Viewing profile — NitpickLawyer

> My concern is what a misaligned model will do when they’re even more competent. I think the alignment talk is a red herring. It won't matter in the end, because there will be (if…

They do address some of these things in the final slides / "lessons learned" section of the defcon talk. Good security practices will continue to be good, but... and there are a lo…

It's not unexpected. Current model gains are mainly from RLing a pretrained model on lots and lots of scenarios. They have the models run scenarios, and RL on successful runs.

First via ssrf -> groovy plugin and second via specially crafted package I believe.

There's more details for the HF related incident in this [1] defcon talk. They said they're gonna have a proper post mortem at the end of the investigation (interesting to see if t…

> detect extremely tiny gravitational waves via an array of hyper-accurate clocks. I wonder how tiny is feasible. Would it get to a level of accuracy where you could detect neutrin…

> I wonder if it will ever be released in that form now. Yeah, I had the same thought. The key thing for them was the price point at which they could deliver a ~30B model. I would …

Back in the kazaa and limewire days, you'd sometimes try to get a movie / episode from a series, wait hours / days for it to download, and when it was done you had a ~50/50 chance …

> lifting ideas from other engines and copy-washing them. You cannot copyright ideas. And that's a good thing. Remember the "one click checkout patent" saga? That is bad for everyo…

1.2 is now on openrouter. You could try it there, they have hard limits per key.

> already rules that output of an LLM can't be copyrighted Mind sharing such cases? I'm not aware of any so far. There's the one with images, but that's commonly miss-understood, t…

> trying to run it locally or with your own hardware is pointless. Serving local models has advantages other than price. If you work in restricted industries, or have a strong need…

> nothing to do with socialism scaremongering What's "scaremongering" to you is "life" to GP. Besides missing their point, you're also saying "you held socialism wrong, we can make…

If anything this shows how pangram is unreliable. Chances are, if someone is running an article generation thing, it'll be 100% generated. Writing 3 / 100 articles "by hand" doesn'…

"Draw a pelican on a bicycle" is not a serious benchmark. "Draw an animation of this long ass scene from a movie, and only call me when everything works e2e" can be.

Heh, yeah. At least we're currently a bit more prepared for a Seveneves scenario than we were say 15 years ago. Speaking of prediction powers, there's one that stands out and turne…

> which is about as hard sci-fi as you can get. Interesting, I always thought that the modern interpretation is that hard sci-fi is "rooted in science" or close to it. The ansible …

> Is it the preferred way only because training is expensive? No, it's the preferred way because that's literally how you train it. Contrary to popular misconceptions, you don't "c…

Depending on how this data is being collected, a factor could be the rise of search / fetch / scrape capabilities, added in lots of places. Agentic this, google killer that, and so…

> Why did I need to know that? In a perfect world such a disclaimer would help anyone reading the output by setting expectations. We each have different experiences with this, and …

> We improve, we learn, we recalibrate our expectations based on what we've learned. That's not what people mean when they say "moving the goalposts". It means that people are adam…

(let's assume that)My 3dprinter is special. It has a bunch of values + an algorithm (i.e. a neural network) that takes input as tokens and outputs a printed object.

A better analogy would be a manufactured object, say 3d printed for simplicity. The 3d printer is given an input, and an object manifests itself after some time. We say that the cr…

> There is going to be a very wide variety of outcomes, all of which are going to coexist. Haha, I feel like this is true for a lot of things that people argue in the ai vs. non-ai…

I'm not convinced. Both are improvements on something implemented by highly trained/credentialed/successful teams. Something "missed" by either team is impressive either way.