towards more
biometric identifiers. When I call my credit card company, they note the
number from which I\'m calling to identify me and my account. I am only
forced to talk with a human if I choose to call from some \"unregistered\"
phone number.

I don\'t know what credit cards you have, but I have to give them my card
number and other identifying information that no one else has. They don\'t
give a crap about my phone number.

*ALL* of my credit cards use ANI to verify my identity. I once tried to
\"activate\" a card from a different phone and the call was routed to a human
attendant (flagged as a problem in activation): \"You\'re not calling from your
HOME PHONE!\" (note the sticker on the face of the new credit card indicated
that you should call this number FROM YOUR HOME PHONE).

I literally don\'t know what you are talking about with \"forced\" to talk to a
person. There\'s virtually no reason to call on a phone if you aren\'t going
to talk. I can much more easily and usefully get the info on my account
that can be handled automatically simply by using the web site. I only use
the phone when I have to talk to someone.

We do nothing \"electronically\". Never a fear of malware on a machine
harvesting our personal data. No \"accounts\" that we access on-line -- we
rarely need to interact with any of these people so why make it easier for
information to be lost/stolen/harvested.

The new AMEX I received last week gave me two options to activate:
- download our app
- go online (and you;ll be prompted to set up an online account, access
your statements on-line -- to save them postage fees, etc.)

\"No thanks. I\'ll telephone customer service. When asked to \"say what
you want, in a few words\", say \"activation\". Ignore their desire to
set up an account for you right then and there. Wait for the \"your
card is activated. Please destroy your old card\" message. Done.

If you mean YOU listen, I don\'t even need to do that. But I don\'t
want to be bothered with the calls. No ring, no messages, no
monitoring the call. That\'s what the $0.10 per call fee will do.
Spammers and the services that connect them to the phone network
won\'t be able to afford to make thousands of calls to bring in a few
hundred dollars.

I don\'t listen, the voice attendant does. It decides who you are by
analyzing the feature-set that it derives (while listening to you)
from your words. It compares this to a database of feature-sets to
form an appraisal of WHO you are. It augments this with any
supplemental authenticators (CID) -- even if not authoritative -- as
well as the CONTENT of your speech and conditions of the call (time of
day, etc.)

So you aren\'t going to explain what you are talking about??? Whatever.

There is a little box. The wires to the PSTN come in one end. An
ethernet connection comes in the other (bringing power with it). Inside
the box is an interface to the telco network. It allows the processor (in
the box) to detect ring cadences, go off-hook, \"listen\" to incoming audio
as well as generate outgoing audio.

That\'s what I thought. An invention for the previous century. Some huge
percentage of phone numbers are cell phones. So your \"invention\" is pretty
much worth less. That\'s why no one has produced it before.

And, as I said, you can implement my box as purely software IN your cell phone!
I need a box because I need access to the PSTN\'s \"wires\". Your cell phone
already has that built-in.

[In my initial demo of the speaker identification system, I had my
doorbell query arriving visitors and welcomed them, by name, based on my
recognition of their individual voices from past phone conversations.]

Wow! So much noise and so little utility. You are designing a product for
the Ozzie and Harriet generation.

Gee, I guess there must be lots of Ozzies and Harriets if TWO of my colleagues
are developing products based on it!

I think you suffer from a lack of imagination. But, then again. you\'re still
living with Forth...

These factors are then used to decide how your call should be handled.
A business relation\'s call might be routed to voice mail \"after
hours\" (because I don\'t think they deserve to cut into my personal
time). A family member\'s call might prompt the attendant to contact
me: \"Your sister is on the phone. Do you want to speak with her,
now?\" A neighbor\'s call (or visit!) might warrant chasing me down --
even if I\'m in the back yard or asleep. And, a call from my other half
(assuming she\'s not home) should go to great pains to find me IF SHE
CLAIMS IT IS URGENT.

You keep going on about handling different calls differently without
explaining how you distinguish the callers. In particular you claim you
are not blocking calls based on CID but then say you do use CID along
with \"feature-sets\" which you fail to explain. Looking back for the
term you used so I could quote it I see your post has gotten longer and
longer.

You want go-nogo to be based on a single criteria.

You keep talking about what I want without having any understanding of what
I am saying or what the real problem is. The problem is the huge number of
spam phone calls. Your idea will do literally nothing for me because I
can\'t connect your little box to my cell phone.

Why can\'t you understand that?

Why can\'t you understand that you can port the code TO your cell phone?
Do you think there is something magical about my \"invention from the previous
century\" that can\'t be replicated on ANOTHER device (cell phone) from the
previous century?

I take a more realistic approach and base it on many criteria. Speaker
identification, alone, requires an analysis of many non-binary criteria --
feature sets.

There are calls I get that have nothing to do with any calls I\'ve received
in the past. They should not go to voicemail. They should not be
intercepted by a silly voice prompting system and routed to voice mail.
There\'s nothing you box does that is useful to most people.

What SHOULD happen with those calls? Should they go directly to you?
How do you differentiate them from spam?

Again, I\'ve folks investing considerable monies and effort to bring devices
based on this technology to market. So, \"most people\" seems a generalization
that is purely YOUR OPINION. People putting THEIR MONEY on the line seem to
think otherwise.

What do you do about Alex Montgomery or Pat Paulson? No, don\'t tell me.
You\'ll just talk about something else that isn\'t workable.

I don;\'t expect calls from Alex Montgomery or Pat Paulson. How are they
any different from Citizens for Education Funding? Or, Macy\'s? Or,
spammers-are-us?

I\'m not interested in reading your book. If you want to explain simply
I\'m happy to listen, but if you just want to go on and on, that makes it
very clear to me that you have developed a very sophisticated
non-solution to a real problem.

You are obviously ignorant of the technologies involved and, thus, can\'t
wrap your head around what is currently possible (and already in use).
Google \"diarization\" (so you know WHY you need it), \"speaker
recognition\", \"speech recognition\", \"dynamic pattern recognition\", \"expert
systems\", \"speech synthesis\", etc.

I did google it and I understand what it does. But you are building a self
driving car without understanding it will have to travel roads that have not
been mapped.

No, you just don\'t see the map!

Simple question: do you believe a sentient entity could act to screen
your calls to weed out spammers? If so, then my point is proven.
If not, there\'s nothing you can do to block them from \"evolving\" around
whatever scheme you concoct.

Charge 10c per call? But, you know the politicians will want an exemption
(as they already have, from the Do Not Call Registry). Likewise, stores
with which you\'ve done business. Police? Fire?

We receive solicitations on the City\'s letterhead for \"Sewer Insurance\"
from a private company. Who\'s to say some determined spammer won\'t \"buy\"
access to your phone line from someone who already HAS access to it?

Any solution that relies on someone else to filter your calls will always
be subject to \"tinkering\". The only effective solution is to filter the
calls yourself.

Familiarize yourself with the technologies and the issues addressed by
each. Familiarize yourself with the \"open\" implementations for each of
these (so you understand how much is already in place). Actually BUILD
some demo apps so you can see what they can and can\'t reliably do. Be
amazed!

Then, sit down and think about how you could make a similar system to do
what I\'ve described. I\'ve two colleagues that are making significant
investments in technology derived from my implementation.

Your system does nothing useful for me. In addition it will ultimately be
circumvented by the spammers. As I\'ve said several times already, spammers
use the sort of technology you are talking about. They already are ahead of
you and will very easily circumvent your system. Cat and mouse except there
are thousands of them ready to spend lots of money to defeat your system.

They\'ve not, thus far. My logs show more than 8000 calls have been blocked
since I installed the proof-of-concept device. Over 600 in the last month
alone! (wanna bet those are all political groups and pollsters trying to
influence swing-state outcome? or, folks advocating for -- and against -- the
increased funding for education? MJ legalization? other local initiatives??

If you can\'t, well... the world needs ditch diggers, too!

That\'s the sort of job that will be open for you if you can\'t understand the
limitations of your approach.

I\'m already \"set for life\". Now working on setting up a foundation
to pursue issues of importance to me after I\'m gone.

You\'re not interested in understanding an implementation, just looking to
poke at (incorrectly perceived) issues. Grep my message for \"cell
tower\"... didn\'t find it, did you? :> I said \"cell phone\". And most
(many) cell phones have oodles of computational power.

OMG! That\'s even worse. The phone has enough resources to do it\'s job and
does not have tons of leftover computational power for such a job. CPU

Boy, you REALLY are clueless!

By your statement, this 8 core, 1.7GHz, 2G phone has \"does not have tons
of leftover computational power for such a job\". I guess facetime takes
zero resources (cuz you\'re not streaming bidirectional video and audio
when the call comes *in*!)

Let\'s assume you are 100.0% correct!

Then, how does this iPhone 6S with a 2 core, 1.8GHz processor ever manage
to receive calls? Is there an equivalent amount of computational power
in those 2 cores as in the first phone\'s *8* cores? If the nature of
\"answering a call\" is constrainedd to using a single core (implementation
issues), then why can\'t I use the other cores for my code?

And, if these TWO phones are just barely capable of performing the required
tasks \"without leftover computational power\", then how is this 600MHz 1 core
phone EVER able to act as a telephone?

Or, my other half\'s 10 year old, disposable \"feature phone\"? I guess 10 year
old processors were much better than this fancy multicore stuff -- they don\'t
make CPUs like they used to!

speeds aren\'t even the limitation. People can\'t keep their phones charged
now. What you are talking about would require crunching for every call,
often not keeping up when other tasks are running and eating up the battery
charge. When I do some surfing trying to keep up with the news the battery
goes very quickly. What you are talking about will slam the battery in
short order.

And you KNOW this, how? Have you even looked at the code required to do this?
Gee, a few messages back, you didn\'t even understand the concept of a \"feature
set\" in regard to speaker recognition. Now, you claim to know how much
resources (battery and MIPS) it requires to do that processing!

You\'ve also chosen to ignore the details I\'ve laid out here. Namely, that
you can use whitelisting to ACCEPT calls! Or, is THAT too much of a drain
on battery, as well?

You clearly are spouting personal opinion without any basis in fact.
Go get an education in the technologies involved. You don\'t even have
to implement a real product with them! Just build their demos and
poke at them. Notice how many CPU cycles are consumed for various
types of actions. Extrapolate that to battery life.

If you conclude that YOU can\'t solve this problem as I\'ve outlined,
then you can humbly admit that it\'s too much for your capabilities.
Leave it to more qualified folks to suss out.

When you actually have some basic knowledge, I *might* consider
engaging in an intellectual conversation. Until then, I\'m talking to
an opinionated dullard who just states what he WANTS to be true.

Bye!

<plonk>

I\'ve got a rescued phone lying on the desk in front of me that has 2G of
RAM, 16G FLASH and an 8 core 1.7GHz CPU. And, the \"DAA\" is already
present in it -- along with a network connection to a conceivably
unlimited amount of \"remote\" storage AND an interface to the callee\'s
\"ear\"!

capacities... which is used by the cell calls, that\'s why the capacities
are there, because they are needed. They didn\'t just toss a bunch of
processors onto a board they don\'t need.

I\'m not interested in continuing the conversation if you can\'t get to
the point.

I still say you have no useful way (by \"useful\" I mean to the rest of
the world) to distinguish spammers from anyone else, because no matter
what you do, the spammers will adapt. That\'s the reason why there is so
much email spam. Do you really think email just needs a better filter?

I\'ve not received a piece of junk email in YEARS! Nor has my other half
-- and she\'s far less \"computer literate\" than I! You obviously have been
doing something wrong!

Yes, not asking them to mail it to me so I can compost it!
 
On Monday, October 26, 2020 at 1:25:24 AM UTC-4, Don Y wrote:
On 10/25/2020 1:15 AM, Ricketty C wrote:
Companies and agencies are starting to realize that folks simply can\'t
keep up with the various artificial authentication mechanisms that they
put in place (PINs, secret questions, etc.) and are moving towards more
biometric identifiers. When I call my credit card company, they note the
number from which I\'m calling to identify me and my account. I am only
forced to talk with a human if I choose to call from some \"unregistered\"
phone number.

I don\'t know what credit cards you have, but I have to give them my card
number and other identifying information that no one else has. They don\'t
give a crap about my phone number.

*ALL* of my credit cards use ANI to verify my identity. I once tried to
\"activate\" a card from a different phone and the call was routed to a human
attendant (flagged as a problem in activation): \"You\'re not calling from your
HOME PHONE!\" (note the sticker on the face of the new credit card indicated
that you should call this number FROM YOUR HOME PHONE).

I can\'t believe you are so illiterate. They don\'t tell you to call from your home phone because few have home phones these days. We have cell phones.. Yes, calling from your phone number helps, but not because ANI or CID are not easy to fake, simply because someone stealing your credit card is not likely to know your phone number.

That\'s the problem. You are thinking about the situation as if you had gone to sleep 20 years ago and just woke up. You don\'t seem to understand that times have changed. Yesterday\'s problems are not today\'s problems and yesterday\'s solutions won\'t work for most people.


I literally don\'t know what you are talking about with \"forced\" to talk to a
person. There\'s virtually no reason to call on a phone if you aren\'t going
to talk. I can much more easily and usefully get the info on my account
that can be handled automatically simply by using the web site. I only use
the phone when I have to talk to someone.

We do nothing \"electronically\". Never a fear of malware on a machine
harvesting our personal data. No \"accounts\" that we access on-line -- we
rarely need to interact with any of these people so why make it easier for
information to be lost/stolen/harvested.

See, you are making my point for me. You are designing a system that works for you and very few others.


The new AMEX I received last week gave me two options to activate:
- download our app
- go online (and you;ll be prompted to set up an online account, access
your statements on-line -- to save them postage fees, etc.)

No credit card company requires you to access statements online, it\'s available if you wish to use it. You can have both, online and paper access which is what I do. Only the electric company and one other bill I have does not allow access to PDF bills unless I forgo paper bills. So each month I send them an email through their web site asking them to email me a PDF which they do at considerable expense compared to just giving me access.


\"No thanks. I\'ll telephone customer service. When asked to \"say what
you want, in a few words\", say \"activation\". Ignore their desire to
set up an account for you right then and there. Wait for the \"your
card is activated. Please destroy your old card\" message. Done.

I have a 94 year old friend who thinks like you do and also has a home phone... VOIP through the cable company. So you are in good company, just not a large group. In fact, a very teeny tiny group.

That\'s what you don\'t get. Your ideas serve nearly no one but yourself. You are not a typical user. If it works for you, fine. Don\'t kid yourself into thinking it will work for others.


If you mean YOU listen, I don\'t even need to do that. But I don\'t
want to be bothered with the calls. No ring, no messages, no
monitoring the call. That\'s what the $0.10 per call fee will do.
Spammers and the services that connect them to the phone network
won\'t be able to afford to make thousands of calls to bring in a few
hundred dollars.

I don\'t listen, the voice attendant does. It decides who you are by
analyzing the feature-set that it derives (while listening to you)
from your words. It compares this to a database of feature-sets to
form an appraisal of WHO you are. It augments this with any
supplemental authenticators (CID) -- even if not authoritative -- as
well as the CONTENT of your speech and conditions of the call (time of
day, etc.)

So you aren\'t going to explain what you are talking about??? Whatever.

There is a little box. The wires to the PSTN come in one end. An
ethernet connection comes in the other (bringing power with it). Inside
the box is an interface to the telco network. It allows the processor (in
the box) to detect ring cadences, go off-hook, \"listen\" to incoming audio
as well as generate outgoing audio.

That\'s what I thought. An invention for the previous century. Some huge
percentage of phone numbers are cell phones. So your \"invention\" is pretty
much worth less. That\'s why no one has produced it before.

And, as I said, you can implement my box as purely software IN your cell phone!
I need a box because I need access to the PSTN\'s \"wires\". Your cell phone
already has that built-in.

I\'m willing to bet you can\'t implement it in a cell phone. You are only thinking of MIPS and don\'t understand the nature of an invasive app like yours. Give it a try and see if you can get access to the features you need. It will be an interesting exercise.


[In my initial demo of the speaker identification system, I had my
doorbell query arriving visitors and welcomed them, by name, based on my
recognition of their individual voices from past phone conversations.]

Wow! So much noise and so little utility. You are designing a product for
the Ozzie and Harriet generation.

Gee, I guess there must be lots of Ozzies and Harriets if TWO of my colleagues
are developing products based on it!

You should be aware of your target audience. You also need to be aware of the limitations of the product.


I think you suffer from a lack of imagination. But, then again. you\'re still
living with Forth...

I don\'t need to defend Forth. You, however, are clearly stuck in the rut of land lines and thinking the problem you want to solve won\'t change with the nature of your solution. I\'ve already mentioned several times about the AI like robocall I\'ve received many times that you won\'t be able to block. Your AI has to be better than their AI and they get to work on it continuously on thousands and even millions of calls. Will you be able to match that? I don\'t think so.


These factors are then used to decide how your call should be handled.
A business relation\'s call might be routed to voice mail \"after
hours\" (because I don\'t think they deserve to cut into my personal
time). A family member\'s call might prompt the attendant to contact
me: \"Your sister is on the phone. Do you want to speak with her,
now?\" A neighbor\'s call (or visit!) might warrant chasing me down --
even if I\'m in the back yard or asleep. And, a call from my other half
(assuming she\'s not home) should go to great pains to find me IF SHE
CLAIMS IT IS URGENT.

You keep going on about handling different calls differently without
explaining how you distinguish the callers. In particular you claim you
are not blocking calls based on CID but then say you do use CID along
with \"feature-sets\" which you fail to explain. Looking back for the
term you used so I could quote it I see your post has gotten longer and
longer.

You want go-nogo to be based on a single criteria.

You keep talking about what I want without having any understanding of what
I am saying or what the real problem is. The problem is the huge number of
spam phone calls. Your idea will do literally nothing for me because I
can\'t connect your little box to my cell phone.

Why can\'t you understand that?

Why can\'t you understand that you can port the code TO your cell phone?
Do you think there is something magical about my \"invention from the previous
century\" that can\'t be replicated on ANOTHER device (cell phone) from the
previous century?

I have no reason to believe you can port your code to the cell phone. Cell phones are not open architectures and phone companies are not open access networks.


I take a more realistic approach and base it on many criteria. Speaker
identification, alone, requires an analysis of many non-binary criteria --
feature sets.

There are calls I get that have nothing to do with any calls I\'ve received
in the past. They should not go to voicemail. They should not be
intercepted by a silly voice prompting system and routed to voice mail.
There\'s nothing you box does that is useful to most people.

What SHOULD happen with those calls? Should they go directly to you?
How do you differentiate them from spam?

Yes, if someone from my insurance company calls to say there\'s a problem with my payment, I want to receive that call. What phone number will they be calling from, I have no idea, they have bazillions of numbers. I received a call not too long ago from my bank in another state. That one I missed and didn\'t recognize the number and had a hard time figuring out if they were for real or not. It finally \"sort of\" showed up in an online search and I realized it was valid, just not one of the regular offices. If a human has a hard time telling how is your tool going to figure it out? A tool like this doesn\'t need to screw up very often before it causes some bad problem.

Someone in your family is injured and in a hospital and you get a call from some number there. Again, it\'s not a call you want obstructed. Many callers you want to talk to won\'t even bother with such an ungainly system.

You can block the stupid, recorded message robo calls. That might help at least for a while. Some of those callers won\'t fork out the money to upgrade to a better system. But many will and your system won\'t stop them. If the phone rings or even if they leave a message you have lost because now I have to deal with it. A recording is actually worse than a call. I can recognize a crap call in a couple of seconds. The recording takes some time just to get into.


Again, I\'ve folks investing considerable monies and effort to bring devices
based on this technology to market. So, \"most people\" seems a generalization
that is purely YOUR OPINION. People putting THEIR MONEY on the line seem to
think otherwise.

People put money into many products that never make it to market. I think the statistic is 90% never pay off. So investment is no indication of a \"good\" idea. What you say is \"my opinion\" is not opinion at all. You are talking about land lines and being called during diner. That\'s not even 20 years old, that\'s more like 40 years old.


What do you do about Alex Montgomery or Pat Paulson? No, don\'t tell me..
You\'ll just talk about something else that isn\'t workable.

I don;\'t expect calls from Alex Montgomery or Pat Paulson. How are they
any different from Citizens for Education Funding? Or, Macy\'s? Or,
spammers-are-us?

Do you not understand your own statements? You seem to have lost context. Go back and read your prior statements.


I\'m not interested in reading your book. If you want to explain simply
I\'m happy to listen, but if you just want to go on and on, that makes it
very clear to me that you have developed a very sophisticated
non-solution to a real problem.

You are obviously ignorant of the technologies involved and, thus, can\'t
wrap your head around what is currently possible (and already in use).
Google \"diarization\" (so you know WHY you need it), \"speaker
recognition\", \"speech recognition\", \"dynamic pattern recognition\", \"expert
systems\", \"speech synthesis\", etc.

I did google it and I understand what it does. But you are building a self
driving car without understanding it will have to travel roads that have not
been mapped.

No, you just don\'t see the map!

Simple question: do you believe a sentient entity could act to screen
your calls to weed out spammers? If so, then my point is proven.
If not, there\'s nothing you can do to block them from \"evolving\" around
whatever scheme you concoct.

That\'s my point although \"any scheme\" is not accurate. I can prevent all spam phone calls by not having a phone.


Charge 10c per call? But, you know the politicians will want an exemption
(as they already have, from the Do Not Call Registry). Likewise, stores
with which you\'ve done business. Police? Fire?

Ok, so expempt the government from paying the fee. How is that a problem??? I don\'t get spam from the Police.


We receive solicitations on the City\'s letterhead for \"Sewer Insurance\"
from a private company. Who\'s to say some determined spammer won\'t \"buy\"
access to your phone line from someone who already HAS access to it?

Not sure what that even means. Are you suggesting the government will sell phone service to spammers? That\'s a government that will be out of office rather quickly.


Any solution that relies on someone else to filter your calls will always
be subject to \"tinkering\". The only effective solution is to filter the
calls yourself.

That\'s a rather large leap. Mandating a charge that is pushed through the phone system on each and every call is not really subject to \"tinkering\". Here in the US we call that theft and in the end someone gets stuck with the bill. That someone will definitely take action to prevent it and that is the point. The phone company is very good at regulating their network... it\'s not open. You have to pay to be connected and I\'m not just talking about the subscriber level.

Ultimately it may require an international agreement, but I think the existing regulations provide for this. The US can impose a $0.10 fee to any calling party for the cost of connecting a call. The other phone systems will push that back to their subscribers through any intermediaries required until ultimately it either is paid by the spammers who can\'t afford it or by their service providers who also can\'t afford it and in each case go out of business rapidly. They can\'t just spring up like eBay vendors because they would be required to provide a bond or deposit so the provider doesn\'t get stuck with the bill.

Not complex, no hacking into phones, just a simple regulatory requirement and no more spam other than selling million dollar homes to billionaires, but they don\'t often call me. I do get calls from people wanting to buy my house. They are some of the more entertaining spammers. I can run them ragged getting many calls back. Once they think you are a possibility, they don\'t want to let go. They like to start off by saying they are calling to make you an offer, but then always ask me how much I want. Then it\'s all on them to stand by that statement and make me an offer. lol

The only real question is what to do with the money collected? I think it should be only used for schools. No one argues against funding schools... or maybe roads.


Familiarize yourself with the technologies and the issues addressed by
each. Familiarize yourself with the \"open\" implementations for each of
these (so you understand how much is already in place). Actually BUILD
some demo apps so you can see what they can and can\'t reliably do. Be
amazed!

Then, sit down and think about how you could make a similar system to do
what I\'ve described. I\'ve two colleagues that are making significant
investments in technology derived from my implementation.

Your system does nothing useful for me. In addition it will ultimately be
circumvented by the spammers. As I\'ve said several times already, spammers
use the sort of technology you are talking about. They already are ahead of
you and will very easily circumvent your system. Cat and mouse except there
are thousands of them ready to spend lots of money to defeat your system.

They\'ve not, thus far.

Of course not. You aren\'t even on their radar yet.

My logs show more than 8000 calls have been blocked
since I installed the proof-of-concept device.

A robodialer makes that many calls in a day!!! Actual robodialers probably operate on many calls at one time through a PBX like interface so thousands of calls per hour.

You would have to change their hit rate by 10x at least to be noticed. Then they would investigate, find your information and learn how you work. They would quickly develop methods of attacking your system by... attacking your customers repeatedly while exploring solutions.


Over 600 in the last month
alone! (wanna bet those are all political groups and pollsters trying to
influence swing-state outcome? or, folks advocating for -- and against -- the
increased funding for education? MJ legalization? other local initiatives??

Your system has some advantages. But I bet if it becomes at all common they require that you allow political election calls through! You may be required to white list their numbers and then you have a wide open door for spammers!!!


If you can\'t, well... the world needs ditch diggers, too!

That\'s the sort of job that will be open for you if you can\'t understand the
limitations of your approach.

I\'m already \"set for life\". Now working on setting up a foundation
to pursue issues of importance to me after I\'m gone.

You\'re not interested in understanding an implementation, just looking to
poke at (incorrectly perceived) issues. Grep my message for \"cell
tower\"... didn\'t find it, did you? :> I said \"cell phone\". And most
(many) cell phones have oodles of computational power.

OMG! That\'s even worse. The phone has enough resources to do it\'s job and
does not have tons of leftover computational power for such a job. CPU

Boy, you REALLY are clueless!

By your statement, this 8 core, 1.7GHz, 2G phone has \"does not have tons
of leftover computational power for such a job\". I guess facetime takes
zero resources (cuz you\'re not streaming bidirectional video and audio
when the call comes *in*!)

The point is your app has significant resource needs and that will always be on top of whatever else the user is doing! So now every time a spammer calls their other software starts dragging and hiccuping. You may think this fades into the background, but I assure you it will be recognized and reported.


Let\'s assume you are 100.0% correct!

Then, how does this iPhone 6S with a 2 core, 1.8GHz processor ever manage
to receive calls?

The phone doesn\'t stress the general processor. Don\'t you know anything about phone technology? They have dedicated resources on a phone CPU to deal with the signal processing, etc. The phone app is not a complex piece of software.


Is there an equivalent amount of computational power
in those 2 cores as in the first phone\'s *8* cores? If the nature of
\"answering a call\" is constrainedd to using a single core (implementation
issues), then why can\'t I use the other cores for my code?

You really need to learn about phones and what they do. So how will your software run on the 2 core phone? Will it only be available to people who buy $1000 phones?


And, if these TWO phones are just barely capable of performing the required
tasks \"without leftover computational power\", then how is this 600MHz 1 core
phone EVER able to act as a telephone?

Again, you are pointing out how your app won\'t be able to work. But you still haven\'t learned enough about how phones work. How many phone apps are their on phones? I mean phone apps that answer the phone and deal with all the low level phone stuff that you will need access to? What will it take to replace the phone app with your app? Have you checked to see if the phone companies will even let you do that?


Or, my other half\'s 10 year old, disposable \"feature phone\"? I guess 10 year
old processors were much better than this fancy multicore stuff -- they don\'t
make CPUs like they used to!

You keep making the same mistake confusing the GP processor with a phone. You could ask how a cordless phone makes a call. It\'s about as relevant.


speeds aren\'t even the limitation. People can\'t keep their phones charged
now. What you are talking about would require crunching for every call,
often not keeping up when other tasks are running and eating up the battery
charge. When I do some surfing trying to keep up with the news the battery
goes very quickly. What you are talking about will slam the battery in
short order.

And you KNOW this, how? Have you even looked at the code required to do this?

I know my phone loses charge much faster when on the phone than when off. Your app will essentially answer EACH AND EVERY call to find out who is there. My calls don\'t even get answered unless I feel like dueling with the spammers. So yes, your app will suck phone batteries dry much faster than without answering all those calls.


Gee, a few messages back, you didn\'t even understand the concept of a \"feature
set\" in regard to speaker recognition. Now, you claim to know how much
resources (battery and MIPS) it requires to do that processing!

That\'s because I actually understand something about phones which you seem to lack. I don\'t need to measure the MIPS of your app to know it will suck power. EVERYTHING running on the processor and especially using the phone as a phone draws extra power than the idle current when the screen is off and the RF circuits are idling and the CPU is coasting.

Do you at lease know what the term \"talk time\" means? Check out the term \"standby time\" and get back to me. No, don\'t get back to me because there will be many other obstacles you don\'t even know are there.

It\'s not so much that you can\'t get past any one of these obstacles. The problem is you don\'t even know they are there. Most inventions and startups fail because of problems they don\'t know about, not the ones they know about at the start. Often it\'s not any one problem, it\'s the sheer number of them. I\'m seeing that right now on a ventilator project. I don\'t want to run it (not that I\'m really qualified either) but they are sailing on the Titanic and they don\'t even know the icebergs are there. They keep bouncing from one iceberg to the next as they come into view. We are pretty far along but I still only give them a 25% chance of making it because they have literally no idea of the issues beyond the technical ones.


You\'ve also chosen to ignore the details I\'ve laid out here. Namely, that
you can use whitelisting to ACCEPT calls! Or, is THAT too much of a drain
on battery, as well?

I literally don\'t give a durn about your white list. Your basic premise is flawed that you can recognize spammers and they won\'t be able to adapt. They have already come a long way which is why you feel a need for this product. They aren\'t done yet and once they adapt to your product it will be reduced to a white/black list device.


You clearly are spouting personal opinion without any basis in fact.
Go get an education in the technologies involved.

That\'s your failing. The problems are not technology based. The problems are that your technology isn\'t any better than the technology the spammers can also and presently use which you have not addressed even once.


You don\'t even have
to implement a real product with them! Just build their demos and
poke at them. Notice how many CPU cycles are consumed for various
types of actions. Extrapolate that to battery life.

You are showing a complete ignorance of how phones work and then tell me I don\'t understand the technology. You need to learn about phone batter use which is the single biggest issue when designing a phone. Any app that wrecks that is doomed.


If you conclude that YOU can\'t solve this problem as I\'ve outlined,
then you can humbly admit that it\'s too much for your capabilities.
Leave it to more qualified folks to suss out.

I never said that. I said that what YOU have described will fail as a product. You also don\'t realize the hurdles that will be required to jump to get onto a phone in the way you describe. You should get a thorough understand of that before you do anything else in your development since it will potentially impact every part of your \"thing\".


When you actually have some basic knowledge, I *might* consider
engaging in an intellectual conversation. Until then, I\'m talking to
an opinionated dullard who just states what he WANTS to be true.

Yes, that is exactly how I feel.

Bye!

plonk

The typical response when someone can\'t actually make a valid argument. Sorry you are spending so much time on something that is very unlikely to succeed.

--

Rick C.

--+- Get 1,000 miles of free Supercharging
--+- Tesla referral code - https://ts.la/richard11209
 
Ricketty C wrote:
On Friday, October 23, 2020 at 9:36:57 PM UTC-4, Robert Baer wrote:
Ricketty C wrote:
On Thursday, October 22, 2020 at 6:13:33 PM UTC-4, Robert Baer wrote:
Ricketty C wrote:
On Thursday, October 22, 2020 at 3:23:06 PM UTC-4, Robert Baer wrote:
Have need for such a beast; must recognize caller ID and determine if
call is from a town/city.

Wait 7 rings before pickup.

Emit loudest noise allowed by Ma Bell specs, primary frequency 3-4KC
for greatest effect; perhaps sawtooth for most grating.
Mixed with second tone about 14Hz above fundamental (understand 14Hz
is least pleasing frequency difference).

Simple-minded way of town/city determination: extract state
designation (seems 90 percent of these fake callers have state included)
and use look-up table for verification.

Second \"level\" would be: verify all capitalized letters, only one
\"word\" excluding state if any.

Anyone willing to take on such a project?

Thanks,
R. Baer

Not sure what you are talking about. Seems a very complicated way to force a dialing robot to listen to robot music. Actually, most of the robocalls I get are from out of state only because my phone number is from out of state. They even pick an exchange local to where my phone number was originally from. So how would this detect a spammer?
* NOT music; never music, machine would emit grating NOISE.
Read what i wrote; all undesired calls are from a town/city and most
disclose state (WA, TX, etc).

Yes, and these calls are pretty much all from India. What\'s your point???
* Maybe, and maybe not. Do not care if they come from Mars.
How about that machine?

Are you actually rational? Can you form a coherent thought???

The calls are nearly all robo calls with no human to hear your noise. When you press the right button they connect you to someone from India. They use random phone numbers local to you and pay for nothing other than being connected to a phone line somewhere in the world, very possibly over the Internet before it reaches a phone network.

Caller ID tells you nothing about this callers. Nothing.
* BULLSHIT! As i said many times before, most of the calls i get have
the state code as a part of the caller ID.
BELLEVUE WA, OGDEN UT, MOUNTAIN VIEW CA, SILVERDALE CA,CASTLE ROCK
WA, ROY WA, MARYSVILLE WA, CAMAS WA, NEW YORK NY, PORT ANGELES WA, ELMA
WA, HAYWARD CA ... need i go on with examples that GIVE USEFUL INFO?

>
 
On 28/10/2020 21:40, Robert Baer wrote:
Ricketty C wrote:
On Friday, October 23, 2020 at 9:36:57 PM UTC-4, Robert Baer wrote:
Ricketty C wrote:
On Thursday, October 22, 2020 at 6:13:33 PM UTC-4, Robert Baer wrote:
Ricketty C wrote:
On Thursday, October 22, 2020 at 3:23:06 PM UTC-4, Robert Baer wrote:
Have need for such a beast; must recognize caller ID and
determine if
call is from a town/city.

      Wait 7 rings before pickup.

      Emit loudest noise allowed by Ma Bell specs, primary
frequency 3-4KC
for greatest effect; perhaps sawtooth for most grating.
      Mixed with second tone about 14Hz above fundamental
(understand 14Hz
is least pleasing frequency difference).

      Simple-minded way of town/city determination: extract state
designation (seems 90 percent of these fake callers have state
included)
and use look-up table for verification.

      Second \"level\" would be: verify all capitalized letters,
only one
\"word\" excluding state if any.

      Anyone willing to take on such a project?

       Thanks,
R. Baer

Not sure what you are talking about.  Seems a very complicated way
to force a dialing robot to listen to robot music.  Actually, most
of the robocalls I get are from out of state only because my phone
number is from out of state.  They even pick an exchange local to
where my phone number was originally from.  So how would this
detect a spammer?
* NOT music; never music, machine would emit grating NOISE.
     Read what i wrote; all undesired calls are from a town/city
and most
disclose state (WA, TX, etc).

Yes, and these calls are pretty much all from India.  What\'s your
point???
* Maybe, and maybe not. Do not care if they come from Mars.
    How about that machine?

Are you actually rational?  Can you form a coherent thought???

The calls are nearly all robo calls with no human to hear your noise.
When you press the right button they connect you to someone from
India.  They use random phone numbers local to you and pay for nothing
other than being connected to a phone line somewhere in the world,
very possibly over the Internet before it reaches a phone network.

Caller ID tells you nothing about this callers.  Nothing.

It tells you that they have the kit and technical ability to spoof CLID.

* BULLSHIT! As i said many times before, most of the calls i get have
the state code as a part of the caller ID.
  BELLEVUE WA, OGDEN UT, MOUNTAIN VIEW CA, SILVERDALE CA,CASTLE ROCK
WA, ROY WA, MARYSVILLE WA, CAMAS WA, NEW YORK NY, PORT ANGELES WA, ELMA
WA, HAYWARD CA ... need i go on with examples that GIVE USEFUL INFO?

And you still don\'t seem to understand how easily they are forged?

UK CLID is pure numeric data.
They typically fake a local number in the hope that you will pick up.

--
Regards,
Martin Brown
 

Welcome to EDABoard.com

Sponsor

Back
Top