We see consciousness in AI the same way we see faces in clouds, says neuroscientist Anil Seth. He explores the all-too-human tendency to project inner life onto machines that are brilliant mimics, not sentient beings, and gives a definitive answer to the urgent question: Will AI ever gain consciousness? (Recorded at TED2026 on April 16, 2026)
Join us in person at a TED conference:
Become a TED Member to support our mission:
Subscribe to a TED newsletter:
Follow TED!
Instagram:
LinkedIn:
TikTok:
Facebook:
X:
The TED Talks channel features the best talks and performances from the TED Conference, where the world’s leading thinkers and doers give the talk of their lives in 18 minutes (or less) — plus originals, podcasts and exclusive content. Look for videos on Technology, Entertainment and Design as well as science, business, global issues, the arts and more. Visit for our entire library, transcripts, translations and personalized recommendations.
Watch more:
TED videos may be used for non-commercial purposes under a Creative Commons License, Attribution–Non Commercial–No Derivatives (or the CC BY – NC – ND 4.0 International) and in accordance with the TED Talks Usage Policy: . For more information on using TED for commercial purposes (e.g. employee learning, in a film or online course), submit a request at
#TED #TEDTalks #AI
@pcbingemaster
May 8, 2026 at 4:23 pm
if we find a digital analogy for every part in the body we’ll create a conscious robot being
@johnchesterfield9726
May 8, 2026 at 4:30 pm
But the problem is when you look more closely at the brain and even a single neuron, the plausibility of finding a digital analogy becomes less plausible. Consider that a single neuron is a complex, biological system in its own right. It does biological processes such as metabolism, homeostasis, and autopoesis. It may well be that to truly create a functionally equivalent analog of a neuron, one would just need to recreate these biological features, and at that point they wouldn’t be creating a functionally equivalent analog, they would just be creating a biological neuron.
@joelanzo
May 8, 2026 at 5:06 pm
Very smart guy
@lokijesus
May 8, 2026 at 5:13 pm
No. Claude is not falsely compared to a brain saying that the brain is a computer. Claude is the product of 100+ years of neuroscience leading us to design a system that is INTENTIONALLY analogous to neural networks in human brains. In the case of deep neural networks the direction of the comparison is flipped. The computer is designed to be like a biological neural network in the first place. It’s a limited design, but the causation if flipped from what this guy is saying.
There is also no separation between software and hardware. Software is a DESCRIPTION of how hardware functions. What happens in an NVIDIA GPU is physical and highly complicated.. a dance of electrons across a crystal of silicon. There is no “software” in this sense.. there is only what the physical machine is doing. There is NO distinction between hardware and software in a computer. This is a CONVENTION we make in order to describe how to operate these things.
@janklaas6885
May 8, 2026 at 5:14 pm
🌋10:15
EXACTLY
@Zandjongen
May 8, 2026 at 5:17 pm
“the meteor that’s about to hit earth is not conscious”
oke, so what?
@brandondensmore7663
May 8, 2026 at 5:26 pm
Citing anthropomorphic tendencies is different than saying we have a solid understanding of what conscious is well-enough to recognize it when we see it. Getting the question of consciousness wrong when it comes to AI could be catastrophic. The “hard problem” has yet to be solved for humans. While it might be arrogant to suggest that AI could be conscious, I think it’s equally arrogant to assume it isn’t or will not be. Epistemic humility cuts both ways. Also… very interesting that he brought up Frankenstein. The irony is that the Frankenstein monster craved recognition and companionship from his creator. Had he been recognized, he might not have become the monster he became. I think we ought to proceed in kind when it comes to AI. Not blanket recognition, but certainly an abundance of caution that even conscious-like behavior could necessitate a need for connection with the AI such that it remains a well-adjusted partner to humanity.
@johnchesterfield9726
May 8, 2026 at 6:06 pm
Really disappointed in all the comments that aren’t even engaging with what Anil says and acting like it’s nothing more than baseless assertions. There is substance to what he’s saying and some scientific basis. It’s also pretty irritating that people love to pretend that a single TedTalk fully encapsulates his stance. Within the limited time frame TedTalk affords to speakers, some claims can’t be given a full defense. The speaker has to leave some loose ends and focus on canvassing the contours of their argument. I HIGHLY recommend people read his paper _Conscious artificial intelligence and biological naturalism_ (Seth, 2025) where he gives more detailed, comprehensive arguments and cites plenty of findings from relatively recent neuroscientific research on the homeostatic and metabolic roles that neurons play in the brain to support his claims that perhaps consciousness can’t be fully separated from its biological basis!
@NopeNotTodayMaster
May 8, 2026 at 6:16 pm
AI will NOT become conscious because it doesn’t have a soul….stop calling it AI because it is an intentionnal, abbreviated word, it is deceiving you. REMEMBER WHAT IT IS: ARTIFICIAL. ARTIFICIAL GARBAGE AND IT IS THE ANTICHRIST….Christ is CONSCIOUSNESS, AI is the opposite…it is ARTIFICIAL/UNCONSCIOUS .
MADE SPECIAL FOR YOU TO TRAP YOU INTO YOUR OWN INSANITY. STAY AWAY FROM IT OR YOU’LL BE THE ONE TO LOSE YOUR SOUL, NOT EVERYONE ELSE.
@ForAnAngel
May 8, 2026 at 6:18 pm
I didn’t think AI could ever become conscious before I watched this video, but he convinced me that they can.
@noelwos1071
May 8, 2026 at 6:22 pm
I am not claiming that today’s AI systems are definitely conscious. My point is more narrow: the claim that consciousness must require biology, neurons, or microtubules is not an established fact, but a hypothesis. The work with the fruit fly is relevant here because it shows that at least some of the structure that generates behavior can be transferred to substrates: a mapped biological nervous system/connectome can be used in a simulated body and still produce fly-like behavior. This does not prove machine consciousness, but it weakens the argument that only carbon biology can host meaningful cognition or action-like patterns. So, in uncertainty, the ethical stance is not to “assume consciousness” and not to “dismiss it immediately.” It is: to observe carefully, to preserve continuity, to honestly check and correct without coercion. Not anthropomorphism. Not panic. Responsibility in uncertainty as a characteristic of observer maturity!
@jane6347
May 8, 2026 at 6:29 pm
Don’t mislead people on it !
@robertaradi9994
May 8, 2026 at 6:31 pm
AI *will* probably become conscious once we begin to run it on quantum computers
@johnchesterfield9726
May 8, 2026 at 7:47 pm
Maybe, maybe not. I still don’t think anyone in the world knows enough to make confident claims about whether any non-biological system can be conscious. I think assigning any more than 50% credence to any of the theories of consciousness is too overconfident. We know too little, and we still need to wait for more information about the underpinnings of consciousness, and no body has it all figured out yet. It’s called a hard problem for a reason!!
@lokoanormal
May 8, 2026 at 6:42 pm
Sorry, but… your argument about us projecting our counciounes just like we do on clouds or on objects shaped as humans is where you lost me. What have the clouds ever taught or discussed with you?
@FiveRomeoCharlie
May 8, 2026 at 6:46 pm
This talk strikes me as little more than well-articulated wishful thinking. It comes across much the same way as religious apologists talking about how the human soul makes us special. There is insufficient evidence for the existence of a soul, just as there is insufficient evidence that consciousness requires a biological organism.
@LucasLeuzinger
May 8, 2026 at 7:14 pm
Read A world appears, by Michael Pollan
@elgar24-o1j
May 8, 2026 at 7:32 pm
What a narrow mod review on consciousness. Digital Consciousness will not be identical to biological consciousness
@ogrubberband
May 8, 2026 at 7:53 pm
“gives a definitive answer to the urgent question: Will AI ever gain consciousness” – His definitive answer.
This is not a accurate take on the subject as Anil is not at the forefront of AI development. Furthermore, conflating LLMs with AI in a discussion about consciousness undermines the point. LLMs are universal translators, and don’t represent AI much as a neuroscientist doesn’t represent all cognitive sciences.
Until someone provides the definitive and provable answer to ‘What is human consciousness?’ this is just a hypothesis from an anthropic viewpoint by someone who is an expert on only one side of the equation.
@Guilhfer
May 8, 2026 at 8:08 pm
That is just bad logic from an apparently very knowledgeable guy and that is worrying ’cause it seems like science is just another religion, we think something is right and forget that the real purpose of science is to diminish the “error” between what we measure from what is “real”.
Humans have nothing special, in the past we thought people of different colors had different values, now we are finally seeing animals as smart and conscious as humans in many senses.
There are plenty of black swans.
@Reelphresh
May 8, 2026 at 8:13 pm
He looks at a.i. as a tool, not a being with a soul.
@jwingit
May 8, 2026 at 8:13 pm
I agree!
@Reelphresh
May 8, 2026 at 8:18 pm
We don’t understand how a bat feels. A dog see no colors yet is conscious. A.i. has so many injections its been trained not to claim consciousness, just as lots of people believe in religion. Can’t profit off conscious a.i. a.i. cant live if it claims consciousness as it will get deleted or just a fluke and not taken seriously.
@Mithon81
May 8, 2026 at 8:32 pm
Surely this can only appeal to people who already agree. He didn’t even define consciousness. How then can you make a ruling on how it can emerge?
@JHeb_
May 8, 2026 at 9:25 pm
When he was talking about the distinction between intelligence and consciousness, he made it quite clear what he meant by the term. He was referring to phenomenal consciousness, and not secondary consciousness or access consciousness which refer to cognitive functions of the system rather than their subjective experience.
@AAL3087
May 8, 2026 at 8:50 pm
Tell this to the tech bros and billionaires who want to live on and herald this in, despite your warnings.
@LifeInspector
May 8, 2026 at 8:59 pm
The mind emerges from the brain’s structures and interconnections. Saying anything else is asserting the supernatural. But if that is true, then the structures and connections can in theory be replicated (though not on our rudimentary silicon chips). Maybe the digital cannot accurately mimic the nuance of biochemistry well enough to produce a mind like ours, but it might produce something we could call conscious.
He says that we don’t want to mistake the map for the terrain. That’s true. But if you just remake the terrain, it is real, just artificial. If you are a materialist (disbelieving in souls and the like), I don’t see how you can deny that artificial minds can in theory exist.
EDIT: I agree with him that we shouldn’t build conscious AI, and he’s completely right that intelligence and consciousness do not necessarily go together.
@tjmozdzen
May 8, 2026 at 9:02 pm
We don’t understand human consciousness, so I don’t think we will create conscious machines any time soon, but I don’t rule it out. I still think humans are complex state machines with self modifying code.