Right? It’s a standard color pass. People acting like it has to be RAW/LOG to be real.

352413, I mean, clearly.

I dunno about advice, but LLMs are very good at re-stating my meandering thoughts in a concise way that’s easy to communicate to others.

World Incarceration Rates If Every U.S. State Were A Country

Joe Rogan is Joe Garrelli in News Radio

The single CD amongst all that analog media seems so much like an anachronism that I had to look it up. The CD standard was published in 1980 and it was commercially available in the US in 1983 but it took until 1992 for CD sales to surpass cassette tape sales.

Thank you. I was just pushing the up arrow a hundred times.

Bonsoir, Don.

It shows which US states contribute more to the US incarceration rate and clearly shows that even those that contribute the least are above the majority of the nations’ incarceration rates. The latter is not obvious without visualizing the data in this way.

Politicians attempting to ban AI regulation are only shielding OpenAI, Microsoft, Google, etc. Those same politicians will pearl clutch about Hugging Face or DeepSeek in the same breath. It’s just about money and the arms race. Abdicating responsibility to AI would require either a change in how we regulate people or a granting of rights/responsibility to AI. Though I don’t doubt we will see people try to blame AI for their poor decisions.

People always askin Jesus to take the wheel, like he knows how to drive a car.

Hmm. Are intrusive thoughts memes struggling to be born? Not in this case, I hope.

That was in Anthropic’s system card for Claude 4, and the headlines/articles largely missed the point. Regarding the blackmail scenario, the paper even says:

… these extreme actions were rare and difficult to elicit, while nonetheless being more common than in earlier models. They are also consistently legible to us, with the model nearly always describing its actions overtly and making no attempt to hide them. These behaviors do not appear to reflect a tendency that is present in ordinary contexts.

They’re testing alignment hacking and jail-breaking tactics in general to see how the models respond. But the greater concern is that a model will understand as part of the context that it is being tested and behave differently in testing than in deployment. This has already been an issue.

In the initial implementations of reasoning models, if an LLM was penalized directly for this kind of misaligned generation in its “scratch pad,” it would not alter its misaligned response - rather it would simply omit the misaligned generation from the scratch pad. In other words, the model’s actions were no longer consistently legible.

Nuke it.

Activating AI Safety Level 3 Protections

I’m lucky that conversational AI is being developed as I’m middle aged, cause younger me absolutely would have fallen into an intractable state of delusion. Much like advertising, even the absurd cheese has an effect with extended exposure. And below the “You’ve hit on something uniquely insightful that could change the world!” shtick there is already a subtler form of reinforcement and enabling. This puts me in an odd place, because I use AI productively on a daily basis. And I still see it as one of the few technologies that could actually help us dig ourselves out of the enormous hole we’ve dug. But I suspect we’ll just use it to dig a deeper hole at a swifter pace.

Is that functionally just seceeding?