Google AI Chief Shane Legg Says There's a 50% Chance We'll Hit AGI in 5 Years

No, although I am more generous than most in according it some degree of sapience. The problems with existing LLMs are twofold: They're not build with a sense of memory, but rather pop into existence as autistic savants with no real agency, and (allied to this) are bad at asking questions, unless explicitly instructed to that end.

With a little effort you can get an LLM chat bot to introspect its text evaluation process, create memory structures to be filled with variables or decision criteria and much more, but it's very hard to get them to 'push back' on the user in any way such as asking for clarification or context absent in the original prompt.

An interesting analogy for this failure: imagine a spy (LLM) is impersonating a military officer (quasi-autonomous agent). An underling suspects a substitution and challenges the officer for proof of identity: our inept spy bends over backwards trying to explain that they're really who they appear to be, instead of bossing the underling about.