Knowledge Processing, Logic, and the Future of AI - Georg Gottlob WLD 2021

[Marco Neumann]

This might be of interest to some of you. To celebrate World Logic Day 2021  Georg Gottlob presents his take on the future of AI.

The presentation has a bit of a slow start but gets better towards the end. Nothing really new here but a sober overview where things stand.

Formal (Symbolic) Logic needs to catch up with the progress made in ML & NN research to stay relevant.

Video

https://www.youtube.com/watch?v=j-HerlYBJEw

Slides

https://www.dropbox.com/s/4p9q06p445526jh/Futute-of-AI-Gottlob%20-%20main-slides.pdf?dl=0

--

---
Marco Neumann
KONA

[Azamat Abdoullaev]

An interesting survey. A couple of points to observe:

We urgently need to“get cracking on building machines equipped with common sense, cognitive models, and powerful tools for reasoning. [...]Together [with ML] these can lead to deep understanding, itself a prerequisite for building machines that can reliably anticipate and evaluate the consequences of their actions. [...]The royal road to better AI is through AI that genuinely understands the world.”

Great! The only thing is, nobody is walking along the royal road, preferring all sorts of side roads.

 If our left hemisphere is now without NNs.

It is misleading to divide it as symbolic and subsymbolic AI. The latter is using binary data as symbols, with statistic inference, induction, deduction, transduction and abduction/transfer learning.  

ML is NOT any true AI, but a computational predictive analytics statistical technique, with no need for "understanding the world".  

What must be really done it is merging MI and HI, as HMIL.

https://cont.ws/@ashacontws/1876284 

--
All contributions to this forum are covered by an open-source license.
For information about the wiki, the license, and how to subscribe or
unsubscribe to the forum, see http://ontologforum.org/info/
---
You received this message because you are subscribed to the Google Groups "ontolog-forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ontolog-foru...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ontolog-forum/CABWJn4RAbNscqmq-kUduLitXTpd0OSqaQ%3D8-9hC-%3DQ5nLBWy2w%40mail.gmail.com.

[Marco Neumann]

I am still not a fan of how terms such as knowledge, learning, intelligence, understanding, recognition etc are used in AI interchangeably with their "original" meaning.

As Georg mentioned the dreadful boasting, the marketing mindset of some, and the money focus in the field (eg asking for lots of money and not delivering) is a real potshot in the talk.

Of course, he is himself guilty of oversimplification here as well again by ending his talk with the brain lateralisation theory analogy.

But I guess what I like about his presentation is ultimately the narrow focus on exploring formal logic as a sound foundation for computational problem-solving.

Georg doesn't seem to be too keen on the ontological reasoning approaches but the solving of existentially quantified Horn clauses. Correct me if I am wrong but that seems to be also the focus currently of the Microsoft Research Laboratory Cambridge yes?

[Azamat Abdoullaev]

Georg doesn't seem to be too keen on the ontological reasoning approaches but the solving of existentially quantified Horn clauses. Correct me if I am wrong but that seems to be also the focus currently of the Microsoft Research Laboratory Cambridge yes?

MRL does everything, computational statistics, data analysis, predictive analytics, ml, etc., except Real AI as Machine Intelligence...

https://www.microsoft.com/en-us/research/lab/microsoft-research-cambridge/publications/?#

https://www.microsoft.com/en-us/research/lab/microsoft-research-cambridge/projects/?#

--
All contributions to this forum are covered by an open-source license.
For information about the wiki, the license, and how to subscribe or
unsubscribe to the forum, see http://ontologforum.org/info/
---
You received this message because you are subscribed to the Google Groups "ontolog-forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ontolog-foru...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/ontolog-forum/CABWJn4ThXLycXEBjQpUjk20kFBaCY7px_yy9QvQ_%3DWERCm_8eQ%40mail.gmail.com.

[Ricardo Sanz]

Hi Marco,

What do you mean with " their "original" meaning"?

Regards,

Ricardo

[Marco Neumann]

Hi Ricardo,

thanks for asking. Actors in the AI field have frequently used very ambitious sounding labels for relatively mondains tasks in the past. Take for example Natural Language Processing when there is frequently only simple text processing gear at work. Or to equate knowledge representation with taxonomic data structuring and thesauri linking. Often interchangeably used for labels such as Knowledge base or knowledge graph. I think the very inflationary use of the word "knowledge" in the field starts to devoid "knowledge" of its actual meaning of what human knowledge is.Computer vision is another example in a spectrum of very rudimentary matching techniques that start to confuse how we think about visual recognition, learning and understanding.  I would still prefer to converse with a cognitive psychologist, a neurologist or even a teacher rather than an AI expert when it comes to elaborating the meaning of knowledge and information.

.

--
All contributions to this forum are covered by an open-source license.
For information about the wiki, the license, and how to subscribe or
unsubscribe to the forum, see http://ontologforum.org/info/
---
You received this message because you are subscribed to the Google Groups "ontolog-forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ontolog-foru...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/ontolog-forum/4d6ddbf6-b17e-4f9b-823e-f64ff96ef850n%40googlegroups.com.

[James Davenport]

There’s a really good description of this inflated terminology in

McDermott,D.,

Artificial Intelligence Meets Natural Stupidity.

Mind Design (ed. J. Haugeland), M.I.T. Press, 1981, pp. 143-160.

 

From: ontolo...@googlegroups.com <ontolo...@googlegroups.com> On Behalf Of Marco Neumann
Sent: 17 January 2021 10:52
To: ontolo...@googlegroups.com
Subject: Re: [ontolog-forum] Knowledge Processing, Logic, and the Future of AI - Georg Gottlob WLD 2021

 

CAUTION:  This email came from outside of the University. To keep your account safe, only click on links and open attachments if you know the person who sent the email, or you expected to receive this communication.

 

To view this discussion on the web visit https://groups.google.com/d/msgid/ontolog-forum/CABWJn4S_b00thNq69gSDGi8gSr%3DkZWvg%2BkP3sjQm--raCxE7ug%40mail.gmail.com.

[Marco Neumann]

Thank you for the reading reference James. I will certainly give it a go, the abstract is already promising:

"As a field, artificial intelligence has always been on the border of respectability, and therefore on the border of crackpottery. Many critics <Dreyfus, 1972>, <Lighthill, 1973> have urged that we are over the border. We have been very defensive toward this charge, drawing ourselves up with dignity when it is made and folding the cloak of Science about us. On the other hand, in private, we have been justifiably proud of our willingness to explore weird ideas, because pursuing them is the only way to make progress."

I have to concede that my views are not new at all here. But I am surprised it's 2021 and the criticism and the justification hasn't changed much. 

To view this discussion on the web visit https://groups.google.com/d/msgid/ontolog-forum/LO2P265MB1341B2CE602572FED67E89BBE0A50%40LO2P265MB1341.GBRP265.PROD.OUTLOOK.COM.