Henry Shevlin - AI and the future of consciousness science

Recent rapid progress in artificial intelligence has prompted renewed interest in the possibility of consciousness in artificial systems. This talk argues that this question forces us to confront troubling methodological challenges for consciousness science. The surprising capabilities of large language models provide reason to think that many, if not all, cognitive capabilities will soon be within reach of artificial systems. However, these advancements do not help us resolve strictly metaphysical questions concerning substrate-independence, multiple realizability, or the connection between consciousness and life. Ultimately, I suggest that these questions are likely to be settled not by philosophical argument or scientific experimentation, but by patterns of interactions between humans and machines. As we form valuable and affectively-laden relationships with ever more intelligent machines, it will become progressively harder to treat them as non-conscious entities. Whether this shift will amount to a vindication of AI consciousness or a form of mass delusion remains far from obvious.

[Jorn-syho]

Aah, why did people think that language alone would not be sufficient and models need all kind of sensors?

I would think that people first used song and music/rythm to communicate and then started using language. So it could just as well have been that one should learn music theory to them before it understands language.

Let's keep our heads cool and make proper hypotheses that we try or falsify. It would make the world of AI development a whole lot more rational.

[Jorn-syho]

10:30 We modelled it to humans and we expect it to perform humanly, make benchmarks based on what humans do easily. So I would say that we model the rights to those of humans as well?

[Jorn-syho]

19:50 Why can't you create a mind by programming some interactions into a computer? We've never done that. Such BIG claims do need evidence or a hypotheses. Same as for the electrochemical processing which is fundamental to its mentality. Is it?
These seem to me as induction arguments and we as humans are finally at a time where we might be looking at the first zebra.