Programmer: AI could certainly become conscious

AI researcher Casper Wilstrup, founder of drug design firm Abzu, tells us that he has been obsessed with consciousness all his life. He first began to think about it seriously when he became aware of the concept of death:

It was a gut-punch realization: the day I die, everything ceases to exist from my point of view. The world might as well go dark. It was like staring into “The Nothing” from Michael Ende’s “The Neverending Story.” It wasn’t even eternal darkness; the very notion of “eternity” itself felt empty if there were no experiences to fill it.

Casper Wilstrup, “With Artificial Intelligence, Philosophy of Mind Has Become an Experimental Science,” Medium, Sep 10, 2023

That’s poetic as well as frightening. What happened next though feels like a distinct change in mood. We are presented with five propositions:

● Consciousness is real, not imaginary.
● We know what it is, even if we’re still figuring out how it happens.
● Consciousness is natural, so it’s something we can study.
● AI can be conscious too. Saying otherwise implies that consciousness isn’t natural.
● We can, in principle, measure an AI’s consciousness.

So Wilstrup wants to be a naturalist (physical nature is all there is) but he does not want to be an eliminationist like Daniel Dennett (1942–2024), who held that consciousness is an illusion.

For support, Wilstrup cites analytical philosopher Galen Strawson who has pronounced the eliminationist view “the silliest claim ever made.

Panpsychism: If a couch can be conscious…

Strawson became a panpsychist (the view that all life forms — or everything in nature — is conscious). That gives some sense of where Wilstrup is heading. Panpsychism, of course, provides a basis for artificial intelligence to be conscious. And Wilstrup offers a panpsychist argument for that, necessarily quoted at some length below:

Sticking with naturalism, I can’t deny that I’m conscious stuff. Neither can you. Denying that would lead you down the road of either saying you don’t exist (which is pretty wild) or saying you’re not natural (which sounds pretty mystical).

Let’s go a little further. Can we say for sure that non-conscious stuff exists? Think about it. I look around, and sure, my couch seems pretty dead to the world. But can I say with certainty that there’s not even a smidgen of consciousness in there? Nope. That’s the essence of panpsychism: the idea that everything, even the stuff that seems totally inert, has some basic level of consciousness. Not saying my couch is pondering life’s big questions, but at a basic, fundamental level, it could be experiencing something.

Why would I make such a claim? Because it ties everything up in a neat, simple package. It explains the only fact I know for sure — that this meaty guy on the couch, me, exists and is conscious.

Stop! Don’t move on until you truly grasp this point. Of course, you could claim that non-conscious stuff exists as well. But if you go that route, you’ve got some serious explaining to do. You see, we know that conscious stuff exists, and now you’re saying that non-conscious stuff exists alongside it. That means you’re suggesting two completely different kinds of stuff coexist. So how would you explain that humans somehow have the consciousness-enabled type of stuff, while a chair doesn’t? How does that sorting process work?

Postulating two kinds of stuff complicates things enormously. It’s a step you should only take if you have solid evidence, which in this case, you don’t. Just because the stuff in the chair doesn’t display the kind of consciousness you experience doesn’t mean it lacks the properties needed to create your specific form of consciousness.

Wilstrup, “An Experimental Science

Well, if a couch can be conscious, so of course, can AI. Wilstrup, who spends his days tinkering in the lab with AI, does not think that today’s versions are there yet. But panpsychism justifies his faith in future developments because, from that perspective, AI turns consciousness studies into an experimental science:

Eventually, we’ll pinpoint what sets a conscious being apart from a non-conscious one. We’ll figure out the specific arrangements that allow unified consciousness to emerge from the proto-conscious properties of individual parts.

The game plan? Back to the lab to build zombie AIs, test, experiment, and theorize. Try, fail, and try again. Sounds familiar, right? It’s science — the good ol’ scientific method in action. This is how AI turns the philosophy of mind from armchair speculation into a hands-on science.

That’s right: With AI, the philosophy of mind has graduated to an experimental science. No more guesswork — just build, test, and learn.

Wilstrup, “An Experimental Science

But are we really any further ahead by claiming that panpsychist assertions are experimental science? If Wilstrup can create conscious machines in his lab, he can. If he can’t, he can’t.

If he can’t, there are likely hard reasons why he can’t, including the type of problem Eric Holloway raises: “there is no way to build a computer that cannot be reduced to the logic of 1’s and 0’s.” Therefore, no matter how sophisticated an AI becomes, it will never get beyond computation.

Wilstrup is welcome to try but he is mainly demonstrating why panpsychism is becoming so popular today, as an alternative to Daniel Dennett’s straightforward eliminative materialism that was once so fashionable.

You may also wish to read: How a prominent chemist came to think that all life forms are conscious. Emeritus professor of chemistry Addy Pross explains the reasoning process: A universal Darwinism can generate cognition in chemistry. Tracing the “chemical roots of consciousness,” he ends up affirming panpsychism, but insisting that nature is a “technologist.” Someone must do the thinking around here, after all.