Maybe I'll end up eating my words completely but I just do not see a massive or even moderate-to-large societal change happening, given the current trajectory of things. GPT4 was released over 2 years ago and 5.0 has been confirmed to just be some sort of composite of tools existing on the current magnitude. It's clear that the current approach isn't going to get anywhere further--as much has been implied or said explicitly by the developers of these models themselves. However there is one existing caveat of sorts to me: the very convincing results of photorealistic video generation. I suppose for video of contemporary urban life it's so good because the internet is just absolutely full of countless hours of that sort of stuff. I wonder if we'll see an at least mostly AI-generated show that's set around modern day. Currently hallucination prevents sufficient consistency, but assuming that can be kept under control eventually, and that does not seem that unreasonable, I'm sure that some studio or another would try it. Would it eventually work for visuals that don't have as much data? Could it make wider inferences to compensate? Currently training an AI on AI seems to field rather poor results, but is that another avenue in the future? Remains to be seen... Of mild note is an AI-generated ad that ran during the recent NBA finals. Apparently it was made by a single person which tracks as I don't think any model is ready to be used for anything past "generic stock footage" in more formal productions, as finer control over the result isn't really there. The ad uses the current shortcomings of the tool to its advantage, going for a "global phenomenon" sort of thing to sidestep its issues with continuity. This doesn't really spell anything new--it's not demonstrating any capability that we haven't seen. But it is mildly notable in the fact that it exists and ran during a high profile event. Regardless, even within the current magnitude of agents there will surely be some lasting effects. Just because there's (likely) an excess of hype doesn't mean that there aren't legitimate applications. Maybe those phone assistants will actually be useful now. I don't follow consumer electronics much so maybe this is already starting to be the case but the capacity for natural conversement already exists, so it's just a matter of implementing it and hooking it up to useful things. Hopefully K-12 education makes appropriate curriculum changes. I'd never say no to more emphasis on critical and original thought. There are probably parts of artistic workflows where AI integration could prove useful, though these sorts of things have yet to be explored much for fairly obvious reasons. I've said more than once that AI as-is is inherently incapable of being superior to more manually authored approaches, and as far as I've seen (and I've seen a lot of products of gen AI) this has been entirely true. Who knows though, given the renewed interest and funding that brings, it's always possible someone will find those breakthroughs. I just wouldn't hold my breath.