Connectionists: An open letter to Geoffrey Hinton: A call for civilized, moderated debate

gabriele.scheler at yahoo.com gabriele.scheler at yahoo.com
Wed Feb 14 03:17:14 EST 2024


 Hybrid AI to which you refer - incorporating logic into LLMS - will not solve the lack of reference. Besides, it is fine to expect much from the future, which however is notoriously hard to predict. If the foundation model is wrong and misguided, adding parts to it may not help. Right now LLMs do not understand.
    On Wednesday, February 14, 2024 at 06:51:18 AM GMT+1, Andras Lorincz <lorincz at inf.elte.hu> wrote:  
 
 I tried to build a tangram together with GPT 4.Even if I forced it – by prompting – to do it though programming instead of using DALL-E. It has no "understanding" of   
   - the world, including neighboring relations
   - physics, including stability
Better prompting may help, of course. 
But this is not the real issue. GPT 4 is an associative system, a kind of implicit and/or procedural memory, something like the person in the Chinese Room if you wish. Cognition and thinking are related to declarative, or explicit memory, using logic, facts and rules, the recollection of factual information, and not predictive associations. 
BUT the mentioned properties can be added. Considerable effort goes into that direction, see, e.g., LINC (A Neurosymbolic Approach for Logical Reasoning by Combining Language Models with First-Order Logic Provers) and CLOMO (Counterfactual Logical Modification with Large Language Models). Soon we will not be able to prove that LLMs/GPTs are not capable of understanding. It will look like they do.
A.





------------------------------------
Andras Lorincz
Fellow of the European Association for Artificial IntelligenceDepartment of Artificial IntelligenceFaculty of Informatics
Eotvos Lorand University
Budapest, Hungary





From: Connectionists <connectionists-bounces at mailman.srv.cs.cmu.edu> on behalf of Weng, Juyang <weng at msu.edu>
Sent: Tuesday, February 13, 2024 6:36 PM
To: gary.marcus at nyu.edu <gary.marcus at nyu.edu>
Cc: connectionists at mailman.srv.cs.cmu.edu <connectionists at mailman.srv.cs.cmu.edu>
Subject: Re: Connectionists: An open letter to Geoffrey Hinton: A call for civilized, moderated debate Dear Gary,     You wrote, "LLMs do not really understand what they are saying".    Those LLMs generated text in a natural language, didn't they?    Why do you say that LLMs do not understand such text?       The truly understandable answer to this question is not as simple as youbelieve!  What you "believe" is not convincing and intuitive to many laymen and media!      That is why Jeffery Hinton can simply give you potshots without valid analysis.    Best regards,-John WengBrain-Mind Institute On Tue, Feb 13, 2024 at 12:49 AM Gary Marcus <gary.marcus at nyu.edu> wrote:
Geoff Hinton recently asked:

Which part of this is a misrepresentation: Gary Marcus believes that LLMs do not really understand what they are saying and Gary Marcus also believes that they are an existential threat.

That’s an easy one. The second, regarding existential threat. I
 do believe that LLMs do not really understand what they are saying. But I donot believe that LLMs as such pose a (literally) existential threat, nor have I ever said such a thing, not in the Senate, not in my Substack, not here, and not anywhere else. (Anyone with evidence otherwise should step forward.)
Ihave in fact said the opposite; e.g., I have said that the human species is hard to extinguish, because we are genetically and geographic diverse, capable of building vaccines, etc.  E.g. in interview with AFP, posted at TechExplore  I said that I thought the extinction threat 'overblown', https:// techxplore.com/news/2023-06- human-extinction-threat- overblown-ai.html.
My actual view is captured here ,https://garymarcus.substack. com/p/ai-risk-agi-risk. 
although a lot of the literature equates artificial intelligence risk with the risk of superintelligence or artificial general intelligence, you don’t have to be superintelligent to create serious problems. I am not worried, immediately, about “AGI risk” (the risk of superintelligent machines beyond our control), in the near term I am worried about what I will call “MAI risk”—Mediocre AI that is unreliable (a la Bing and GPT-4) but widely deployed—both in terms of the sheer number of people using it, and in terms of the access that the software has to the world. ..

Lots of ordinary humans, perhaps of above average intelligence but not necessarily genius-level, have created all kinds of problems throughout history; in many ways, the critical variable is not intelligence but power, which often caches out as access. In principle, a single idiot with the nuclear codes could destroy the world, with only a modest amount of intelligence and a surplus of ill-deserved access. …

We need to stop worrying (just) about Skynet and robots taking over the world, and think a lot more about what criminals, including terrorists, might do with LLMs, and what, if anything, we might do to stop them.
LLMs may well pose an existential threat to democracy, because (despite their limited capacity for understanding and fact checking) they are excellent mimics, and their inability to fact-check is to the advantage of bad actors that which to exploit them. 
But that doesn’t mean they are remotely likely to literally end the species. 
I hope that clarifies my view.Gary


--Juyang (John) Weng  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20240214/fb98f9d4/attachment.html>


More information about the Connectionists mailing list