I've been posting a lot of tools that I've been building, the target for these tools are independent researchers, Small/Medium admins, and those who don't have the resources/needs of a full large enterprise. Here is a breakdown of what they are and what they do

Slowly, the world is discovering the value of provisioning hardware-bound cryptographic credentials. The Internet is slowly catching up to the level of security of tapping to pay for a cup of coffee.

I've been telling people for 25 years that Jane Street is not interested in formal methods. No more! And we're actively hiring to form a new formal methods team!

kinda crazy that someone's full-time job was to steer claude to sabotage ML research capabilities for paying customers

I wonder how much of opus notably degrading in the run up to mythos was them being short on compute and how much was them testing this "degrade silently" process.

It's like Anthropic took a look at the idea behind the Fast16 malware and said "there's a business case here..." Truly awful, if you're going to guardrail something, guardrail it, don't just sabotage people's work.

If they can nerf LLM research and security, they can nerf whatever you use the model for. At any time, for whatever reason. Build on top of something like that at your own peril.

Fable really is something else, 10/10 Anthropic you guys are brilliant. I asked it to take a look at code that I'd been working on for my TLA IDE, it promptly ignored the plan file I asked Opus to write, decided I didn't want to be working on TLA , removed my code and then told me I violated a guardrail. Huge fan guys, I see why this is 2x Opus

Funnily enough, I was told by a notable mathematician that when doing research-level math, unless you are a generational genius, you kind of have to blackbox at least *some* results, or your progress will be way too slow….

I have started a collection of essays, blog posts, etc discussing AI in mathematics. I do not agree with everything written, but all are valuable to read - the more different views the better! Please reply with your own suggestions. thomasbloom.org/AIlinks.html

My general view has always been that it's not worth rushing to be the first to do work that someone else could do next week; clearly the marginal impact of such work is low. Worth thinking about this in the rush to publish results where the intellectual labor was done by AI.

Threat Intel Teams: Respectfully, what the actual HELL are we doing with malware names? InvisibleFerret? OtterCookie? PondRAT? KANDYKORN?? BeaverTail?? calm yourselves We're trying to write bangers on state-backed crime and you handed us the rejected Pokemon nobody picked. ☯️

Very excited to share our interview with @polynoamial on AI for math — the Erdős unit distance problem, saturating the IMO, the future of math research, and more!

My guess is that eventually AI has advantage over us in all aspects of mathematical research. But I think there are two possible worlds where this doesn't happen soonish: (1) AI exceeds our capabilities in all measurable capabilities, but there are still some (very intangible) intangibles where humans have higher capabilities. In this world we see AI making huge numbers of amazing discoveries but for some weird reason humans sometimes also do so. This could happen if there are some skills that are useful for math research but very hard to measure or train for. I think this is unlikely but possible. (2) Rather than AI exceeding humans in all capabilities, AI becomes more like *one* (or several) amazing mathematician(s). By which I mean, very technically strong but with a certain point of view, which makes them more likely to make certain discoveries than others. In this world humans are useful because of their greater cognitive diversity--they sometimes think of approaches/questions/ideas that don't occur to the AI. This happens now--there are mathematicians much stronger than I am but I'm still useful (plausibly for reasons other than comparative advantage) because I have some different points of view. I think this is closer to what the near-term trend looks like, but the medium-term is less clear. In the (IMO more likely, in the medium-to-long term) world where AI does have absolute advantage over us in all aspects, I don't think research ends. Plausibly we have AI producing huge numbers of papers, asking all sorts of interesting questions, and so on. Probably these questions do not include all questions we might ask, so there's still some role where we ask questions (even if the models are better at this, me still might care about questions they haven't asked)! And IMO the goal of science isn't just "papers," it also includes things like "understanding." Maybe the models have this in their weights or something, but I also want to understand stuff. So we have to extract this understanding from papers or model weights or w/ever. This might look more like hermeneutics than research currently does, but my experience is that trying to understand mathematics that someone else has produced is not actually so different from research. You end up doing a lot of work, developing your own intuition, and so on.

We can all spot the issue here, also we can all agree it would be really funny if this happened.

it’s wild that in a democracy it’s easier for a company to lobby making full disclosure an illegal act, than it is for the public to demand that knowingly selling insecure software for use in critical infrastructure an act of gross or criminal negligence

I do not know how to explain this, but as a cat parent, you quickly learn that “absence of cat (chill)” and “absence of cat (suspicious)” are two very different kinds of silence.

Vendors don't always know what their product would and wouldn't catch. But they do have a great deal of info about how it was built, the effort and care that went into it, the tradeoffs that were made, etc. They know where they cut corners and where they didn't. Buyers don't.

“I’m getting too old for this,” I whisper to myself right before doing something I’m too old for.