IMO these questions are impossible to place bets on without very strict resolution criteria. Could be anywhere from 1% to 99% probability depending on your definition of AGI.

@nsokolsky That's a good point. I'm currently on the lookout for an objective set of criteria that clicks for me. If you have any recommendations please let me know. If I don't find any that suit my framing of this question then I'll write a list of specifications in the description.

@RemNi “No intellectual task remains where a human or a group of humans are capable of producing a better output than an AI”

This is solved for chess and Go but not for (say) writing good standup material.

@nsokolsky that is not true for either. https://manifold.markets/0482/by-2030-will-we-have-computer-able

Your definition also sounds closer to the related ASI questions

For this what is chatgpt lacking besides "learn quickly, and learn from experience."?

@singer GPT4 can't solve these two problems:

/MatthewBarnett/will-gpt4-get-the-monty-fall-proble

/dreev/will-an-llm-be-able-to-solve-confus

Courtesy of @MatthewBarnett and @dreev

And then beyond that I'd say it's unable to reason and plan over long contexts like a human, even when provided with databases and retrieval capabilities. GPT4 is also unable to process video data (although there are ways to hack this in to the vision encoder by supplying frames from a video). Those are just a few examples, but there's potentially a long list.