Imagine if Github said "if we detect you're building a competitor to Github, we will silently degrade the results of your CI actions so that tests sometimes randomly fail"
I've been wondering when we'll see it unambiguously showing up in there. I suspect this time next year it'll be visible for sure, maybe Q4 of this year?
At least here in SF the ideal thing would be that any vehicle dropping off in the bike lane gets fined or ticketed. This includes Waymo, Uber, cabs, personal cars, whatever. In practice it's very rare to get a ticket for this, which is why customers expect it from both Waymo and Uber.
This is super cool and exactly what I've been looking for for personal projects I think. I wanna try it out, but the "agent" part could be more seamless. How does my coding agent know how to work this thing?
I'd suggest including a skill for this, or if there's already one linking to it on the blog!
Saying nothing about the actual performance of this model, it does strike me how .... minimal(?) this announcement is. Their safety section is like 2 paragraphs about bioweapons. Go look at the reports for OpenAI and Anthropic's model releases. It's like 50+ pages of tests, examples, reports, and benchmarks across a bunch of safety and wellfare metrics.
If Meta wants to be seen as a cutting edge massive lab they need to come across as one instead of looking like a school project version of a frontier model.
I wonder if this will actually be why the models move to "neuralese" or whatever non-language latent representation people work out. Interpretability disappears but efficiency potentially goes way up. Even without a performance increase that would be pretty huge.
I just can't find myself summoning the energy to be mad about markdown. It's good enough for like 99% of the things I use it for. Sometimes I get annoyed at specific extension support or whatever when I realize I shouldn't be using markdown for that task.
> The devices are either dangerous, or they're not
That's not actually how it works though, it's all a risk and percentages. Nobody says "driving is either safe or it's not" or "delivering a baby is either safe or it's not"
I’m curious if you think viewpoints have also gotten more extreme in this period. It feels like the gap in political ideologies has widened a lot since I was younger.
Yeah as long as the chatbot is empowered to fix a bunch of basic problems I'm okay with them as the first line of support. The way support is setup nowadays humans are basically forced to be robots anyway, given a set of canned responses for each scenario and almost no latitude of their own. At least the robot responds instantly.
Nice, I like the idea. It sounds like qualitatively you haven't had any performance regressions while doing this, but have you tested it at all on any sort of benchmark or similar eval? I'm curious how well the actual system performs with less context like this. I mean it's possible it actually improves...
Another interesting thing here is that the gap between "burned out but just producing subpar work" and "so crispy I literally cannot work" is even wider with AI. The bar for just firing off prompts is low, but the mental effort required to know the right prompts to ask and then validate is much higher so you just skip that part. You can work for months doing terrible work and then eventually the entire codebase collapses.