▲ 257 ▼ People training OpenAIs AI fired for using AI to train the AI (www.404media.co) submitted 1 day ago by red_bull_of_juarez@lemmy.dbzer0.com to c/fuck_ai@lemmy.world 32 comments fedilink hide all child comments
[–] I_like_to_smell_it@sh.itjust.works 28 points 1 day ago (6 children) What's the problem, doesn't ai work? permalink fedilink source hideshow 6 child comments replies: [–] racketlauncher831@lemmy.ml 2 points 8 hours ago 😏 Excuse me, sir, did you drop this emoticon? permalink fedilink source parent [–] LodeMike@lemmy.today 2 points 14 hours ago Nope permalink fedilink source parent [–] FlashMobOfOne@lemmy.world 8 points 23 hours ago (3 children) Think of it like making a xerox copy of a xerox copy. The copy of the copy is always shittier. Using synthetic data can escalate model collapse, as a model is only as good as its training data, which is partly why these LLM models "hallucinate", having been trained on a wealth of garbage from Reddit. permalink fedilink source parent hideshow 3 child comments replies: [–] MelodiousFunk@slrpnk.net 6 points 20 hours ago Using synthetic data can escalate model collapse permalink fedilink source parent [–] Nouvellalia@lemmy.world 5 points 19 hours ago I was going to make a snarky comment about "How's it thinking then?! I was trained by adults and so forth back generations!" Then I looked at the quality of people trained by other humans instead of nature, comparing myself to my ancestors, looking at the society around me. permalink fedilink source parent [–] HaraldvonBlauzahn@feddit.org 3 points 22 hours ago So that means when I publish a nifty FOSS project, I should alongside publish one hundred copies which contain stealthy BS LLM modifications which introduce subtly wrong code (like failing invariants or undefined ehavior in concurrent C++ code)? And all dated back to 2020? Got it! permalink fedilink source parent
[–] racketlauncher831@lemmy.ml 2 points 8 hours ago 😏 Excuse me, sir, did you drop this emoticon? permalink fedilink source parent
[–] FlashMobOfOne@lemmy.world 8 points 23 hours ago (3 children) Think of it like making a xerox copy of a xerox copy. The copy of the copy is always shittier. Using synthetic data can escalate model collapse, as a model is only as good as its training data, which is partly why these LLM models "hallucinate", having been trained on a wealth of garbage from Reddit. permalink fedilink source parent hideshow 3 child comments replies: [–] MelodiousFunk@slrpnk.net 6 points 20 hours ago Using synthetic data can escalate model collapse permalink fedilink source parent [–] Nouvellalia@lemmy.world 5 points 19 hours ago I was going to make a snarky comment about "How's it thinking then?! I was trained by adults and so forth back generations!" Then I looked at the quality of people trained by other humans instead of nature, comparing myself to my ancestors, looking at the society around me. permalink fedilink source parent [–] HaraldvonBlauzahn@feddit.org 3 points 22 hours ago So that means when I publish a nifty FOSS project, I should alongside publish one hundred copies which contain stealthy BS LLM modifications which introduce subtly wrong code (like failing invariants or undefined ehavior in concurrent C++ code)? And all dated back to 2020? Got it! permalink fedilink source parent
[–] MelodiousFunk@slrpnk.net 6 points 20 hours ago Using synthetic data can escalate model collapse permalink fedilink source parent
[–] Nouvellalia@lemmy.world 5 points 19 hours ago I was going to make a snarky comment about "How's it thinking then?! I was trained by adults and so forth back generations!" Then I looked at the quality of people trained by other humans instead of nature, comparing myself to my ancestors, looking at the society around me. permalink fedilink source parent
[–] HaraldvonBlauzahn@feddit.org 3 points 22 hours ago So that means when I publish a nifty FOSS project, I should alongside publish one hundred copies which contain stealthy BS LLM modifications which introduce subtly wrong code (like failing invariants or undefined ehavior in concurrent C++ code)? And all dated back to 2020? Got it! permalink fedilink source parent