Travis

Current status: Retardmaxxing.

Emil Michael now appears to be making an argument that no generative AI should be used in the DoW supply chain (all uncertainties involving model sentience and general unpredictability are common to all language models, not specific to Claude).

To say a little bit more about why I think a lawyer would need to see the full contract to evaluate whether OpenAI's redlines are protected: • its not uncommon in contracts to have language that says "party A shall do x" • and then later language that says "in such and such situation, party A does not have an obligation to do x" and that supersedes the prior obligation. National security law as I understand it is particularly notorious for these kinds of escape hatches and unintuitive outcomes. So I think the additional language Sam shared seems good, but its really extremely hard to evaluate whether it actually does anything without the full contract (which is unlikely to become public). It could say those things, and then have later clauses which effectively allow them to be disregarded on interpreted in unintuitive ways. To use an analogy for engineer types (and sorry if I butcher this) - its a bit like trying to figure out what a larger codebase does where you only see a fragment of code or one specific function. You can talk about what that fragment does, but without seeing it in its entirety in context, maybe it operates very differently than how you might expect if it were shared in isolation. This is particularly true because unlike codebases, contracts are drafted adversarially, where the parties may deliberately be trying to make things work in unintuitive ways that give them more leeway. I realize this is not that satisfying, and overall I take the announcement that the terms were updated as some modest but positive news (and if the language shared is really in there and functions how you would intuitively expect it is good!). It would also be further evidence if the Pentagon made emphatic and specific comments to the effect that their interpretation was the same (so far the pentagons comments have largely been "we will follow the law, and there are laws against x y z"). If the Pentagon publicly acknowledged specific ways this contract constrained them above and beyond just a commitment to follow the law and department policy, that would make me more confident these concessions are real. Overall I've gone on for long enough in this post that I feel compelled to restate two things that I hope folks don't lose track of: • parts of this are complicated, but one thing that isn't complicated is the Pentagons retaliation against Anthropic for a contract negotiation is absurd and should be undone immediately. • In the absence of Congressional action these negotiations do matter and serve an important role. But it would really really be great if we had a Congress interested and able in helping to clarify some of these issues.

AI companies absolutely have to devolve power and defer to democracy. And to the extent that they want to influence policy, the right way to do it is through policy advocacy. I'd like to see Anthropic advocate for legislation on the areas they have proposed in contracts.

We’ve seen this movie before. When the dust settles, a lot of patriotic founders will point to this exact moment as the match that lit the fire in them. 🇺🇸

Rashida Tlaib: ‘Epstein's Not The Last Person! Follow The Money! Who Was Funding It?’

ICE is sexually abusing males in their custody by grabbing and bursting their testicles according to the ACLU and the guardian.

82 million American men buy Only Fans

My radical proposal is that the AI companies should share the fruits of their innovation... By allowing anyone with an Internet connection to have unlimited access to their choice of state of the art AI models on their own terms to the extent of their desires.