Connectionists: LLMs and Chomsky again

[first last]

I have just joined the connectionist mailing list after some 10 years or so (1), because I realized some people here will push the LLM agenda even among people who are knowledgeable about neural networks, or should be.
Here is an example:
His apple example sounded convincing, so I tested GPT on it:
> 
> Me:  If I hold an apple in my hand and let it go what will happen?
> 
> GPT:  The apple will fall to the ground due to the force of gravity.
> 
> Me: What about any such object?
> 
> GPT: Any object released from a person's hand will fall to the ground due to the
> force of gravity.
> 
> Me:  What would happen if there wasn't a force of gravity?
> 
> GPT:  If there were no force of gravity, objects would not be able to stay on
> the ground and would float away in all directions
> 
> Chomsky defined thinking by the correct answers to these questions.Of course the person who posted this seems slightly underpowered intellectually.  (The apple would not fall. How would it get to the ground? It needs to be on the ground before it is unable to stay there. The system did not follow the conversation, it just popped out a new stored item.) I am a former linguist myself. Now a 92 year old man understands this better than most others buying into the hype.

It is a fact that we have a system which scraped lots of text from all kinds of sources and now reproduces it. This is a text generating (text glueing) system, but certainly not a language model of any kind. And I would predict it is useless. 

I put together a few loose remarks in support of Chomsky, naming real issues, on researchgate (2). In general, it seems very clear to me that one way to build better AI is to use real neuron based models (3). Other approaches may deviate from the brain-based model altogether. LLMs are a dead-end, but they were expensive and they will be deployed regardless. I can imagine only few legitimate uses for them.
 Gabriele Scheler

-------------------------------(1) Some of you may know that I had to report Sejnowski to the FBI for repeatedly breaking into my house and drugging me. I ended up in hospital.  https://www.researchgate.net/publication/365925517_NIPS_NeurIPS_and_Neuroscience_A_personal_historical_perspective Obviously he hasn't been convicted yet, I deplore that.Chomsky himself, personally, was ready to listen to me, when I told him about the military connections of TS and his crimes. Few from the AI community have.  And yes, ethics matter in science.(2) https://www.researchgate.net/publication/366567941_Topics_for_a_biological_language_model(3) https://www.researchgate.net/publication/361118319_Sketch_of_a_novel_approach_to_a_neural_model (This paper is getting read: )




--
Dr. Gabriele Scheler
Carl Correns Foundation for Mathematical Biology1030 Judson Dr
Mountain View, Ca. 94040                                                        
http://www.theoretical-biology.org
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20230312/5c74a412/attachment-0001.html>