| alt.fool's-paradise.computers [message #410091] |
Mon, 19 July 2021 09:45  |
|
|
Originally posted by: A.T. Murray
Here in a.f.c. we live in a Fool's Paradise of believing that
our lives make a difference in the world. Thus we publish things like
Saving Private Ryan
An article by Steve Lohr in today's Sunday New York Times about IBM Watson's lack of success affords us a chance to discuss how True AI (i.e.Mentifex AI) can pass a "second grade reading comprehension test" -- which is currently beyond the power of IBM Watson's "AI".
Suppose that a mentifex-class Strong AI Mind reads in a second-grade-level story, "A boy and a girl are in the kitchen. The girl puts a plate on the table. The boy pushes the plate off the table."
To test reading comprehension, we can expect questions and answers such as the following.
Q. Where is the boy/girl/plate/table?
A. The boy/girl/plate/table is in the kitchen.
Q. Where is the plate?
A. The plate is off the table.
Q. Where is the kitchen?
A. I do not know.
The Mentifex AI Mind understands the situation of the story only in terms of the language being used, namely subject-verb-object (SVO) sentences and prepositional phrases. At first the AI knows that the plate is on the table, because the girl puts it there. Then the AI knows not exactly where the plate is, but only that it is "off the table" because of the act of "push" by the boy.
If there is no prior experience of the words involved, the AI does not understand each individual word, such as what a boy is or what a plate is. The AI makes a concept out of each word but with only a shallow conceptual knowledge. From the story alone, the AI understands each concept only in terms of the other concepts. Over time, the AI will develop deeper knowledge about each concept and about the world in general.
It seems a shame for IBM to spend so much time, money and talent on developing Watson only to see it fall into desuetude. Any entity wanting to go beyond Watson could look into the Mentifex AI mind-design and finally do Watson in the correct True AI way.
Mentifex (Arthur T. Murray)
http://ai.neocities.org/NLU.html -- Natural Language Understanding;
https://www.mail-archive.com/agi@agi.topicbox.com -- list of messages
|
|
|
|
| Re: alt.fool's-paradise.computers [message #410098 is a reply to message #410091] |
Mon, 19 July 2021 14:31  |
|
|
Originally posted by: greymaus
On 2021-07-19, A.T. Murray <mentificium@gmail.com> wrote:
> Here in a.f.c. we live in a Fool's Paradise of believing that
> Q. Where is the plate?
> A. The plate is off the table.
>
> Q. Where is the kitchen?
> A. I do not know.
>
> The Mentifex AI Mind understands the situation of the story only in terms of the language being used, namely subject-verb-object (SVO) sentences and prepositional phrases. At first the AI knows that the plate is on the table, because the girl puts it there. Then the AI knows not exactly where the plate is, but only that it is "off the table" because of the act of "push" by the boy.
>
> If there is no prior experience of the words involved, the AI does not understand each individual word, such as what a boy is or what a plate is. The AI makes a concept out of each word but with only a shallow conceptual knowledge. From the story alone, the AI understands each concept only in terms of the other concepts. Over time, the AI will develop deeper knowledge about each concept and about the world in general.
>
> It seems a shame for IBM to spend so much time, money and talent on developing Watson only to see it fall into desuetude. Any entity wanting to go beyond Watson could look into the Mentifex AI mind-design and finally do Watson in the correct True AI way.
>
> Mentifex (Arthur T. Murray)
> http://ai.neocities.org/NLU.html -- Natural Language Understanding;
> https://www.mail-archive.com/agi@agi.topicbox.com -- list of messages
Am I correct in what I have read, that the IBM machine has to be
retructured during the game, that outside the fools gambit, the possibilites in
this basically simple game are not capable of being all considered at
one time, that the possibilities (20x-> infinity) are too large, even
before the Queen gets going?. I find that comforting.
--
greymausg@mail.com
Down the wrong maushole.
|
|
|
|
Members
Search
Help
Register
Login
Home
