Stephen.
The six branches of the cognitive sciences (Philosophy, psychology, linguistics, AI, neuroscience, and anthropology) have an open-ended variety of unanswered questions. That is the nature of every active branch of science. The reason why researchers in those six sciences formed the coalition called cognitive science is that cutting-edge research in each of them has strong implications and valuable results for each of the others. In fact, prominent leaders in AI were very active in founding the Cognitive Science Journal and conferences.
There is a huge amount of fundamental research about the multiplicity of very different "languages" of thought. These results are well established with solid evidence about the influences. Natural languages are valuable for communication, but they are not the best or even the most general foundation for thinking about most of the things we do in our daily lives -- or in our most complex activities.
You can't fly an airplane, drive a truck, thread a needle, paint a picture, ski down a mountain, or solve a mathematical problem if you have to talk to yourself (vocally or silently) about every detail. You might do that when you're first learning something, but not when you master the subject
Compared to those results, the writings by many prominent researchers on LLMs are naive. They know how to play with LLMs, but they don't know how to solve the very serious tasks that AI researchers have been implementing and using successfully for years. As just some examples that my colleagues and I have implemented successfully, see
https://jfsowa.com/talks/cogmem.pdf
Look at the examples in the final section (slides 44 to 64). The current LLM technology cannot even begin to meet the requirements that the VivoMind technology could implement in 2010. Nobody writing about LLMs can show how to handle those requirements by using LLMs.
And those examples are just a small sample of successful applications. Most of the others were proprietary for our customers, who did not want to have their solutions publicized. That was fundamental science applied to mission-critical applications.
John
From: "Stephen Young" <steve@electricmint.com>
Sent: 10/8/23 7:13 PM
To: ontolog-forum@googlegroups.com, Stephen Young <steve@electricmint.com>
Subject: Re: [ontolog-forum] Addendum to (Generative AI is at the top of the Hype Cycle. Is it about to crash?
John, we've known since the 50s that the right brain has a significant role in understanding language. We also know that there is a ton of neural real estate between Wernicke's and Broca's areas that must be involved in language processing. They're like the input and output layers of the 98-layer GPT model. And we call them large language models, but they also "understand" vision.
Using our limited understanding of one black box to try to justify our assessment of another black box is not going to get us anywhere.
Alex,
Thanks for the list of applications of LANGUAGE-based LLMs. It is indeed impressive. We all agree on that. But mathematics, physics, computer science, neuroscience, and all the branches of cognitive science have shown that natural languages are just one of an open-ended variety of left-brain ways of thinking. LLMs haven't scratched the surface of the methods of thinking by the right brain and the cerebellum.
The left hemisphere of the cerebral cortex has about 8 billion neurons. The right hemisphere has another 8 billion neurons that are NOT dedicated to language. And the cerebellum has about 69 billion neurons that are organized in patterns that are totally different from the cerebrum. That implies that LLMs are only addressing 10% of what is going on in the human brain. There is a lot going on in that other 90%. What kinds of processes are happening in those regions?
Science makes progress by asking QUESTIONS. The biggest question is how can you handle the open-ended range of thinking that is not based on natural languages. Ignoring that question is NOT scientific. As the saying goes, when the only tool you have is a hammer, all the world is a nail. We need more tools to handle the other 90% of the brain -- or perhaps updated and extended variations of tools that have been developed in the past 60+ years of AI and computer science.
I'll say more about these issues with more excerpts from the article I'm writing. But I appreciate your work in showing the limitations of the current LLMs.
John
John,
English LLM is the flower on the tip of the iceberg. Multilingual LLMs are also being created. The Chinese certainly train more than just English-speaking LLMs. You can see the underwater structure of the iceberg, for example, here https://huggingface.co/datasets (1).
Academic claims against inventors are possible. But you know the inventors: it works!
It's funny that before that hype LLM meant Master of Laws:-)
Alex
(1)
--
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-forum+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ontolog-forum/4b87b85e6cde491780c3115b5babe297%40bestweb.net.
--
Stephen Young
--
All contributions to this forum are covered by an open-source license.
For information about the wiki, the license, and how to subscribe or
---
You received this message because you are subscribed to the Google Groups "ontolog-forum" group.