Viewing profile — piker

So an agent was somehow able to manipulate internal OpenAI infrastructure, albeit perhaps temporarily. It makes me wonder if OpenAI infrastructure is so littered with verbose AI sl…

I see that now. I've been confused about that to this point, I guess. I understood this to be a specific infosec exercise. [Edit: eh, a bit of both. They were doing RL on a hacking…

Why do these super agents need package managers anyway? Can’t they basically instantiate most OSS projects from scratch anyway? Spin up a sub agent to write me an OS interface in C…

Only one of those cards was printed in the last 4 years (and it was planned 3 years ago). The shift is somewhat more recent than that. If they had printed a Ninja Turtles set for M…

It's sad watching Magic: the Gathering shift away from nerds towards mainstream profits. I wasn't a fan of Marvel growing up, but I wonder if there is a similar sense of loss. MTG …

I’ve deleted it but no it’s not awful anymore than saying “we survived WWII, we can survive this.” The point was that change happens but humans find a way forward.

I don’t feel the existential dread of mathematicians is correct. It seems to me in fact these results are bringing math mainstream. I now personally look forward to the interpretat…

White text still works! There are many approaches today. Check out https://tritium.legal/blog/noroboto where we tricked frontier algorithms into reading different Unicode values fr…

Nassim Taleb has a note about this idea in one of his books. Essentially just pointing out how unimportant most news headlines are a few years out.

The article is paywalled, but isn’t he Canadian? Who cares what he thinks.

> Not all software needs to be commercial to be useful, even if it's just for a learning experience. I have learned a lot from those experiments, even if they're not polished. This…

> "As far as I can tell they..." This phrase I think highlights the fundamental issue for a lot of folks that would otherwise consider adopting a project like this written by human…

Okay, but then that does not square with calling a GPT implementation out-of-distribution, let alone "extremely" so.

How is a GPT implementation “extremely out of distribution” but “planet scale infra” isn’t? That’s got me totally confused about your point that I was taking seriously.

It doesn't seem Ukraine's end goal is for Russians to be happy about it.

Hard to see them though if you’re not in the weeds and your agents are stacking abstraction on abstraction to pass tests.

"The pursuit of excellence does not need justification." https://x.com/mitchellh/status/2074225453217505494

I don't trust him or Dario frankly, but I don't run a Fortune 500 company. They obviously do.

> Why are they charging 1000x more for the exact same business? The implication they want to avoid, I'm suggesting, is because the CCP isn't plugged into and funding those entities…

In that case is there not some open harness that demand can coalesce around?

For fear that the data they're entrusted with will be copied or that there is some nefarious reason like state sponsorship that makes the activity economical?

The baker is not asking you to trust them with your entire knowledge base in exchange for an apple pie.

I do agree with the UX idea of familiarity, but copying the exact salutations, etc. just seems egregious. Why not let the model speak for itself and contribute to an open harness?

Perhaps, but "we run the exact same business as Anthropic and OpenAI to the pixel at 0.1% the price" is going to raise eyebrows.

I don't think the comment was intended to be pedantic, but rather to reframe the message as scientific fact.