• 0 Posts
  • 48 Comments
Joined 1 year ago
cake
Cake day: July 12th, 2023

help-circle

  • The tag does two things. One, for people like me (ND ASD) the tag allows us to actually understand what you’re trying to say. Kind of a dick move to exclude people from the conversation. Two, and far more important, it’s no longer a question if you’re being sarcastic. That doesn’t sound important until you remember that 4chan literally brought back nazis because they thought ironic nazism was hilarious. And then all the closeted nazis used that to say “none of those people are being ironic, they all mean what they’re saying” to new, properly primed recruits.









  • Generative AI, as it is being built right now, is a dead-end. It won’t get much better than it currently is (markedly worse once the next-gen is forced to scrape data that includes AI generated data) and hallucinations are always going to be the reality for them.

    It’s why there’s this big push over the last couple of years to get these products to market. Not because you’re going to corner some burgeoning industry (though the hype definitely is designed to look like that), but because this is a grift now and you have to get the goods while there’s still goods to get. Need to recoup those R&D dollars somehow.



  • If the option is 5 people dying every day or 10 people dying every day, yes harm reduction says we should limit the amount of people dying every day. That way we have more time to save the people.

    I’m not a fucking lib, I’m a leftist, and I’m telling you that Trump is worse than the status quo. He only really started to fuck shit up towards the end of his presidency and that was because he only just realized he could change the rules and get rid of the bureaucracy that actually keeps our government limping along. Next time he, or anyone with an R next to their name, takes office they’re going to destroy anything of value left in our government so they can hoard the ashes. This will continue to be the reality of the situation until being a fascist is no longer politically viable.

    Yes, it’s a hostage situation. Yes, it’s royally fucked, but the last 4 years has seen an incredible shift in both public support of good policy and politicians promoting good policy. There is a light, there is good coming of all this shit, it just requires your support for the good that’s happening, and your disdain to be pointed at the bad. Don’t point your disdain at the “not good enough” and shrug at the unbelievable bad. That’s how things get worse.







  • Having read the article and then the actual report from the Sakana team. Essentially, they’re letting their LLM perform research by allowing it to modify itself. The increased timeouts and self-referential calls appear to be the LLM trying to get around the research team’s guardrails on it. Not because it’s become aware or anything like that, but because its code was timing out and that was the least effort way to beat the timeout. It does handily prove that LLMs shouldn’t be the one steering any code base, because they don’t give a shit about parameters or requirements. And giving an LLM the ability to modify its own code will lead to disaster in any setting that isn’t highly controlled like this.

    Listen, I’ve been saying for a while that LLMs are a dead end towards any useful AI, and the fact that an AI Research team has turned to an LLM to try and find more avenues to explore feels like the nail in that coffin.