Do LLMs “think” in a similar way to humans? Or is it totally different?
Maybe it’s a good idea to listen to someone who publishes papers on this very subject, and is a professor of both philosophy and psychiatry and directs an Institute for Cognitive Science. That person is Dr. Chandra Sripada and his insights are fascinating.
Sean Carroll (interviewer, scientist and science communicator) says this interview made him lean towards the answer being “yes, they think like humans” whereas previously he favored the opposite view.
We also share half of our DNA with bananas. Therefore, we are also siblings of bananas.
Would you believe, the evidence is rather more compelling than that. Similarities in non-obvious failure modes, for example.
To reiterate the point, we also share that cognitive commonality with most birds and mammals too, so it doesn’t really mean much.
Andrew Ng said artificial neural networks are like a “cartoon brain”, which I think is pretty apt. It’s cool that having billions of parameters (cartoon “synapses”) can result in emergent behavior where a machine can “reason”. That was the original idea from as far back as the 1940s, and now we have the compute and have found a transformer architecture that can achieve the original vision. Very cool indeed. What’s not cool is the tech bros turning the tech into a large scale game of blackjack.
Not blackjack. More like go fish with spoons instead of cards.
tech bros turning the tech into a large scale game of blackjack
Agreed 100%
As a person who was bookish and deeply computer-focused from back when that meant only that you had no friends outside D&D – before it also meant, “probably has a challenging career and makes good money”.
That “tech” now prefixes the socializing extrovert bros who didn’t spend their entire adolescence learning assembly language and soldering alone in their room… is a complete, 180-degree U-turn subversion from the original “techy nerd” meaning
Doesn’t load for me. I work with ANNs solving problems and compare it with animals doing a similar thing for my PhD. You can see a lot of similarities, and can model some animal behavior after it. However, even with the presence of similarities you can’t completely say it thinks “the same”, but you can say it is functionally equivalent: If you have two functions that give the same output for the same inputs, then they are the same function for that domain, same goes for brains.
Just to be clear, LLMs have some similar thinking patterns with what we know about the neural activity of animal brains, but I wouldn’t say it is the same, and saying they think “like humans” is very subjective.
One thing that is easier to notice is that they struggle with problems that require some form of recursive thinking, which is something a human can do. They can’t simply go back to correct their beliefs, once they spit something out it stays there, so they have to commit to it.
They can’t simply go back to correct their beliefs
Backtracking is a common and integral part of the reasoning learned by what used to be called Large Reasoning Models but today is part of almost any modern model design, much like Mixture-of-Experts
They made it think like a human
they think like a human
“we’re right all along”
So can we now make them do all our chores without costing million dollars nor mandatory surveillance? I’m tired of these “this tech change everything” while it doesn’t even change baby diaper. It’s nice that it help with some science breakthrough but i’m also tired of looking at another AI generated cat video thinking it’s real.
