Skip to main content
Ask the ExpertSpiegeloog 416: Language

Ask the Expert: How can the current developments in AI deepen our understanding of human intelligence?

By February 9, 2022No Comments
Ingmar Visser
Steven Scholte
Ingmar Visser’s (Developmental Psychology) question

Dear Steven,

Artificial intelligence is once again on the rise after what is called the AI Winter. Psychology and AI are of course closely intertwined and share a common interest in understanding the mechanisms underlying intelligent behavior. How can current developments in AI deepen our understanding of human intelligence? Is there a limit here or will AI be truly intelligent this time around?

Ingmar

Steven Scholte’s (Brain and Cognition) answer

Dear Ingmar,

Thanks for this inspiring question, I have pondered on this for some weeks.

Human intelligence is a special category. It maybe is the new ‘language’. In the last century, language was often used as an argument for the manifest superiority of humans over animals.  Developments have made it clear that language is much less complex than was assumed and science has exorcised miracles like the hypothesized ‘Language acquisition device (LAD)’ (a proposed evolutionary derived black box system enabling humans to learn such a complex system). Also the hard boundary between humans and other great apes has turned out to be much more gradual. The miracle of language has not been solved, it has evaporated. Miracles are the domain of people like patriarch Krill, not science (and notice how hilarious it is that the big conclusion of the LAD was that because we do not understand language it is what makes us special).

I think general intelligence is the next frontier in the battle of keeping humans special, but in drawing this line between us, on the one hand, and computer systems and animals on the other. General intelligence is often seen is a general problem solver, working on abstract concepts extracted by the mind/brain using slave systems (vision, audition, memory). In an endless problem space, of which the best and brightest solutions are regularly selected by us, homo sapiens sapiens.

For me, this immediately breaks. Something like chess (for a long time considered to be a hallmark of cognition) turns out to be relatively simple, something like walking is, in comparison, very difficult. That an ant and a chess grandmaster are perfectly able to walk does not show that walking is simple, just that walking is a relevant problem to solve.

With chess and walking it is clear what you refer to. General intelligence is a concept of which it is unclear if it exists as such. It might ‘really’ exist, it might be a efficient mid-term concept to be filled in, it might obscure solutions (like the LAD did in language research).

My bet would be that many of the mechanisms of the mind/brain (vision, memory, movement) solve very complex problems without us becoming aware of it. Psychology/AI will help show how complex these problems really are (vision, memory, movement). I think Psychology /AI like research will also show these mechanisms reduce the problem space substantially. This would reduce the problem space for general intelligence, and maybe even to such an extent that it will not take a miracle to solve.

Of course, one of the really great things of science is that we might learn what the correct answer is. I am looking forward to discovering how our conceptual framework will change, but any clear answer, and given the rapid progress we might get there, will fill me with great pleasure.

Steven

Steven Scholte’s question is for Han van der Maas (Psychological Methods)

Dear Han,

You are a great expert of intelligent systems. Also, you are very knowledgeable about chess related research. How has your view on chess (as a model of cognition/problem solving) changed in the last 20 years? Has this altered your view of Good Old Fashioned Artificial Intelligence?

Steven

Ingmar Visser’s (Developmental Psychology) question

Dear Steven,

Artificial intelligence is once again on the rise after what is called the AI Winter. Psychology and AI are of course closely intertwined and share a common interest in understanding the mechanisms underlying intelligent behavior. How can current developments in AI deepen our understanding of human intelligence? Is there a limit here or will AI be truly intelligent this time around?

Ingmar

Steven Scholte’s (Brain and Cognition) answer

Dear Ingmar,

Thanks for this inspiring question, I have pondered on this for some weeks.

Human intelligence is a special category. It maybe is the new ‘language’. In the last century, language was often used as an argument for the manifest superiority of humans over animals.  Developments have made it clear that language is much less complex than was assumed and science has exorcised miracles like the hypothesized ‘Language acquisition device (LAD)’ (a proposed evolutionary derived black box system enabling humans to learn such a complex system). Also the hard boundary between humans and other great apes has turned out to be much more gradual. The miracle of language has not been solved, it has evaporated. Miracles are the domain of people like patriarch Krill, not science (and notice how hilarious it is that the big conclusion of the LAD was that because we do not understand language it is what makes us special).

I think general intelligence is the next frontier in the battle of keeping humans special, but in drawing this line between us, on the one hand, and computer systems and animals on the other. General intelligence is often seen is a general problem solver, working on abstract concepts extracted by the mind/brain using slave systems (vision, audition, memory). In an endless problem space, of which the best and brightest solutions are regularly selected by us, homo sapiens sapiens.

For me, this immediately breaks. Something like chess (for a long time considered to be a hallmark of cognition) turns out to be relatively simple, something like walking is, in comparison, very difficult. That an ant and a chess grandmaster are perfectly able to walk does not show that walking is simple, just that walking is a relevant problem to solve.

With chess and walking it is clear what you refer to. General intelligence is a concept of which it is unclear if it exists as such. It might ‘really’ exist, it might be a efficient mid-term concept to be filled in, it might obscure solutions (like the LAD did in language research).

My bet would be that many of the mechanisms of the mind/brain (vision, memory, movement) solve very complex problems without us becoming aware of it. Psychology/AI will help show how complex these problems really are (vision, memory, movement). I think Psychology /AI like research will also show these mechanisms reduce the problem space substantially. This would reduce the problem space for general intelligence, and maybe even to such an extent that it will not take a miracle to solve.

Of course, one of the really great things of science is that we might learn what the correct answer is. I am looking forward to discovering how our conceptual framework will change, but any clear answer, and given the rapid progress we might get there, will fill me with great pleasure.

Steven

Steven Scholte’s question is for Han van der Maas (Psychological Methods)

Dear Han,

You are a great expert of intelligent systems. Also, you are very knowledgeable about chess related research. How has your view on chess (as a model of cognition/problem solving) changed in the last 20 years? Has this altered your view of Good Old Fashioned Artificial Intelligence?

Steven

Spiegeloog Editors

Author Spiegeloog Editors

Spiegeloog editorial staff.

More posts by Spiegeloog Editors