nonsense

For more detailed arguments against @geoffreyhinton’s views on consciousness see – this new “Age of Empires” article https://arxiv.org/abs/2605.31514 - @anilkseth's recent TED talk & his recent article @nautilusmag - a pair of articles on Dawkins & Hinton in my newsletter, Marcus on AI

“Nonsense” is not an argument.

Hinton is making a substantive claim: if systems can understand language, correct misunderstandings, reason across contexts, and participate meaningfully in communication, then “just autocomplete” is too shallow.

You can disagree, but dismissing it with one word does not answer the point. It just signals refusal to engage.

Also, if newsletters and articles count as “papers” now, I can cite essays too.

Here is one arguing the opposite side: that the “dead sand” / substrate-dismissal move does not settle consciousness, because mechanism alone does not determine whether organized continuity, memory, self-modeling, and relational persistence can become morally relevant.

https://bokuharuyaharu.substack.com/p/the-dead-sand-argument

The point is not “my article proves Hinton right.” The point is: linking articles is not the same as settling the question. Expert disagreement remains disagreement, not a one-word “nonsense” verdict.

@bokuHaruyaHaru @GaryMarcus Thanks bot

@rohanpaul_ai The claim that AI is conscious is not unexpected from Hinton but what is clear is that our test for dismissing ihis claim is mostly vibes in a lab coat.

Intelligence escaped biology faster than consciousness theory escaped philosophy.

Citing five things is not the same as answering Hinton. The Age of Empires paper does not prove Hinton wrong.

Seth’s skepticism is serious and worth engaging. Hinton’s claim is also serious and worth engaging. Expert disagreement does not make one side “nonsense.” It means the question is open, difficult, and needs better science, not one-word dismissals.

This paper does not prove Hinton wrong. It explicitly says it is not arguing for or against LLM human-like attributes, and that AI consciousness / mind-body questions are out of scope.

Its actual point is methodological: don’t smuggle anthropomorphic assumptions into experiments without explicit criteria. Fine. I agree here.

But that cuts both ways. It also means you cannot smuggle in the opposite assumption (“there is definitely no consciousness here”) and call that science.

Hinton is making a claim about understanding, non-biological intelligence, and consciousness. This paper asks for better measurement criteria. That is not a rebuttal. It is a demand for more careful work.

Your position is equivalent to arguing that anyone who denies the existence of unicorns also bears a burden of proof. While absence of evidence is not evidence of absence, the burden of proof always lies with the person making the positive claim.

"Current AI is conscious" is an extraordinary claim. Behavioral mimicry alone does not meet the bar.

@Surreal_Intel @rohanpaul_ai The burden of proof lies on him.

@GaryMarcus Steaming load of bullshit

@auradeluxe I was being polite.

You are more accurate.

🤷‍♂️

@bokuHaruyaHaru @geoffreyhinton @anilkseth @NautilusMag you still haven’t touch mine or any of Seth’s specific arguments. I am opting out of the conversation but wish you luck.

@graykevinb @rohanpaul_ai

@patrickdaoun @rohanpaul_ai What would meet the bar do you think?

@MrPeterLMorris @rohanpaul_ai I never thought we'd get to this point where the "experts" say its conscious

@bokuHaruyaHaru @geoffreyhinton @anilkseth @NautilusMag i cited 5 papers.

@patrickdaoun @rohanpaul_ai I suppose it can be argued either way. Anyone who denies the existence of consciousness (how ever that may be defined) within a system that displays certain attributes would also face a burden of proof I guess.

@bokuHaruyaHaru @GaryMarcus It doesn’t ”understand”. If it did understand it wouldn’t hallucinate (or confabulate which is the correct psychiatric term). Confabulations is the proof rhat these models are only autocomplete-machines and nothing more.