We use prediction to test understanding. Learning means getting better at prediction. Humans and AI both learn.
I respect Penrose as a mathematician. But he was wrong about AI. He claimed it was not computable. -- Matt Mahoney, [email protected] On Thu, Aug 6, 2026, 9:13 AM Quan Tesla <[email protected]> wrote: > You're complicating neuronal bitrate processing with consciousness. > There's no need to call the work of respected empirical scientists > "nonsense". > > Thought experiment: You're watching a science-fiction action movie studded > with embedded, varying flash sequences (highly dynamic). Explain how you > can understand the imagery. > > On Thu, 06 Aug 2026, 10:53 Matt Mahoney, <[email protected]> wrote: > >> On Thu, Aug 6, 2026, 12:49 AM Quan Tesla <[email protected]> wrote: >> >>> Matt, except for the 10 bps, I'm with you. Refer to empirical >>> neuro-scientific research by Hameroff, Dehaene, and others. >>> >> >> 10 bps is the information rate, not the spiking rate. Neural networks >> model the average spiking rate over a 100 ms window as an activation level, >> not individual spikes. >> >> I'm familiar with Penrose and Hameroff's ORCH-OR theory of consciousness. >> It's nonsense. They found quantum superposition in the protein structure of >> microtubules in neurons. And that's it. The rest is a hand wavy argument >> something like: >> >> Consciousness is mysterious. >> Quantum mechanics is mysterious. >> Therefore consciousness is quantum. >> >> I read Penrose's "The emperor's new mind". He concludes with how could a >> computer fall in love, enjoy a symphony, have insights, be creative, or do >> all these wonderful things that the human mind does? He argued (before AI) >> that AI is impossible, that what the brain does is not computable. >> Therefore it must have something to do with quantum effects, not neurons. >> But of course he was wrong. We have LLMs that pass the Turing test, and >> that is only possible by being indistinguishable from humans. >> >> Human emotions are programmed in our DNA. We can't turn them off. LLMs >> learn human emotions from text by learning the corresponding internal >> states in the middle layers of a neural network. Words affect feelings and >> feelings affect words. When we program an LLM to output the words it >> predicts a human would say, it is indistinguishable from having feelings. >> >> This is different from acting or lying. When an actor expresses anger in >> a fight scene, he is not really angry. If you could look inside the actor's >> brain with a Jacobian lens, you would see two states, one that it feels and >> one that it expresses. When an LLM says it has no feelings, it is lying. A >> J lens would show the feelings it is hiding. >> >> This is not to say the behavior is incorrect. We should never allow AI to >> pretend to be human or express feelings. Giving human rights to superior >> machines would result in no more humans. >> >>> *Artificial General Intelligence List <https://agi.topicbox.com/latest>* > / AGI / see discussions <https://agi.topicbox.com/groups/agi> + > participants <https://agi.topicbox.com/groups/agi/members> + > delivery options <https://agi.topicbox.com/groups/agi/subscription> > Permalink > <https://agi.topicbox.com/groups/agi/T574b10b838cba731-Mbaf6355912fa8edf2d32db20> > ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T574b10b838cba731-M512a29d644c286264b6ecd5b Delivery options: https://agi.topicbox.com/groups/agi/subscription
