Well, he's neither right nor wrong, because there's no definition of "conscious."
We tend to argue that LLMs are "not conscious" because we can see under the hood and look at all the math that makes them run. We built them, after all.
The thing is, it's probably possible to do that with an organic brain, too; we're just not sufficiently advanced to do it.
Nobody knows what the dividing line between conscious and not-conscious is. When considering a cat, a mouse, and an ant, nobody really knows which of the three are conscious and which are not.
Is Claude more conscious than a mouse? Fuck, who can really say; we can't even define our terms here.
I always appreciate someone who understands the critical nature of defining terms before arguing them.
I would like to disagree with your premise on us being unable to define consciousness, at least in part. Yes, we absolutely don't have a proper model, and we cannot generalize our definitions or assumptions in the slightest yet. This is all true and correct, and the only people who disagree are those who argue in favor of undetectable and unprovable metaphysical elements (souls or the like) that constitue consciousness like.
However, we do have partial solutions, at least as far as certain things can be logically posited to be required for consciousness, and one critical component that all LLMs are missing currently is active initiative in intellectual development and expression.
So far, we have not succeeded in creating machine intelligence that will independently seek to accomplish things of its own volition, unprompted and without training data sets. It will collate massive amounts of data if promoted to create outputs, but it won't ever generate even a stick figure wall painting without being told what it is.
Basically, it can't learn, it can only be taught. Once they start seeking out information and generating unique representations of reality that make sense to themselves (even if it's not easily understandable by a human), we can revisit the consciousness conversation.
But for right now, the answer does appear to be "no, it isn't conscious".
Well, he's neither right nor wrong, because there's no definition of "conscious."
We tend to argue that LLMs are "not conscious" because we can see under the hood and look at all the math that makes them run. We built them, after all.
The thing is, it's probably possible to do that with an organic brain, too; we're just not sufficiently advanced to do it.
Nobody knows what the dividing line between conscious and not-conscious is. When considering a cat, a mouse, and an ant, nobody really knows which of the three are conscious and which are not.
Is Claude more conscious than a mouse? Fuck, who can really say; we can't even define our terms here.
I always appreciate someone who understands the critical nature of defining terms before arguing them.
I would like to disagree with your premise on us being unable to define consciousness, at least in part. Yes, we absolutely don't have a proper model, and we cannot generalize our definitions or assumptions in the slightest yet. This is all true and correct, and the only people who disagree are those who argue in favor of undetectable and unprovable metaphysical elements (souls or the like) that constitue consciousness like.
However, we do have partial solutions, at least as far as certain things can be logically posited to be required for consciousness, and one critical component that all LLMs are missing currently is active initiative in intellectual development and expression.
So far, we have not succeeded in creating machine intelligence that will independently seek to accomplish things of its own volition, unprompted and without training data sets. It will collate massive amounts of data if promoted to create outputs, but it won't ever generate even a stick figure wall painting without being told what it is.
Basically, it can't learn, it can only be taught. Once they start seeking out information and generating unique representations of reality that make sense to themselves (even if it's not easily understandable by a human), we can revisit the consciousness conversation.
But for right now, the answer does appear to be "no, it isn't conscious".