Raw LLM Responses

I dont understand all the praise about this interview. This person is supposedly an expert in AI, and not once have I heard him speak about the shortcomings of transformers and the doubts that it can take us to AGI. This feels like bullshit on many levels. Ask any AI to do some midly difficult intellectual job (like reconciling some financial data between a few files and using judgement like any accountant or auditor would) and it will fall on its face miserably. What we have now is not intelligence. We have some highly efficient search engine that is really good at impersonating humans and retrieve info they have been trained on. But the moment you ask anything out of it's training data, it fails like a 5 yo. How can it be so good at solving highly complex math problems but be fooled by a common brain teaser in which we introduce a small variation than any child would notice. Those things are far far far from being able to generalise what they "know'". And then wont be unless there is a major architectural breakthrough. This (those) breakthrough(s) could happen in the next few years or decades or never. So the current state of the art is irrelevant in the context of AGI, let alone ASI. And, yet, everybody seems to thing it is around the corner. The level of delusion of seemingly intelligent people is mind-boggling.