Why AI Isn’t Going to Become Conscious | Anil Seth | TED

14:57Added Aug 30, 2026

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 o...

Watch on YouTube →
Contributed by 刘嘉琪

Transcript

Transcript format
  1. 00:04So for centuries,
  2. 00:05people have fantasized about playing God
  3. 00:08by creating artificial versions of ourselves.
  4. 00:11From Mary Shelley's "Frankenstein"
  5. 00:14to Hal in Stanley Kubrick's "2001"
  6. 00:16and Ava in Alex Garland's "Ex Machina,"
  7. 00:19this is a dream reinvented with every breaking wave of technology,
  8. 00:23and with AI, the wave is a big one.
  9. 00:25The AI we have is already smart, at least in some ways.
  10. 00:29But could it ever be conscious?
  11. 00:33Will a robot ever gaze at a sunset and experience the beautiful colors,
  12. 00:37the reds and the oranges?
  13. 00:38Will it feel a sense of beauty or a rush of joy?
  14. 00:43Or will computers, however smart they get,
  15. 00:46always remain dark on the inside,
  16. 00:50always an object and never a subject?
  17. 00:54Whether AI can be conscious is one of the most consequential questions
  18. 00:59we face in our time.
  19. 01:01If computers can be conscious or sentient or aware,
  20. 01:03we'd be entering a new era in human history.
  21. 01:06We'd have new entities that have their own inner lives,
  22. 01:10new inventions that matter for their own sakes
  23. 01:13and not only for their effects on us.
  24. 01:16Conscious AI might suffer the click of a mouse,
  25. 01:19perhaps in ways we wouldn't even recognize.
  26. 01:22And if silicon can be sentient,
  27. 01:24then maybe our messy flesh-and-blood bodies
  28. 01:27will soon be superseded by machines that never age
  29. 01:31and cannot die.
  30. 01:34Now over the last few years,
  31. 01:35progress in AI has been simply astonishing.
  32. 01:38And who knows what's around the corner.
  33. 01:40Many experts think that conscious AI is possible.
  34. 01:45Quite a few think it's inevitable, and some, some think it's here already.
  35. 01:50I think they're wrong.
  36. 01:51I want to tell you why and why this matters so much.
  37. 01:55So I've been studying brains, minds and consciousness for nearly 30 years,
  38. 01:59and one thing I've learned is that to answer the question,
  39. 02:02"can AI be conscious,"
  40. 02:04we need to start by looking within ourselves
  41. 02:06at the makeup of our own human minds.
  42. 02:09Now we humans, we tend to see the world in our own terms.
  43. 02:12We know we’re conscious, and we like to think that we’re intelligent.
  44. 02:16So we think the two go together.
  45. 02:19And this is why some people think
  46. 02:21that consciousness might just glimmer into existence
  47. 02:24as AI gets smarter and smarter.
  48. 02:27But consciousness and intelligence are different things.
  49. 02:30Intelligence is all about doing.
  50. 02:32It's solving a crossword puzzle,
  51. 02:34assembling some furniture, navigating a tricky family situation.
  52. 02:38Consciousness, on the other hand, it's all about feeling and being.
  53. 02:43It's the difference between normal, wakeful awareness
  54. 02:46and the oblivion of general anesthesia.
  55. 02:49It's the bitter tang of coffee,
  56. 02:51it's the warmth of a log fire.
  57. 02:53The joy of seeing a loved one.
  58. 02:56Just because consciousness and intelligence go together in us
  59. 03:00does not mean that they go together in general.
  60. 03:03The assumption that they do,
  61. 03:05well, that's a reflection of our own psychology,
  62. 03:07not an insight into the nature of reality.
  63. 03:12Take language models like Claude or GPT,
  64. 03:15trained on vast quantities of written texts.
  65. 03:18They reflect back to us an image of ourselves,
  66. 03:20of our collective, digitized past.
  67. 03:23We talk about ourselves endlessly,
  68. 03:25and so do they.
  69. 03:27We wonder about consciousness and the meaning of it all.
  70. 03:31And so, it seems, do they.
  71. 03:33But language models are not conscious.
  72. 03:37They simulate consciousness.
  73. 03:38We project consciousness into them
  74. 03:41in the same way we might project faces into clouds,
  75. 03:45or even the image of Mother Teresa in a cinnamon bun.
  76. 03:48I don't know if you can see that.
  77. 03:50There she is, Mother Teresa in a cinnamon bun.
  78. 03:52(Laughter)
  79. 03:54We are built to be seduced, like Narcissus, by our own reflections.
  80. 03:59And so we see ourselves in our algorithms.
  81. 04:03I'm always struck that nobody really worries
  82. 04:05whether DeepMind's AlphaFold is conscious.
  83. 04:08AlphaFold predicts the structure of proteins
  84. 04:10rather than words and sentences,
  85. 04:12but under the hood, it's not much different from Claude or GPT,
  86. 04:16algorithms running on silicon and trained on vast reservoirs of data.
  87. 04:20AlphaFold just doesn't pull our psychological strings in the same way.
  88. 04:24So if we think that Claude is conscious, but AlphaFold isn't,
  89. 04:28that says more about us than it says about AI.
  90. 04:32(Applause)
  91. 04:36But how can I be so sure
  92. 04:37those systems, like Claude or GPT, aren't conscious?
  93. 04:41Well, nothing is for certain when it comes to consciousness,
  94. 04:44but the very idea of conscious AI depends on a deeper assumption,
  95. 04:48a kind of myth, really.
  96. 04:49And this is the myth that the brain is a computer
  97. 04:52that just happens to be made of meat rather than metal.
  98. 04:56Now consciousness in this story is a special algorithm,
  99. 04:59a collection of computations
  100. 05:01that just happens to be carried out in the wetware of the brain in us,
  101. 05:05but which could equally be carried out in silicon, in AI.
  102. 05:10But the computer is just one
  103. 05:12in a long line of technological metaphors that we've reached for
  104. 05:16when trying to understand the deep complexity of the brain.
  105. 05:21One time, the brain was a system of plumbing.
  106. 05:25Later, it was a telephone exchange.
  107. 05:28And for the last few decades, it's been a computer.
  108. 05:30And this most recent metaphor has been extremely powerful.
  109. 05:34But it is still a metaphor.
  110. 05:37And we will always get into trouble
  111. 05:38when we confuse a metaphor with the thing itself.
  112. 05:42The map with the territory.
  113. 05:45For one thing, in a real brain,
  114. 05:47there's no sharp separation between the mindware and the wetware.
  115. 05:52Unlike the separation that you get between software and hardware in a computer.
  116. 05:57And this really matters because in a computer,
  117. 06:00you can describe and understand everything about an algorithm,
  118. 06:03whether it's a language model or a word processor,
  119. 06:05without worrying about all the silicon shenanigans
  120. 06:08going on underneath.
  121. 06:09The computation, the algorithm is all that matters.
  122. 06:13But for brains,
  123. 06:14you just cannot separate what they do from what they are.
  124. 06:19And this means that what they do
  125. 06:20is unlikely to be a matter of computation of algorithm alone.
  126. 06:25Look closely at a brain, at any brain,
  127. 06:28and it becomes less and less plausible that all that's going on
  128. 06:31is just turning some numbers into other numbers
  129. 06:34in this endless dance of zeros and ones.
  130. 06:38Yes, there are neural circuits
  131. 06:40which exchange signals and may do computation,
  132. 06:42or at least something like it.
  133. 06:44But there's so much else that escapes the confines of the digital.
  134. 06:48Neurotransmitter chemicals course through the brain circuitry,
  135. 06:52electromagnetic fields sweep through the cortex like weather systems.
  136. 06:57Even a single neuron is such a beautiful biological machine.
  137. 07:02A far cry from the simplified, cartoon-like neurons
  138. 07:06that power today's AI.
  139. 07:09The brain is not, or at least not just, a computer made of meat.
  140. 07:14And so consciousness is very unlikely to be a matter of computation alone.
  141. 07:19And if this is true, then conscious AI is off the table,
  142. 07:23at least for AI as we know it today.
  143. 07:26Let me put it another way.
  144. 07:28What if we simulated every last detail about the brain
  145. 07:32in some massive supercomputer?
  146. 07:33Now if the fine details of the brain do matter for consciousness,
  147. 07:37well, wouldn't this be enough
  148. 07:38for consciousness to happen inside a machine?
  149. 07:41Well, a computer simulation of a hurricane does not create real wind.
  150. 07:46[A] computer simulation of a black hole doesn’t suck the Earth
  151. 07:49into its algorithmic singularity.
  152. 07:52Making these simulations more detailed can make them more useful,
  153. 07:56but it does not make them any more real.
  154. 08:00We can have a simulation of the brain,
  155. 08:02and you can make it as detailed as you want.
  156. 08:04This might make it more useful,
  157. 08:06but it's not going to make it any more conscious.
  158. 08:11Now consciousness, has to be said does remain a bit mysterious,
  159. 08:14but perhaps one reason for this
  160. 08:16is that we've been so constrained by our metaphors,
  161. 08:19by the idea that it just has to be some kind of information processing.
  162. 08:24After all, if you think the brain literally is a computer,
  163. 08:27then what else could it really be?
  164. 08:30But once we see the brain more clearly for what it really is,
  165. 08:34many new possibilities arise.
  166. 08:36And my own view, developed over many years,
  167. 08:39is that consciousness is intimately connected to our nature
  168. 08:42as living creatures.
  169. 08:44Unlike the abstract universe of computation,
  170. 08:48life is all about materiality.
  171. 08:50Unlike algorithms,
  172. 08:51living systems are deeply embedded in flows of energy and matter,
  173. 08:56and they continually regenerate their own conditions for existence
  174. 09:00and for persistence over time.
  175. 09:03I think we can draw a direct line from the molecular furnaces of metabolism,
  176. 09:08one billion biochemical reactions in every cell, in every second,
  177. 09:13all the way to the neural circuits
  178. 09:14that underlie each and every experience that we have,
  179. 09:18whether it's the sight of a blue sky or a pang of envy.
  180. 09:23Every conscious experience is imbued, however subtly,
  181. 09:27with a tinge of aliveness,
  182. 09:28with some core relevance for our future survival prospects.
  183. 09:32And at the heart of every experience, beneath even emotion, is this simple,
  184. 09:37shapeless and formless but fundamental feeling of being alive.
  185. 09:43And in this story,
  186. 09:44it's life, not computation,
  187. 09:47that breathes the fire into the equations of experience.
  188. 09:51And if this is right,
  189. 09:52then conscious AI will need to be living AI.
  190. 09:58Let me bring things together.
  191. 10:00First, we're built to see consciousness where it isn't,
  192. 10:03thanks to deep-seated psychological biases
  193. 10:06that bundle language, intelligence
  194. 10:08and consciousness together.
  195. 10:10Second, the brain is not, or not just, a computer.
  196. 10:13So consciousness is unlikely to be
  197. 10:15a matter of computation of algorithm alone.
  198. 10:19Brains are the kinds of things for which you can't separate
  199. 10:21what they do from what they are.
  200. 10:23And silicon is not up to the job.
  201. 10:26And third, many other things about our biological brains and bodies
  202. 10:30might matter for consciousness,
  203. 10:31including a deep connection between consciousness and life.
  204. 10:36Artificial intelligence is computer software.
  205. 10:40It is not a living mind.
  206. 10:42It might give the impression of being conscious,
  207. 10:45but it is vanishingly unlikely that it actually is.
  208. 10:50I want to close by returning to why this matters so much.
  209. 10:54Take the idea of AI welfare.
  210. 10:58Now there are already influential groups advocating
  211. 11:00that AI systems should have their own rights
  212. 11:03based mainly on the idea that they might be or become conscious.
  213. 11:08Now if real artificial consciousness is on the way,
  214. 11:12maybe through some other technology or other pathway,
  215. 11:14then this is entirely justified.
  216. 11:16We humans have a terrible track record
  217. 11:19in our ethical treatment of non-human animals and of other humans,
  218. 11:22and we don't want to make the same mistake again.
  219. 11:25This is one reason why even trying to build conscious AI is a very bad idea.
  220. 11:33But if conscious AI is just an illusion created by design,
  221. 11:37as I think it is,
  222. 11:39then by extending rights to these systems,
  223. 11:41we’d be sacrificing our ability to control, to regulate them
  224. 11:45and perhaps even to turn them off and for no good reason at all.
  225. 11:49And this is one reason why even AI that merely seems to be conscious
  226. 11:53is very dangerous for our society, too.
  227. 11:56And unlike real artificial consciousness,
  228. 11:59conscious-seeming AI is either already here
  229. 12:03or coming very, very soon.
  230. 12:06There are other reasons we should avoid creating AI that seems to be conscious.
  231. 12:10It makes us more psychologically vulnerable.
  232. 12:12We might be more likely to do what an AI says
  233. 12:15if we believe that it really feels for us,
  234. 12:18that it really understands us,
  235. 12:20even if what it's telling us to do is very bad for us.
  236. 12:24And finally, the very idea of conscious AI undermines our human nature.
  237. 12:30The mirror of AI goes both ways.
  238. 12:33We see ourselves in our algorithms,
  239. 12:36but we also see our algorithms in ourselves.
  240. 12:40And when we do,
  241. 12:41when we think of the mind as a collection of computations
  242. 12:44floating free from their bases in biology,
  243. 12:47we reduce and we diminish what it is to be a living,
  244. 12:50breathing human being in a real world.
  245. 12:54(Applause)
  246. 12:58"Frankenstein,"
  247. 13:00which Mary Shelley wrote when she was just 19,
  248. 13:03is often taken as a cautionary tale,
  249. 13:06a warning against the hubris of bringing something to life,
  250. 13:09a Promethean sin like stealing fire from the gods.
  251. 13:14The idea of conscious AI is a new Promethean dream
  252. 13:18wrapped up in a silicon rapture.
  253. 13:22And if conscious AI is possible,
  254. 13:25then so is the prospect of uploading our own conscious minds
  255. 13:28and floating off to eternity in a silicon cloud of living,
  256. 13:32or at least existing forever
  257. 13:34in the pristine circuits of some future supercomputer.
  258. 13:39Now the seductive power of this vision --
  259. 13:41of being at such a pivotal point in the history of life on Earth --
  260. 13:45I suppose it's understandable.
  261. 13:47And back in the real world, talk of conscious AI does other things too --
  262. 13:51it conjures this sense of technological wonder and magic
  263. 13:55which might keep share prices aloft and regulators at bay.
  264. 14:02But we should resist.
  265. 14:04The sacrament of the algorithm is most likely an empty dream,
  266. 14:08delivering not post-human Paradise, but silicon oblivion.
  267. 14:13We need a different story.
  268. 14:16One in which we're more part of nature,
  269. 14:18not apart from it,
  270. 14:20with consciousness more closely tied to living flesh and blood,
  271. 14:23not to the dead sand of silicon.
  272. 14:27AI might claim the prize of intelligence, at least in some ways,
  273. 14:31but consciousness,
  274. 14:32consciousness remains ours to celebrate
  275. 14:35and to share with other living creatures.
  276. 14:39So let's not sell our minds so easily to our machine creations.
  277. 14:43If we do,
  278. 14:44we not only overestimate them,
  279. 14:47we underestimate ourselves.
  280. 14:49Thank you.
  281. 14:51(Cheers and applause)
  282. 14:55Thank you very much.
Why AI Isn’t Going to Become Conscious | Anil Seth | TED — Transcriptly