On listener question why we don't iterate AI improving on AI ad infinitum toward ASI. From almost two decades working on developing AI, that is one of typical exmples of general public misunderstanding of how AI(LLMs) work. Because of the Math involved attempt to iterate AI like that will not lead to improvement but very quickly to degradation.
More importantly there have been many different Math approaches analyzing whether ASI/AGI is possible while it runs on curent computing technology. All those different approaches came to the same conclusion - AI run on current computing paradigm will never achieve self aware state. Absolute minimum required is Quantum Computing. Once we have QC at the level of current computing AGI becomes plausible. However, my own hypothesis is that will not be enough, I think that self organizing properties of the life will be needed too. Hence, the first (and only) AGI type we create will in my opinion be a biological living organism designed to be way more capable than us. Meanwhile AI will continue to be an incredibly powerful tool for many things.
More importantly there have been many different Math approaches analyzing whether ASI/AGI is possible while it runs on curent computing technology. All those different approaches came to the same conclusion - AI run on current computing paradigm will never achieve self aware state. Absolute minimum required is Quantum Computing. Once we have QC at the level of current computing AGI becomes plausible. However, my own hypothesis is that will not be enough, I think that self organizing properties of the life will be needed too. Hence, the first (and only) AGI type we create will in my opinion be a biological living organism designed to be way more capable than us. Meanwhile AI will continue to be an incredibly powerful tool for many things.