@tautologer it is the first publicly available model that i am explicitly not allowed to use for my work, because anthropic holds the view that the work i do to facilitate open model research is harmful. capability and alignment research are coupled. anthropic wants to be the only lab.

I'm confused why so many are upset and surprised that they can't use Claude anymore to catch up with Anthropic - or at least they won't get the same benefits (it is not clear how much worse the model is vs Mythos for frontier RE tasks)

The safety case is equally strong as is the competition case

Open-source AI is inherently unsafe and one def doesn't want to enable an RSI loop on top of open-source models any time soon

Also people in AI should be happy that it means their jobs are safe for a bit longer

Lastly this is consistent with everything they have done all the way back to GPT-2's release

Claude Fable 5 is the first ever officially Decel model

Literally a Sophon that will sabotage

@OpenAI please mog them ASAP

welp my vision here was probably wrong and indeed there will be an extreme asymmetry of outcomes

The public Mythos is the first model that is deliberately designed to perform worse on technical tasks than its predecessors. AI companies will have models that help them to do research, but they will increasingly block others from benefiting. Not your weights, not your brain!

@BlackHC It's very very very unsafe. Absolutely disastrously unsafe.

You shouldn't be surprised that people are horrified about this.

really good model though from my limited testing it’s made me smile a few times. Big Model Smell

Quite a few people are very disappointed, to say the least, about Anthropic’s limitations on its new Fable model for AI researchers.

@EnoReyes has the best post about it.

As gross as this is, at least they disclose it. The infamous Gemini model that made images of diverse nazis also used hidden prompt modification, and in a way that affected way more outputs, but that was before these kinds of disclosures became normative (thanks in part to Ant)

By far the most serious problem with this release is the silent ML research nerf, which is incredibly unethical and makes me update negatively wrt my expectations of Anthropic leadership behavior a wide range of plausible ethically challenging situations

I’ll be honest that it would have been much more difficult to defend Anthropic against the DoW incursion had that incident occurred after this one. This is the company literally telling their customers, “we reserve the right to silently sabotage you.” I’d still have defended them, because the government trying to destroy a firm is still wrong, but man would it have been a harder case to make.

if only vetted institutions (big labs, governments, large enterprises) get unrestricted frontier capability, especially for AI research itself, those players compound their lead while everyone else works with a capped tool. The gap becomes self-reinforcing

We were going to build models anyways, but honestly this feels so petty and annoying. Now I just feel driven by spite.

it’s “self-interested” in that i picked my side intentionally because i believe it is the right side to be on. i’ve had plenty of opportunities to join the other side. i’m picking the side with less capital because i want to live in a world where this stuff is open.

the models in the competitive world are less powerful and more controllable. you aren’t going to convince me to trust dario that he’s gonna make sure everything is okay.

@tenobrus Come the f*** on man It's not about RSI It's about reducing the speed of your competitors stop buying that horseshit

This is why open-source AI matters. If the tools for building the future can be silently throttled by the same few labs competing to own that future, that’s not democratizing intelligence instead it’s building a new colonial infrastructure.

@natolambert Next obvious stage seems to be model actively fudging experiments and introduce subtle bugs so no one else can do LLM development. Feels like Monsanto’s terminator seeds.

Stated vs revealed intent of safetyism.

QED.

@willccbb @tautologer how dare you help people build better product sir

it’s a tweet. “the work i do to facilitate open model research” implies i believe that open model research is a good thing. i do. you know this. we’ve talked about it before. you’re being pedantic.

the open model world is a race to the bottom on price. this is good for limiting the pace of capability gain.

the most antisocial and competitive company in this race is anthropic. they are frequently alone at the frontier. they do not collaborate. they cut off anyone they see as a threat to their dominance or pricing power.

i trust jensen, then demis, then elon, then sam, then dario.