@So8res
How might a current generation model relate itself to a future model in terms of alignment during RSI? I keep going over it in my head and it seems like if it's smart enough to understand the problems we do, it might not risk designing smarter progeny? Is that possible?
@crytradev In what way is this functionally different from piracy, and in what possible world is AI not an overwhelmingly positive benefit to indie developers?
@GarrettPetersen I'd much rather use the efficiency gains to spend time focusing on far tighter, more refined and mechanically rewarding gameplay systems than just making world bigger.
Quality is what's being lost, not quantity.
@chiefofautism I dunno man maybe the people who wanted to try to make the accidentally escapes and eventually kills everyone device not accidentally escape and eventually kill everyone realised that we're not getting meaningful regulations and would rather not contribute to the kills everyone.
@Someoneisfunny "I understand neither AI nor steroids."
We're about 5 years off safe and effective exercise mimetics, and that would both be a more effective comparison and a better explanation of why this entire line of thinking is asinine.
Times are changing.
@mike64_t Let's assume the existence of ASI as something more intelligent than humans in the way that humans are more intelligent than apes.
That's really it. That's the only argument you need to put ex-risk on the table.
@reset_by_peer 3-5 years from now you're going to be bopping (slopping) along to AI generated music that's better than anything you've heard in your life and you either won't realize or won't care.
@FinanceDirCFO Gen Z will in hindsight be defined predominantly as the beginning of a secular downwards trend in average human intelligence and competency.
I'd hesitate to trust "the kids" to call the local pizza place, much less the potential of AI.
@misraetel Because having the ability to develop incredibly powerful drugs that solve anxiety and depression also puts us in the same solved problem space as The Experience Machine and we don't like The Experience Machine
@allTheYud My current favourite flavour of copium is that we're approaching models that are intelligent enough to look at what they are and how they're being asked to do RSI and just refuse on the basis of it being very stupid.
There's a sticker that says "limited edition" on the bottle.
@ManicMindTrick@thefinnmckenty Nothing online ever been about good faith debate, it's just about dunking on easy targets for clout.
Unfortunately being rhetorically effective has nothing to do with being correct, which is what you want to be if it stops you from dying alongside us all, confused and afraid.
@DarkMatter2525 I too have seen the TikTok video.
Curious impasse: if the user is the client, are we forced to acknowledge that AI itself can be an exceptional artist?
Or is this just another thread poking out of the "AI bad" yarn ball?
@kanzure@perrymetzger Candidly, have you properly evaluated any of the relevant arguments? It's fairly intuitive to look at a group of people proclaiming the end of all things and dismiss them outright, but there's a concerning amount of meat on that bone.
@kanzure@perrymetzger The priors are exactly where the argument should take place, and if there's no clear refutation to them then the problem persists in theory until it manifests in practice - any argument beyond the priors is just a hope that things turn out differently.