is confident enough that AI isn’t conscious that her blog is titled “AIs Without Minds.” My view, in contrast, is that we don’t know whether AIs have minds and should have a good deal of uncertainty about it. I plan to write a long piece laying out my reasoning but in short:
I think biological naturalism is likely false and consciousness arises when the right functions are performed. On many leading theories, it’s
at least reasonably plausiblethat the LLMs of today are conscious—and likely that the LLMs of the future will be.I think the sophisticated behavior of LLMs should raise our credence that they’re conscious. Just as we’d be pretty sure that smart aliens who solve Millennium Prize Problems are conscious, even if they had silicon brains, we should similarly think it’s within the realm of plausibility that AI is conscious.
I think we should have a lot of uncertainty about what conditions are needed for consciousness. Given LLMs’ sophisticated behavior, this makes it reasonably likely that they’re conscious.
For this reason, I decided to read Burns’ blog to see if there are any convincing arguments for AI unconsciousness. I did not find any of the arguments convincing and found it worth explaining why. I admit that I have not read every essay on her blog, so it is possible that there are some arguments that I’ve missed.
Burns has another essay titled The AI Solipsism Trap (And Why You Shouldn’t Fall For It). In this essay, Burns objects to the argument that we should infer AI consciousness from AI behavior just as we should infer human consciousness from human behavior. Burns opens with the question “How Can I Know AI isn’t Conscious?” Inquiring minds want to know!
Her argument is that we have some direct faculty for inferring that there are other human minds. This ability is explained by features of developmental psychology and is a natural part of proper human development. This is not because we undergo some explicit process of reasoning, but instead it develops naturally. We know that others have minds like we know how to walk—innately, not through some explicit process of reasoning.
It’s certainly true that our beliefs that others have other minds aren’t the result of explicit reasoning. Kids think other kids have minds even before explicitly considering the problems of other minds. It isn’t as if five-year-olds are doing inference to the best explanation! But if that belief is to be justified, it should be that the reasoning we do *implicitly *can be made cogent explicitly.
Analogy: suppose that you innately feel that there are an even number of particles in the universe. This belief was adaptive. This wouldn’t justify believing that there are an even number of particles. Perhaps an innate sense can carry with it some pro tanto justification, but there should be some cogent reasoning behind this. Or, similarly, our belief in induction is innate—but if induction is justified, it really must be that past uniformity licenses high credence in future uniformity.
Now, I don’t think our reasons for inferring AI consciousness are as solid as our reasons for inferring human consciousness. Humans have brains that are very similar to one another; LLMs are very different. So I don’t agree with the original argument.
But section one of Burns’ article opened How Can I Know AI isn’t Conscious? I’ve been clickbaited! There is no argument given for this. All that is argued is that one reason to infer that AI is conscious is unsound. But note “one argument for P is unsound,” is not the same claim as “we can know that P is false.” If I wrote a blog post that opened “how can I know that my brother is a squid monster,” and the whole post was about how one argument against his being a squid monster was unsound, you would naturally feel some disappointment. So this piece, I claim, gives literally no reason at all to think AI isn’t conscious. I’m reminded of a passage from Alvin Plantinga in Where The Conflict Really Lies:
I announce to my wife, “I’m getting a $50,000 raise for next year!” Naturally she asks me why I think so. “Because the arguments against its being astronomically improbable fail! For all we know, it’s not astronomically improbable!”
Another piece is titled Biology and Consciousness: Where’s Your Evidence? The Perversion of the new ‘Silicon Science’. At risk of being perverted, I did not find the arguments convincing. The subtitle boldly claims “The evidence for ‘biological naturalism’ is the entire history of successful science.” Big iff true! Burns begins by complaining that the demand to explain why one thinks what they think “is always hurled at the biological naturalist, and rarely ever at the computational functionalist.”
Set aside that there are views other than biological naturalism and computational functionalism. I don’t think Burns’ claim about what is asked of computational functionalists is accurate—people often ask computational functionalists why they think what they think. But rather than haggling over this empirical question, we should simply agree on what should happen: anyone who is arguing for high confidence in one specific view should have a good argument.
Another general methodological precept of science —so utterly elementary that it usually goes completely undiscussed — is that science is the study of the natural world.
[Let me repeat that, because it’s so basic you might have missed it. Science studies nature.]1Thatis the assumption functionalism now requires that we defend.
But algorithms implementing functions are also part of nature. It isn’t as if computational functionalists think that there are ghosts flitting about and demanding contrary evidence. Instead, biological naturalists are often very confident that no non-biological system could be conscious, yet do not have any convincing reasons for this conviction.
Both biological naturalists and computational functionalism are claims about the physical systems that give rise to consciousness. Both say that certain natural systems give rise to consciousness. Thinking science is about the natural world gives no reason to prefer biological naturalism to computational functionalism.
Of course, this is not to say that science doesn’t develop artificial tools in an effort to understand the world, or that socio-political forms of hierarchy do not impact those efforts, oftentimes blurring the lines between science and engineering (consider, as an
[obvious example], the making of the first atomic bomb, which required taking newly understood theoretical physics and transforming it into a massive industrial engineering project).
It is just to say that the scientific enterprise generally assumes as its object of study the organic world.
What the heck? Physics does not study the organic world. Chemistry does not chiefly study the organic world. Science tends to study the physical world. But it’s not just concerned with organic things, as the physical world is filled with lots of non-organic materials. Should we, prior to the advent of planes, have been confident that only biological organisms could fly, on grounds that “the scientific enterprise generally assumes as its object of study the organic world.”
Note the slip between “science is the study of the natural world,” and “the scientific enterprise generally assumes as its object of study the organic world.” These are not the same thing, because only some of the things in the natural world are organic.
It’s because the evidence is so overwhelmingly pervasive it is difficult to point to.
If you must point to something, you’d have to point to the entire history of successful science as one’s ‘proof’. The entirety of science is proof of the biological naturalists basic assumption: that organic matter is what consciousness resides in.
Evidence for a hypothesis is a fact that’s likelier if the hypothesis is true than if it’s false. What about the “entirety of science,” is more strongly predicted if biological naturalism is true than if it’s false. Biological naturalism and rivals make exactly the same predictions concerning the results of ordinary scientific experiments.
Because that is essentially what is being asked of the biological naturalist when they are required to show what exactly it is about nature that is important for consciousness. They are being asked to justify the methodological nature of all science.
But this is an absurd ask. And in reality, the opposite should be occurring: it is the computational functionalist that needs to justifying themselves.
Curiously however, that justification is rarely given (I would genuinely love to read a single argument for substrate independence).
There are many arguments for substrate independence. For the most famous, see Chalmers’ piece on fading and dancing qualia (I’m not sure what I think about those arguments). For some arguments against biological naturalism, see my piece with here. By default, if the reasons in favor of CF and BN are equal, you should have equal credence in both (plus some credence in other views). So if you are very confident in BN, there should be some specific reasons for thinking it’s right.
To the demand that biological naturalists defend the view that biology matters to consciousness, here is my response: the existence of ordinary science is
self-justifying, or if you like is vindicated through its past historical explanatory successes.Granting this, the burden of proof is on those who wish to
this successful methodological precept, and the evidence, given how well this precept works, should be overwhelming.alter
Views other than biological naturalism are *compatible *with the existence of ordinary science. How could a methodological constraint in science tell us about what physical structures you need for consciousness? This is a very baffling suggestion. In fact, the history of scientifically explaining mental states has mostly done so in terms of computational functions.
Note that the argument that Burns has given is a fully general argument for thinking every biological process needs biology. If we should deny machine consciousness because science methodologically assumes that things require biology, should we have assumed, prior to inventing airplanes, that flight needs biology? Similarly, before the advent of LLMs, should we have assumed that cogent text production needed biology?
As best I can tell, this argument only works if we equivocate between “organic” and “natural.” But those are not the same thing and methodological points that are plausible when applied to the natural world are *totally incredible *when applied to the organic world. Science *does *study the natural world, it does not only study the *organic *world. Elsewhere, Burns elaborates:
This particular claim —that is, responding that computational functionalism is likely false is one that needs to be ‘established’ — is an instance of what I have called in previous work ‘
[the perversion of the new silicon science]’.As I recently
[argued,]the assumption that those who believe consciousness is biological need to prove this point has the science completely backwards. There are basic methodological precepts that exist in science for a reason:they have proven highly successful.Assuming that natural phenomena arise from organic matter is one of those precepts.
This is wrong on two counts. First, lots of natural phenomena don’t arise from organic matter. For example, planets are natural, but they don’t arise from organic matter. Second, these theories have been successful primarily by suggesting that biology is one of the things that gives rise to natural phenomena, not the only possible thing. There aren’t that many natural properties that can only arise from biology.
If this methodological constraint was correct, at best it would show that biology is one of the things that gives rise to consciousness, just as, across many other sciences, we often think biology gives rise to all sorts of other physical properties. But it would not show that ONLY biology gives rise to consciousness. Certainly it is not taken as a basic principle of science that only biology can give rise to natural properties.
So it seems that Ellen Burns’ master argument for AI insentience is unconvincing. This is a more common trend I’ve found among those who are confident that you need biology for consciousness. While they are often quite confident in their views, when I peek under the hood at their arguments, I find them to have very little weight. There is no good argument for being very confident that you need biology to have a mind.
