For real? Anthropic sounds more evil that my wildest imagination. If true, I hope this kind of evil company gets destroyed sooner than later.
1. "We want to own it and be the only megacorp that exists in the future", and
2. "Proliferation of open-weight models is like giving everyone a nuclear reactor in their home: it is extremely dangerous to the world"
Anthropic has previously stated (or, at best, implied) that they believe/have-reason-to-believe the second.
Some other organizations, well the public has every reason to believe they believe the first.
The end result of "hardware shouldn't be capable of running non-validated models" is the same between the two.
I can't comment on which of the two Anthropic is (because I have literally never in my life talked to anyone in an influential leadership position there). Just keep in mind that even people who are diametrically opposed can have matching recommendations (again, for polar opposite reasons).
The problem with #2 is why should we trust them? If they truly believe #2 then they shouldn't exist either.
This is like the very first thing Dario addresses if you've ever read a word he's wrote. TLDR race dynamics make this inevitable, and it's preferable if we get to superintelligence before China
Remember their "LHR" GPU
Edit: well, maybe massively decreasing memory bandwidth?
The underlying premise from Dario is pretty much that they are the enlightened vanguard who can be trusted with power, while the public is too foolish to have autonomy.
But people will never converge on a single definition of virtue in the standard of Robespierre's ideals. Because people naturally hold conflicting values, Robespierre had to resort to the guillotine to force ideological conformity and erase dissent. You cannot enforce a "Republic of Virtue" without becoming an authoritarian gatekeeper. When Dario demands only certain AI models that we can use, his so-called "safety" stops being a precaution and becomes a tool to impose his values on everyone else. And that, makes him dangerous and disgusting.
Either through technical means, or simpler: by law for American providers. It's hard to believe the administration was not aware of Jensen's affiliations all this time. And time is pressing for Anthropic and OpenAI.
Only conjectures for now, time will tell
Makes you wonder what would happen if the US ever wanted to enforce "certified US AI" on its chips
Nvidia can certainly put up some guardrails that will sound impressive to the dinosaurs in Congress, but can it really stop China from using those GPUs anyway?
This is a simplification, but is this not in some way similar to how the Switch 1 has hacked due to a design flaw on Nvidia's Tegra chip?
In the same way, even if their GPUs get this protection, wouldn't a state actor figure out how to get around it?
Most of the hardware functionality from a GPU that models need is generic and used every time you use a GPU for any compute. You can implement language models in a GLSL shader; modern GPUs have been programmable for 20 years, that's the point of their existence. The best thing they could do is just literally nerf the consumer hardware substantially to make tokens unusable or stop selling consumer hardware. They don't need spyware to do that. It would also make their products dogshit.
More broadly, "Compute" is basically a resource that our economy relies on to a large degree, it has been that way for decades. And it turns out it's very bad when you just destroy the thing your economy relies on. And doing so doesn't solve problems like "Current computers exist today" or "Other people make computers". So who gives a shit about Jensen? Comments here talking about "taking him out"? What's actually more likely is just that really powerful people today are often corrupt and also really fucking stupid. He played a stupid game and lost.
None of this stuff is putting the current open weights genie back in the bottle IMO. Amodei using RoboStormtrooperGruppen9000 bots and going door to door to take peoples GPUs at gunpoint is a more plausible method of doing so.
Open-weight models, at an absolute minimum the ones that already exist, are here to stay, and will continue ([0]) to get drastically cheaper to run.
[0] I shall continue to reference one my favorite articles in recent months: https://jyn.dev/tokens-too-cheap-to-meter/
lol had to read this several times due to all the negatives
finally got it - "I hope NVIDIA continue to allow inference..."
It's sad that people talk here about some personal affairs when it's clearly about governmental control, but it's not the first time a scandal is used as a weapon.
This was a hit job coordinated by captured media and the people behind the NVDA board to work towards a tech oligopoly in AI. It's glaringly obvious, at least to me.
I mean, none of this "spy" stuff is proven, and the various alphabet agencies are backtracking the claim. That means it's fair game to compare Jensen's "scandal" to the "scandals" involving OpenAI and Anthropic's CEOs. Jensen is the patron saint of morality and virtue comparatively.