I agree, up to this point most everything we've seen are cartoonishly simplified demonstrations in virtual worlds, or low stakes requests like retrieving a can of soda for you. I don't think these are simple little games because that's all AI is capable of right now, I think they're simple tasks just for demonstrative purposes and the AIs themselves are actually capable of much more as-is.

Couple this with the fact that the public is informed of AI progress much later than the AIs creation itself AND the fact the public can't know too many specifics because it's a national security risk AND it could be hooked up to GPT-4 AND it's multi modal AND OpenAI has 10 billion dollars to throw at compute AND we have AIs helping us create the next batch of AIs (plus much more going for it) then you have an insane amount of reasons why the truly life-changing stuff is so much closer at hand than you might otherwise intuit