A couple of weeks ago, I heard some AI guru say that
the role of AI should be as our assistants. THE role, or
A role? Anyway...

Following that, it occurred to me that I'm using stuff that
when I first thought about AI thirty years ago, I was
these benefits as what AI would give us. Well, "AI" was
not really needed -- just a minor giveaway of privacy
reached the same goal.

Wouldn't it be nice if AI would fill in my credit card number
for me when I order something? along with my address, etc.?
Yeah, Amazon knows that because it reads my cookies and
knows who I am. We don't call it AI.

Amazon has a further assisting function. Once I'm at the
site, using its search bar, typing two or three letters pops
up the suggestion of what I'm about to look for. AI sitting
here with me, sort of, but we don't call it AI.

When I login to a site, it is great that I don't need to
remember that unique password -- Firefox fills in both
the logname and the password. It is a simple file lookup,
after a bit of recognizing that the page is looking for
something.

I wasn't impressed with word-completion when I first
experienced it; now I like it. I haven't tried any of the
ChatBots for writing text. But I can imagine becoming
fairly comfortable with 'suggestions' if they weren't too
pushy.

OBaue: What should we call examples of AI when the
achievements are too tiny to for the intelligence to baffle us?

--
Rich Ulrich

On 18/04/2024 07:47, Rich Ulrich wrote:
> A couple of weeks ago, I heard some AI guru say that
> the role of AI should be as our assistants. THE role, or
> A role? Anyway...
>
> Following that, it occurred to me that I'm using stuff that
> when I first thought about AI thirty years ago, I was
> these benefits as what AI would give us. Well, "AI" was
> not really needed -- just a minor giveaway of privacy
> reached the same goal.
>
> Wouldn't it be nice if AI would fill in my credit card number
> for me when I order something?

You don't need AI for that. Any more than you need AI to open an
automatic door when you approach it.

> along with my address, etc.?
> Yeah, Amazon knows that because it reads my cookies and
> knows who I am. We don't call it AI.
>
> Amazon has a further assisting function. Once I'm at the
> site, using its search bar, typing two or three letters pops
> up the suggestion of what I'm about to look for. AI sitting
> here with me, sort of, but we don't call it AI.

Why not? It is using text-based prediction in some language context? It
is therefore making use of AI.

>
> When I login to a site, it is great that I don't need to
> remember that unique password -- Firefox fills in both
> the logname and the password. It is a simple file lookup,
> after a bit of recognizing that the page is looking for
> something.

No AI needed. Same as filling your credit card details online.

>
> I wasn't impressed with word-completion when I first
> experienced it; now I like it. I haven't tried any of the
> ChatBots for writing text. But I can imagine becoming
> fairly comfortable with 'suggestions' if they weren't too
> pushy.
>
> OBaue: What should we call examples of AI when the
> achievements are too tiny to for the intelligence to baffle us?
>

Simple algorithms.

If you want to clarify the difference between AI and a smart look-up,
you need to bear in mind the basic definition of AI;
1- It is self-learning i.e it does not need human intervention
2- It adapts ('learns') over time, as new data becomes available to it.

Do your credit card and Firefox login examples meet the above criteria?
(No, therefore not AI.)

Rich Ulrich wrote:

> Wouldn't it be nice if AI would fill in my credit card number
> for me when I order something? along with my address, etc.?
> Yeah, Amazon knows that because it reads my cookies and
> knows who I am. We don't call it AI.

I certainly don't. There's no I involved, just a mere computer look-up
in a table. The same goes for your other examples.

A pocket calculater telling me that 5*6=30 has no AI either.

--
Bertel
Kolt, Denmark

On Thu, 18 Apr 2024, at 08:58:36, occam posted:
>On 18/04/2024 07:47, Rich Ulrich wrote:
>> A couple of weeks ago, I heard some AI guru say that
>> the role of AI should be as our assistants. THE role, or
>> A role? Anyway...
>>
>> Following that, it occurred to me that I'm using stuff that
>> when I first thought about AI thirty years ago, I was
>> these benefits as what AI would give us. Well, "AI" was
>> not really needed -- just a minor giveaway of privacy
>> reached the same goal.
>>
>> Wouldn't it be nice if AI would fill in my credit card number
>> for me when I order something?
>
>You don't need AI for that. Any more than you need AI to open an
>automatic door when you approach it.
>
>> along with my address, etc.?
>> Yeah, Amazon knows that because it reads my cookies and
>> knows who I am. We don't call it AI.
>>
>> Amazon has a further assisting function. Once I'm at the
>> site, using its search bar, typing two or three letters pops
>> up the suggestion of what I'm about to look for. AI sitting
>> here with me, sort of, but we don't call it AI.
>
>Why not? It is using text-based prediction in some language context? It
>is therefore making use of AI.

A conditioned response, or simply foreseen by the programmers? It's hard
to recognise "intelligence" there.
>
>>
>> When I login to a site, it is great that I don't need to
>> remember that unique password -- Firefox fills in both
>> the logname and the password. It is a simple file lookup,
>> after a bit of recognizing that the page is looking for
>> something.
>
>No AI needed. Same as filling your credit card details online.
>
>>
>> I wasn't impressed with word-completion when I first
>> experienced it; now I like it. I haven't tried any of the
>> ChatBots for writing text. But I can imagine becoming
>> fairly comfortable with 'suggestions' if they weren't too
>> pushy.
>>
>> OBaue: What should we call examples of AI when the
>> achievements are too tiny to for the intelligence to baffle us?
>>
>
>Simple algorithms.
Or conditioned responses. I'm growing attached to that phrase.
>
>If you want to clarify the difference between AI and a smart look-up,
>you need to bear in mind the basic definition of AI;

I wouldn't call those below a definition at all. They are criteria.

>1- It is self-learning i.e it does not need human intervention
>2- It adapts ('learns') over time, as new data becomes available to it.
>
I still interpret AI as Artificial Insemination.

>Do your credit card and Firefox login examples meet the above criteria?
>(No, therefore not AI.)

--
Paul W

Rich Ulrich <rich.ulrich@comcast.net> wrote or quoted:
>I wasn't impressed with word-completion when I first
>experienced it; now I like it.

First off, you're just using the term "AI" as a fancy-schmancy
literary device (a metaphor), and you know that for real.

What you're talkin' about are just some input shortcuts, where
the human only puts in the part the computer ain't got yet,
and the rest gets filled in by the computer.

These input shortcuts have been around in computer programs forever.
They're just bein' applied and expanded to other areas these days.

Some folks say that natural language works on the same principle:
you only say what ain't already clear from the context.

Then there's also the general principle of abbreviations
without considerin' the context. That's been a thing in
natural language for a long time too.

In natural language, you can't go too ham on the abbreviations,
or else it'll mess with the understandability - you need
a lil' redundancy too. And that leads us to the field of
data compression . . .

On 18/04/2024 09:51, Paul Wolff wrote:
> On Thu, 18 Apr 2024, at 08:58:36, occam posted:
>> On 18/04/2024 07:47, Rich Ulrich wrote:
>>> A couple of weeks ago, I heard some AI guru say that
>>> the role of AI should be as our assistants.  THE role, or
>>> A role?  Anyway...
>>>
>>> Following that, it occurred to me that I'm using stuff that
>>> when I first thought about AI thirty years ago,  I was
>>> these benefits as what AI would give us. Well, "AI" was
>>> not really needed -- just a minor giveaway of privacy
>>> reached the same goal.
>>>
>>> Wouldn't it be nice if AI would fill in my credit card number
>>> for me when I order something?
>>
>> You don't need AI for that. Any more than you need AI to open an
>> automatic door when you approach it.
>>
>>> along with my address, etc.?
>>> Yeah, Amazon knows that because it reads my cookies and
>>> knows who I am. We don't call it AI.
>>>
>>> Amazon has a further assisting function.  Once I'm at the
>>> site, using its search bar, typing two or three letters pops
>>> up the suggestion of what I'm about to look for.  AI sitting
>>> here with me, sort of, but we don't call it AI.
>>
>> Why not? It is using text-based prediction in some language context? It
>> is therefore making use of AI.
>
> A conditioned response, or simply foreseen by the programmers? It's hard
> to recognise "intelligence" there.

The model behind the responses is beyond 'forseen by programmers'. That
approach to machine translation (in the 60's - 70's - 80's) failed
miserably. In the case of AI, it has a conditional look up table which
is constantly being updated *without* human supervision. This
'unsupervised learning' is the strength (and sometime failure) of
current AIs.

>>
>>>
>>> When I login to a site, it is great that I don't need to
>>> remember that unique password -- Firefox fills in both
>>> the logname and the password. It is a simple file lookup,
>>> after a bit of recognizing that the page is looking for
>>> something.
>>
>> No AI needed. Same as filling your credit card details online.
>>
>>>
>>> I wasn't impressed with word-completion when I first
>>> experienced it; now I like it.  I haven't tried any of the
>>> ChatBots for writing text.  But I can imagine becoming
>>> fairly comfortable with 'suggestions' if they weren't too
>>> pushy.
>>>
>>> OBaue: What should we call examples of AI when the
>>> achievements are too tiny to for the intelligence to baffle us?
>>>
>>
>> Simple algorithms.

> Or conditioned responses. I'm growing attached to that phrase.

Don't. You are at least 60 years too late. 'Conditional statements'
(e.g. IF ...THEN ...ELSE instructions) have been the mainstay of
programming languages forever.

>>
>> If you want to clarify the difference between AI and a smart look-up,
>> you need to bear in mind the basic definition of AI;
>
> I wouldn't call those below a definition at all. They are criteri>
>> 1- It is self-learning i.e it does not need human intervention
>> 2- It adapts ('learns') over time, as new data becomes available to it.
>>
> I still interpret AI as Artificial Insemination.
>
>> Do your credit card and Firefox login examples meet the above criteria?
>> (No, therefore not AI.)
>

On 2024-04-18 09:42:40 +0000, occam said:

> On 18/04/2024 09:51, Paul Wolff wrote:
>> On Thu, 18 Apr 2024, at 08:58:36, occam posted:
>>> On 18/04/2024 07:47, Rich Ulrich wrote:
>>>> A couple of weeks ago, I heard some AI guru say that
>>>> the role of AI should be as our assistants.  THE role, or
>>>> A role?  Anyway...
>>>>
>>>> Following that, it occurred to me that I'm using stuff that
>>>> when I first thought about AI thirty years ago,  I was
>>>> these benefits as what AI would give us. Well, "AI" was
>>>> not really needed -- just a minor giveaway of privacy
>>>> reached the same goal.
>>>>
>>>> Wouldn't it be nice if AI would fill in my credit card number
>>>> for me when I order something?
>>>
>>> You don't need AI for that. Any more than you need AI to open an
>>> automatic door when you approach it.
>>>
>>>> along with my address, etc.?
>>>> Yeah, Amazon knows that because it reads my cookies and
>>>> knows who I am. We don't call it AI.
>>>>
>>>> Amazon has a further assisting function.  Once I'm at the
>>>> site, using its search bar, typing two or three letters pops
>>>> up the suggestion of what I'm about to look for.  AI sitting
>>>> here with me, sort of, but we don't call it AI.
>>>
>>> Why not? It is using text-based prediction in some language context? It
>>> is therefore making use of AI.
>>
>> A conditioned response, or simply foreseen by the programmers? It's hard
>> to recognise "intelligence" there.
>
> The model behind the responses is beyond 'forseen by programmers'. That
> approach to machine translation (in the 60's - 70's - 80's

and well into the 21st century. When I first tried GoogleTranslate to
or from French or Spanish, the results were abysmal. That was about 15
years ago, I think, but now they are quite good. Curiously, at the same
time as it couldn't manage French or Spanish it did a remarkably good
job of Hausa. Not that I can read Hausa, but I can recognize an
intelligible English sentence when I see one.

> ) failed
> miserably. In the case of AI, it has a conditional look up table which
> is constantly being updated *without* human supervision. This
> 'unsupervised learning' is the strength (and sometime failure) of
> current AIs.
>
>>>
>>>>
>>>> When I login to a site, it is great that I don't need to
>>>> remember that unique password -- Firefox fills in both
>>>> the logname and the password. It is a simple file lookup,
>>>> after a bit of recognizing that the page is looking for
>>>> something.
>>>
>>> No AI needed. Same as filling your credit card details online.
>>>
>>>>
>>>> I wasn't impressed with word-completion when I first
>>>> experienced it; now I like it.  I haven't tried any of the
>>>> ChatBots for writing text.  But I can imagine becoming
>>>> fairly comfortable with 'suggestions' if they weren't too
>>>> pushy.
>>>>
>>>> OBaue: What should we call examples of AI when the
>>>> achievements are too tiny to for the intelligence to baffle us?
>>>>
>>>
>>> Simple algorithms.
>
>> Or conditioned responses. I'm growing attached to that phrase.
>
> Don't. You are at least 60 years too late. 'Conditional statements'
> (e.g. IF ...THEN ...ELSE instructions) have been the mainstay of
> programming languages forever.
>
>>>
>>> If you want to clarify the difference between AI and a smart look-up,
>>> you need to bear in mind the basic definition of AI;
>>
>> I wouldn't call those below a definition at all. They are criteri>
>>> 1- It is self-learning i.e it does not need human intervention
>>> 2- It adapts ('learns') over time, as new data becomes available to it.
>>>
>> I still interpret AI as Artificial Insemination.
>>
>>> Do your credit card and Firefox login examples meet the above criteria?
>>> (No, therefore not AI.)

--
Athel -- French and British, living in Marseilles for 37 years; mainly
in England until 1987.

On 18/04/2024 11:57, Athel Cornish-Bowden wrote:
>>
>> The model behind the responses is beyond 'forseen by programmers'. That
>> approach to machine translation (in the 60's - 70's - 80's
>

> and well into the 21st century. When I first tried GoogleTranslate to or
> from French or Spanish, the results were abysmal. That was about 15
> years ago, I think, but now they are quite good. Curiously, at the same
> time as it couldn't manage French or Spanish it did a remarkably good
> job of Hausa. Not that I can read Hausa, but I can recognize an
> intelligible English sentence when I see one.

In the 90s I used to run a language translation technologies web site
for the EC, documenting progress in the domain, including the feeble
research attempts by the EC.

We had a light-hearted section on the web which had examples of
automatic translation failures. The funniest examples were those which
made the full cycle, e.g. English -> German -> English. It was a hoot!

On Thu, 18 Apr 2024, at 11:57:16, Athel Cornish-Bowden posted:
>On 2024-04-18 09:42:40 +0000, occam said:
>
>> On 18/04/2024 09:51, Paul Wolff wrote:
>>> On Thu, 18 Apr 2024, at 08:58:36, occam posted:
>>>> On 18/04/2024 07:47, Rich Ulrich wrote:
>>>>> A couple of weeks ago, I heard some AI guru say that
>>>>> the role of AI should be as our assistants.  THE role, or
>>>>> A role?  Anyway...
>>>>> Following that, it occurred to me that I'm using stuff that
>>>>> when I first thought about AI thirty years ago,  I was
>>>>> these benefits as what AI would give us. Well, "AI" was
>>>>> not really needed -- just a minor giveaway of privacy
>>>>> reached the same goal.
>>>>> Wouldn't it be nice if AI would fill in my credit card number
>>>>> for me when I order something?
>>>> You don't need AI for that. Any more than you need AI to open an
>>>> automatic door when you approach it.
>>>>
>>>>> along with my address, etc.?
>>>>> Yeah, Amazon knows that because it reads my cookies and
>>>>> knows who I am. We don't call it AI.
>>>>> Amazon has a further assisting function.  Once I'm at the
>>>>> site, using its search bar, typing two or three letters pops
>>>>> up the suggestion of what I'm about to look for.  AI sitting
>>>>> here with me, sort of, but we don't call it AI.
>>>> Why not? It is using text-based prediction in some language
>>>>context? It
>>>> is therefore making use of AI.
>>> A conditioned response, or simply foreseen by the programmers? It's
>>>hard
>>> to recognise "intelligence" there.
>> The model behind the responses is beyond 'forseen by programmers'.

You haven't understood what I meant by "foreseen". But I don't mind.

>> That
>> approach to machine translation (in the 60's - 70's - 80's
>
>and well into the 21st century. When I first tried GoogleTranslate to
>or from French or Spanish, the results were abysmal. That was about 15
>years ago, I think, but now they are quite good.

The European Patent Office was well ahead of the pack, I believe, in
developing machine translation, initially of patent specifications.
There exist millions of patent documents translated by expert humans
into and from all the languages in which patents are relevant. That has
been a huge resource for optimising the programming feedback, and the
motivation was huge too - the cost of all those translators. Patent
application owners also had motivation for ensuring that their original
work was properly rendered internationally. What company wants to spend
hundreds of thousands to get patents in say twenty countries that don't
say accurately what they wrote their patents to say?

>Curiously, at the same time as it couldn't manage French or Spanish it
>did a remarkably good job of Hausa. Not that I can read Hausa, but I
>can recognize an intelligible English sentence when I see one.
>
>> ) failed
>> miserably. In the case of AI, it has a conditional look up table which
>> is constantly being updated *without* human supervision. This
>> 'unsupervised learning' is the strength (and sometime failure) of
>> current AIs.
>>
>>>>
>>>>> When I login to a site, it is great that I don't need to
>>>>> remember that unique password -- Firefox fills in both
>>>>> the logname and the password. It is a simple file lookup,
>>>>> after a bit of recognizing that the page is looking for
>>>>> something.
>>>> No AI needed. Same as filling your credit card details online.
>>>>
>>>>> I wasn't impressed with word-completion when I first
>>>>> experienced it; now I like it.  I haven't tried any of the
>>>>> ChatBots for writing text.  But I can imagine becoming
>>>>> fairly comfortable with 'suggestions' if they weren't too
>>>>> pushy.
>>>>> OBaue: What should we call examples of AI when the
>>>>> achievements are too tiny to for the intelligence to baffle us?
>>>>>
>>>> Simple algorithms.
>>
>>> Or conditioned responses. I'm growing attached to that phrase.
>> Don't. You are at least 60 years too late. 'Conditional statements'
>> (e.g. IF ...THEN ...ELSE instructions) have been the mainstay of
>> programming languages forever.

I'm not speaking of conditional statements. I meant what I wrote.
--
Paul W

On 2024-04-18 11:47:54 +0000, Paul Wolff said:

> On Thu, 18 Apr 2024, at 11:57:16, Athel Cornish-Bowden posted:
>> On 2024-04-18 09:42:40 +0000, occam said:
>>
>>> On 18/04/2024 09:51, Paul Wolff wrote:
>>>> On Thu, 18 Apr 2024, at 08:58:36, occam posted:
>>>>> On 18/04/2024 07:47, Rich Ulrich wrote:
>>>>>> A couple of weeks ago, I heard some AI guru say that
>>>>>> the role of AI should be as our assistants.  THE role, or
>>>>>> A role?  Anyway...
>>>>>> Following that, it occurred to me that I'm using stuff that
>>>>>> when I first thought about AI thirty years ago,  I was
>>>>>> these benefits as what AI would give us. Well, "AI" was
>>>>>> not really needed -- just a minor giveaway of privacy
>>>>>> reached the same goal.
>>>>>> Wouldn't it be nice if AI would fill in my credit card number
>>>>>> for me when I order something?
>>>>> You don't need AI for that. Any more than you need AI to open an
>>>>> automatic door when you approach it.
>>>>>
>>>>>> along with my address, etc.?
>>>>>> Yeah, Amazon knows that because it reads my cookies and
>>>>>> knows who I am. We don't call it AI.
>>>>>> Amazon has a further assisting function.  Once I'm at the
>>>>>> site, using its search bar, typing two or three letters pops
>>>>>> up the suggestion of what I'm about to look for.  AI sitting
>>>>>> here with me, sort of, but we don't call it AI.
>>>>> Why not? It is using text-based prediction in some language context? It
>>>>> is therefore making use of AI.
>>>> A conditioned response, or simply foreseen by the programmers? It's hard
>>>> to recognise "intelligence" there.
>>> The model behind the responses is beyond 'forseen by programmers'.
>
> You haven't understood what I meant by "foreseen". But I don't mind.
>
>>> That
>>> approach to machine translation (in the 60's - 70's - 80's
>>
>> and well into the 21st century. When I first tried GoogleTranslate to
>> or from French or Spanish, the results were abysmal. That was about 15
>> years ago, I think, but now they are quite good.
>
> The European Patent Office was well ahead of the pack, I believe, in
> developing machine translation, initially of patent specifications.
> There exist millions of patent documents translated by expert humans
> into and from all the languages in which patents are relevant. That has
> been a huge resource for optimising the programming feedback, and the
> motivation was huge too - the cost of all those translators. Patent
> application owners also had motivation for ensuring that their original
> work was properly rendered internationally. What company wants to spend
> hundreds of thousands to get patents in say twenty countries that don't
> say accurately what they wrote their patents to say?
>
>> Curiously, at the same time as it couldn't manage French or Spanish it
>> did a remarkably good job of Hausa. Not that I can read Hausa, but I
>> can recognize an intelligible English sentence when I see one.
>>
>>> ) failed
>>> miserably. In the case of AI, it has a conditional look up table which
>>> is constantly being updated *without* human supervision. This
>>> 'unsupervised learning' is the strength (and sometime failure) of
>>> current AIs.
>>>
>>>>>
>>>>>> When I login to a site, it is great that I don't need to
>>>>>> remember that unique password -- Firefox fills in both
>>>>>> the logname and the password. It is a simple file lookup,
>>>>>> after a bit of recognizing that the page is looking for
>>>>>> something.
>>>>> No AI needed. Same as filling your credit card details online.
>>>>>
>>>>>> I wasn't impressed with word-completion when I first
>>>>>> experienced it; now I like it.  I haven't tried any of the
>>>>>> ChatBots for writing text.  But I can imagine becoming
>>>>>> fairly comfortable with 'suggestions' if they weren't too
>>>>>> pushy.
>>>>>> OBaue: What should we call examples of AI when the
>>>>>> achievements are too tiny to for the intelligence to baffle us?
>>>>>>
>>>>> Simple algorithms.
>>>
>>>> Or conditioned responses. I'm growing attached to that phrase.
>>> Don't. You are at least 60 years too late. 'Conditional statements'
>>> (e.g. IF ...THEN ...ELSE instructions) have been the mainstay of
>>> programming languages forever.

Not forever. When I learned Fortran IV in the 1960s it had arithmetic
if and logical if, but neither corresponded to if...then...else.
>
> I'm not speaking of conditional statements. I meant what I wrote.

--
Athel -- French and British, living in Marseilles for 37 years; mainly
in England until 1987.

On 18/04/2024 at 13:51, Athel Cornish-Bowden wrote:
> On 2024-04-18 11:47:54 +0000, Paul Wolff said:
>
>> On Thu, 18 Apr 2024, at 11:57:16, Athel Cornish-Bowden posted:
>>> On 2024-04-18 09:42:40 +0000, occam said:
>>>
>>>> On 18/04/2024 09:51, Paul Wolff wrote:
>>>>> On Thu, 18 Apr 2024, at 08:58:36, occam posted:
>>>>>> On 18/04/2024 07:47, Rich Ulrich wrote:
>>>>>>> A couple of weeks ago, I heard some AI guru say that
>>>>>>> the role of AI should be as our assistants. THE role, or
>>>>>>> A role? Anyway...
>>>>>>> Following that, it occurred to me that I'm using stuff that
>>>>>>> when I first thought about AI thirty years ago, I was
>>>>>>> these benefits as what AI would give us. Well, "AI" was
>>>>>>> not really needed -- just a minor giveaway of privacy
>>>>>>> reached the same goal.
>>>>>>> Wouldn't it be nice if AI would fill in my credit card number
>>>>>>> for me when I order something?
>>>>>> You don't need AI for that. Any more than you need AI to open an
>>>>>> automatic door when you approach it.
>>>>>>
>>>>>>> along with my address, etc.?
>>>>>>> Yeah, Amazon knows that because it reads my cookies and
>>>>>>> knows who I am. We don't call it AI.
>>>>>>> Amazon has a further assisting function. Once I'm at the
>>>>>>> site, using its search bar, typing two or three letters pops
>>>>>>> up the suggestion of what I'm about to look for. AI sitting
>>>>>>> here with me, sort of, but we don't call it AI.
>>>>>> Why not? It is using text-based prediction in some language
>>>>>> context? It
>>>>>> is therefore making use of AI.
>>>>> A conditioned response, or simply foreseen by the programmers? It's
>>>>> hard
>>>>> to recognise "intelligence" there.
>>>> The model behind the responses is beyond 'forseen by programmers'.
>>
>> You haven't understood what I meant by "foreseen". But I don't mind.
>>
>>>> That
>>>> approach to machine translation (in the 60's - 70's - 80's
>>>
>>> and well into the 21st century. When I first tried GoogleTranslate to
>>> or from French or Spanish, the results were abysmal. That was about
>>> 15 years ago, I think, but now they are quite good.
>>
>> The European Patent Office was well ahead of the pack, I believe, in
>> developing machine translation, initially of patent specifications.
>> There exist millions of patent documents translated by expert humans
>> into and from all the languages in which patents are relevant. That
>> has been a huge resource for optimising the programming feedback, and
>> the motivation was huge too - the cost of all those translators.
>> Patent application owners also had motivation for ensuring that their
>> original work was properly rendered internationally. What company
>> wants to spend hundreds of thousands to get patents in say twenty
>> countries that don't say accurately what they wrote their patents to say?
>>
>>> Curiously, at the same time as it couldn't manage French or Spanish
>>> it did a remarkably good job of Hausa. Not that I can read Hausa, but
>>> I can recognize an intelligible English sentence when I see one.
>>>
>>>> ) failed
>>>> miserably. In the case of AI, it has a conditional look up table which
>>>> is constantly being updated *without* human supervision. This
>>>> 'unsupervised learning' is the strength (and sometime failure) of
>>>> current AIs.
>>>>
>>>>>>
>>>>>>> When I login to a site, it is great that I don't need to
>>>>>>> remember that unique password -- Firefox fills in both
>>>>>>> the logname and the password. It is a simple file lookup,
>>>>>>> after a bit of recognizing that the page is looking for
>>>>>>> something.
>>>>>> No AI needed. Same as filling your credit card details online.
>>>>>>
>>>>>>> I wasn't impressed with word-completion when I first
>>>>>>> experienced it; now I like it. I haven't tried any of the
>>>>>>> ChatBots for writing text. But I can imagine becoming
>>>>>>> fairly comfortable with 'suggestions' if they weren't too
>>>>>>> pushy.
>>>>>>> OBaue: What should we call examples of AI when the
>>>>>>> achievements are too tiny to for the intelligence to baffle us?
>>>>>>>
>>>>>> Simple algorithms.
>>>>
>>>>> Or conditioned responses. I'm growing attached to that phrase.
>>>> Don't. You are at least 60 years too late. 'Conditional statements'
>>>> (e.g. IF ...THEN ...ELSE instructions) have been the mainstay of
>>>> programming languages forever.
>
> Not forever. When I learned Fortran IV in the 1960s it had arithmetic if
> and logical if, but neither corresponded to if...then...else.

There was a nice simple language. I was still using it in the late '80s.
We had moved to F77 by then though.

>>
>> I'm not speaking of conditional statements. I meant what I wrote.
>

--
Chris Elvidge, England
I WAS NOT TOUCHED "THERE" BY AN ANGEL

On 2024-04-18 16:00:45 +0000, Chris Elvidge said:

> On 18/04/2024 at 13:51, Athel Cornish-Bowden wrote:
>> On 2024-04-18 11:47:54 +0000, Paul Wolff said:
>>
>>>
>>> [ … ]

>>>>> Don't. You are at least 60 years too late. 'Conditional statements'
>>>>> (e.g. IF ...THEN ...ELSE instructions) have been the mainstay of
>>>>> programming languages forever.
>>
>> Not forever. When I learned Fortran IV in the 1960s it had arithmetic
>> if and logical if, but neither corresponded to if...then...else.
>
> There was a nice simple language. I was still using it in the late '80s.

Me too.

> We had moved to F77 by then though.
>
>
>>>
>>> I'm not speaking of conditional statements. I meant what I wrote.

--
Athel -- French and British, living in Marseilles for 37 years; mainly
in England until 1987.

On 18/04/2024 14:51, Athel Cornish-Bowden wrote:

<snip>

>>>>

>>>> 'Conditional statements'
>>>> (e.g. IF ...THEN ...ELSE instructions) have been the mainstay of
>>>> programming languages forever.
>
> Not forever. When I learned Fortran IV in the 1960s it had arithmetic if
> and logical if, but neither corresponded to if...then...else.

Emm...actually they are. Both logical and arithmetic 'if' statements
are of the form 'if...then...else'

If the condition is true, 'then' follows (e.g. GO to 50)

If the statement is not true, the 'else' is implicit. You continue down
the list of instructions.

On Thu, 18 Apr 2024 08:58:36 +0200, occam <occam@nowhere.nix> wrote:

>On 18/04/2024 07:47, Rich Ulrich wrote:
>> A couple of weeks ago, I heard some AI guru say that
>> the role of AI should be as our assistants. THE role, or
>> A role? Anyway...
>>
>> Following that, it occurred to me that I'm using stuff that
>> when I first thought about AI thirty years ago, I was
>> these benefits as what AI would give us. Well, "AI" was
>> not really needed -- just a minor giveaway of privacy
>> reached the same goal.
>>
>> Wouldn't it be nice if AI would fill in my credit card number
>> for me when I order something?
>
>You don't need AI for that. Any more than you need AI to open an
>automatic door when you approach it.

Well, you do need a lookup list. And there is some recognition
going on, to know the context, that a program WANTs a credit card.

>
>> along with my address, etc.?
>> Yeah, Amazon knows that because it reads my cookies and
>> knows who I am. We don't call it AI.
>>
>> Amazon has a further assisting function. Once I'm at the
>> site, using its search bar, typing two or three letters pops
>> up the suggestion of what I'm about to look for. AI sitting
>> here with me, sort of, but we don't call it AI.
>
>Why not? It is using text-based prediction in some language context? It
>is therefore making use of AI.

I'm defending lookup lists. This looks like another use of lookup,
but this time you accept it.

>
>>
>> When I login to a site, it is great that I don't need to
>> remember that unique password -- Firefox fills in both
>> the logname and the password. It is a simple file lookup,
>> after a bit of recognizing that the page is looking for
>> something.
>
>No AI needed. Same as filling your credit card details online.
>
>>
>> I wasn't impressed with word-completion when I first
>> experienced it; now I like it. I haven't tried any of the
>> ChatBots for writing text. But I can imagine becoming
>> fairly comfortable with 'suggestions' if they weren't too
>> pushy.
>>
>> OBaue: What should we call examples of AI when the
>> achievements are too tiny to for the intelligence to baffle us?
>>
>
>Simple algorithms.
>
>If you want to clarify the difference between AI and a smart look-up,
>you need to bear in mind the basic definition of AI;
>1- It is self-learning i.e it does not need human intervention
>2- It adapts ('learns') over time, as new data becomes available to it.
>
>Do your credit card and Firefox login examples meet the above criteria?
>(No, therefore not AI.)

I also see disparagement of lookup lists in a separate Reply,
On Thu, 18 Apr 2024 09:16:10 +0200, Bertel Lund Hansen
<gadekryds@lundhansen.dk> wrote:

>Rich Ulrich wrote:
>
>> Wouldn't it be nice if AI would fill in my credit card number
>> for me when I order something? along with my address, etc.?
>> Yeah, Amazon knows that because it reads my cookies and
>> knows who I am. We don't call it AI.
>
>I certainly don't. There's no I involved, just a mere computer look-up
>in a table. The same goes for your other examples.

Translation of natural language seems to be accepted as AI.

But my impression is that the BEST translation of idiomatic
expressions or sayings is done by lookup, enabled by huge
amounts of cheap storage and fast, parallel, processors.
"The spirit is willing but the flesh is weak" should be found
in its entirety.

Translation in context, I imagine, is done by scoring other
words found in what is being observed. And using huge sets
of data for comparison.

I have gained a different perspective on my long-time hope that
computers should make excellent tutors, replacing conventional
schools. The 'AI' will have tabled lookups that dictate (almost) every
response, based on huge files of interactions. The trickier part
will be the tables (and algorithms) that reduce, slightly, the
variety in what is said or done by the tutee.

What is left to be AI, if all of this is 'just table lookup'?
- there is the stuff that is generated on command, with minor
description. Texts. Pictures. But it seems to me that those bots
are using databases which are themselves 'lookup lists' with
various attributes scored and saved.

Maybe, it is a lookup list with algorithms if I can imagine how
it might be done, and it is AI if I can't.

When I was a kid and I first saw an automatic door open as
we approached (1950s), I was astounded. Magic?!

--
Rich Ulrich

On 18/04/24 22:51, Athel Cornish-Bowden wrote:
> On 2024-04-18 11:47:54 +0000, Paul Wolff said:
>> On Thu, 18 Apr 2024, at 11:57:16, Athel Cornish-Bowden posted:
>>> On 2024-04-18 09:42:40 +0000, occam said:

>>>> Don't. You are at least 60 years too late. 'Conditional
>>>> statements' (e.g. IF ...THEN ...ELSE instructions) have been
>>>> the mainstay of programming languages forever.
>
> Not forever. When I learned Fortran IV in the 1960s it had arithmetic
> if and logical if, but neither corresponded to if...then...else.
>>
>> I'm not speaking of conditional statements. I meant what I wrote.

I' sorry to say this, but some people now classify the 1960s as pre-forever.

Like a former guitar teacher of mine, who insisted that rock and roll
was invented in the 1980s.

--
Peter Moylan http://www.pmoylan.org
Newcastle, NSW

> OBaue: What should we call examples of AI when the
> achievements are too tiny to for the intelligence to baffle us?

A nervous system for interconnected machines serving as a super-human robot.
That was Arindam's last C program, back in 2007, as a contractor, before he was shifted elsewhere by the management.
Top end product for now - military drone.
So maybe, not a tiny achievement after all.

bt

Peter Moylan wrote:

> Like a former guitar teacher of mine, who insisted that rock and roll
> was invented in the 1980s.

I think that we should commemorate Rosetta Tharpe with a song from 1936,
"This Train":

https://www.youtube.com/watch?v=jOrhjgt-_Qc

--
Bertel
Kolt, Denmark

On 19/04/2024 00:27, Rich Ulrich wrote:
> On Thu, 18 Apr 2024 08:58:36 +0200, occam <occam@nowhere.nix> wrote:
>
>> On 18/04/2024 07:47, Rich Ulrich wrote:

<snip>

>>>
>>> Wouldn't it be nice if AI would fill in my credit card number
>>> for me when I order something?
>>
>> You don't need AI for that. Any more than you need AI to open an
>> automatic door when you approach it.
>
> Well, you do need a lookup list. And there is some recognition
> going on, to know the context, that a program WANTs a credit card.

The point of AI is not just the recognition (which is only a part) but
also adaptability ('learning') and adjusting to new situations.

>
>>
>>> along with my address, etc.?
>>> Yeah, Amazon knows that because it reads my cookies and
>>> knows who I am. We don't call it AI.
>>>
>>> Amazon has a further assisting function. Once I'm at the
>>> site, using its search bar, typing two or three letters pops
>>> up the suggestion of what I'm about to look for. AI sitting
>>> here with me, sort of, but we don't call it AI.
>>
>> Why not? It is using text-based prediction in some language context? It
>> is therefore making use of AI.
>
> I'm defending lookup lists. This looks like another use of lookup,
> but this time you accept it.

It *is* a look-up table, but it may well be a slightly *different*
lookup table than the last time you typed in your search term. Why?
Because the AI has modified the look up table (since your last effort)
based on your choice last time around. A fixed 'look up table' is
different to an adaptive lookup table, which adjusts itself every time
new data goes into its database.

>
>>
<snip>

>
> Translation of natural language seems to be accepted as AI.

Yes.

>
> But my impression is that the BEST translation of idiomatic
> expressions or sayings is done by lookup, enabled by huge
> amounts of cheap storage and fast, parallel, processors.
> "The spirit is willing but the flesh is weak" should be found
> in its entirety.

Yes. (No contradiction here with your statement above.)

>
> Translation in context, I imagine, is done by scoring other
> words found in what is being observed. And using huge sets
> of data for comparison.

Now ask yourself, will the 'scoring' of the other words be the same next
time you use the AI? In a truly adaptive system, the answer is NO. With
every addition of new data into the 'training' data set, the scoring
with change, albeit slightly.

It is important to not get stuck on the term 'look-up table'. You have
to ask: how often is this table revised/reviewed? And by whom? If it is
done automatically by the computer, then that is 'AI'.

On 2024-04-18 17:03:19 +0000, occam said:

> On 18/04/2024 14:51, Athel Cornish-Bowden wrote:
>
> <snip>
>
>>>>>
>
>>>>> 'Conditional statements'
>>>>> (e.g. IF ...THEN ...ELSE instructions) have been the mainstay of
>>>>> programming languages forever.
>>
>> Not forever. When I learned Fortran IV in the 1960s it had arithmetic if
>> and logical if, but neither corresponded to if...then...else.
>
>
> Emm...actually they are. Both logical and arithmetic 'if' statements
> are of the form 'if...then...else'
>
> If the condition is true, 'then' follows (e.g. GO to 50)
>
> If the statement is not true, the 'else' is implicit. You continue down
> the list of instructions.

Yes, if you go through some mental contortions you can interpret them
that way, but I'd be surprised if all early users of Fortran IV did.

Leaving the "else" implicit is a great way to confuse people.

--
Athel -- French and British, living in Marseilles for 37 years; mainly
in England until 1987.

Rich Ulrich <rich.ulrich@comcast.net> wrote or quoted:
>Maybe, it is a lookup list with algorithms if I can imagine how
>it might be done, and it is AI if I can't.

Yo, artificial intelligence has been defined for a while now
- kinda jokingly - as "the moving goalpost of what computers can't
do yet". (Can't pinpoint a single source for that definition, tbh.)

on 4/19/2024, Athel Cornish-Bowden supposed :
> On 2024-04-18 17:03:19 +0000, occam said:
>
>> On 18/04/2024 14:51, Athel Cornish-Bowden wrote:
>>
>> <snip>
>>
>>>>>>
>>
>>>>>> 'Conditional statements'
>>>>>> (e.g. IF ...THEN ...ELSE instructions) have been the mainstay of
>>>>>> programming languages forever.
>>>
>>> Not forever. When I learned Fortran IV in the 1960s it had arithmetic if
>>> and logical if, but neither corresponded to if...then...else.
>>
>>
>> Emm...actually they are. Both logical and arithmetic 'if' statements
>> are of the form 'if...then...else'
>>
>> If the condition is true, 'then' follows (e.g. GO to 50)
>>
>> If the statement is not true, the 'else' is implicit. You continue down
>> the list of instructions.
>
> Yes, if you go through some mental contortions you can interpret them that
> way, but I'd be surprised if all early users of Fortran IV did.
>
> Leaving the "else" implicit is a great way to confuse people.

I've read an author who felt that else clauses meant that you got the
logic wrong, but I've definitely had places where I couldn't figure out
a nice or tidy way to avoid an else. In Fortran, of course, you used
gotos for that.

/dps

--
We’ve learned way more than we wanted to know about the early history
of American professional basketball, like that you could have once
watched a game between teams named the Indianapolis Kautskys and the
Akron Firestone Non-Skids. -- fivethirtyeight.com

On 19/04/24 19:20, Snidely wrote:
> on 4/19/2024, Athel Cornish-Bowden supposed :
>> On 2024-04-18 17:03:19 +0000, occam said:
>>
>>> On 18/04/2024 14:51, Athel Cornish-Bowden wrote:
>>>
>>> <snip>
>>>
>>>>>>>
>>>
>>>>>>> 'Conditional statements'
>>>>>>> (e.g. IF ...THEN ...ELSE instructions) have been the mainstay of
>>>>>>> programming languages forever.
>>>>
>>>> Not forever. When I learned Fortran IV in the 1960s it had
>>>> arithmetic if
>>>> and logical if, but neither corresponded to if...then...else.
>>>
>>>
>>> Emm...actually they are. Both logical and arithmetic 'if' statements
>>> are of the form 'if...then...else'
>>>
>>> If the condition is true, 'then' follows (e.g. GO to 50)
>>>
>>> If the statement is not true, the 'else' is implicit. You continue down
>>> the list of instructions.
>>
>> Yes, if you go through some mental contortions you can interpret them
>> that way, but I'd be surprised if all early users of Fortran IV did.
>>
>> Leaving the "else" implicit is a great way to confuse people.
>
> I've read an author who felt that else clauses meant that you got the
> logic wrong, but I've definitely had places where I couldn't figure out
> a nice or tidy way to avoid an else. In Fortran, of course, you used
> gotos for that.

Unless you had a compiler that implemented COMEFROM.

--
Peter Moylan http://www.pmoylan.org
Newcastle, NSW

Fri, 19 Apr 2024 10:10:05 +0200: Athel Cornish-Bowden <me@yahoo.com>
scribeva:

>On 2024-04-18 17:03:19 +0000, occam said:
>
>> On 18/04/2024 14:51, Athel Cornish-Bowden wrote:
>>
>> <snip>
>>
>>>>>>
>>
>>>>>> 'Conditional statements'
>>>>>> (e.g. IF ...THEN ...ELSE instructions) have been the mainstay of
>>>>>> programming languages forever.
>>>
>>> Not forever. When I learned Fortran IV in the 1960s it had arithmetic if
>>> and logical if, but neither corresponded to if...then...else.
>>
>>
>> Emm...actually they are. Both logical and arithmetic 'if' statements
>> are of the form 'if...then...else'
>>
>> If the condition is true, 'then' follows (e.g. GO to 50)
>>
>> If the statement is not true, the 'else' is implicit. You continue down
>> the list of instructions.

Only true if what follows 'THEN' breaks the normal program flow, like
GO TO does. Or was it the only possibility in Fortran? I never used
the language myself, my first was Algol 68, then some PL/1, and very
little Cobol.

>Yes, if you go through some mental contortions you can interpret them
>that way, but I'd be surprised if all early users of Fortran IV did.
>
>Leaving the "else" implicit is a great way to confuse people.

--
Ruud Harmsen, https://rudhar.com

On 19/04/2024 06:02, bertietaylor wrote:

>> Rich Ulrich's details missing

>> OBaue: What should we call examples of AI when the
>> achievements are too tiny to for the intelligence to baffle us?
>

<Arindam bullshit deleted>

Learn how to reply to message correctly you ignorant chubhan. Leave the
OP references so as everyone knows who you are replying to.

Le 18/04/2024 à 06:47, Rich Ulrich a écrit :
>
> OBaue: What should we call examples of AI when the
> achievements are too tiny to for the intelligence to baffle us?

AM, artificial mediocrity, perhaps.

I think bafflement may come at different points. We understand what a
self-driving car is doing when it stays in lane or stops at lights,
while likely being unable to understand how it does it.

Then, at a higher level, if AI managed to stabilise all the world's
economies or prevent disputes escalating into war, everything it does
might be beyond our comprehension.

Perhaps self-taught AI marks the divide, the point at which humans can
no longer unravel its inner workings.