agnosticmantis

Incredible results. How can we know that the modifications were not existing on the internet?

"Now let's see if Astra can write a neural network from scratch in zero-shot."

This strongly suggests retrieving pre-trained weights from memory, no?

How are other frontier shops handling this? Are Astra and Carbon/Argon also this anticompetitive? Or are they more chill about AI/LLM research?

Wouldn't breaking the news before the deal is sealed cause the price to blow up?

It'd attract all sorts of billionaires to bid on it and make it unaffordable for the public.

(The story seems to be old though, so hopefully it did end up being sold to the public.)

Which mathematicians though? Only fields medalists? Grad students? Any hobbyist wanting access?

These models are too expensive for broad access unfortunately.

Long term this will be the only reasonable author list: Chad G. Peter {1}, Mat H. Lean {2}.

1: Author 2: Verifier

/s

Any user scripts that automatically redirect to a working instance?

If Teslas could be appreciating assets [0], why not electric planes? /s

If anything, planes have real autopilots!

[0]: https://www.businessinsider.com/musks-claim-teslas-appreciat...

Good opportunity for a Natural Experiment to study the productivity impact of LLMs.

Presumably AI is going to process the images and detect issues and suspicious changes.

On paper it looks like a solid application of vision models.

> Anthropic is ramping up its efforts very quickly in biology and medicine

I hope they have better sandboxes in their biology lab than they did in their cyber security lab.

More seriously, these sociopaths will soon release lab-made viruses into the world to prove AI's capability and to force governments into regulation of open models. It should be obvious by now that with their savior complex, the goals (cure cancer, AI safety, save humanity) justify any means.

Very cool indeed, your response makes a lot of sense!

Wondering if the hallucination also occurs to other animals?

Would dogs taking the L. asiatica see little humans or little dogs? Probably neither, but then does that mean the fungus had targeted humans genetically?

We need more and stronger open source/weight models considering how deeply and intimately these bots are going to be integrated in our lives.

I hope that Dario Amodei fails in his quest to regulate open models out of existence to line his pocket under the guise of safety. Amodei/anthropic will end up being the most harmful force in the next few decades where progress in AI is concerned.

I hope people realize sooner than later that this is a replay of early Microsoft vs open source situation and Amodei is the new Gates on a crusade against open source/Linux.

Devil's advocate: if demand is constrained/finite in a given market, there's an upper bound on revenue, no matter how hard you try. In that case it'd make sense to work on efficiency to maximize profits given fixed revenue.

(Not saying AI helps with that yet.)

Looks like HackerNews has also started showing ads.

(Or maybe there's a /s missing?)

A person's total luck is constant over a lifetime. The remaining half of the candidates already spent some of their luck in this selection, so they'll be on average less lucky than the discarded half.

IIUC, this experiment proved the agent was secure under the "anti-prompt-injection" rules. But did it have any utility? (i.e. not having an agent at all would be even safer!)

Would MBP M5 32GB RAM be adequate for small model post-training and small RL experiments?

Planning to snatch one from BB/Amazon.

Any policy/weapon/tech used against an outgroup (e.g. brown people across the ocean) can and will eventually be used against the ingroup (citizens voting for it) by an inner ingroup.

Sooner or later any ingroup will itself be divided into an inner ingroup in power and a powerless outer ingroup who'll be treated just like the old outgroup.

As long as Tesla doesn't insure their cars (whenever in self-driving mode) completely for the price of their FSD, it's not FSD, just scammy marketing by a chronic charlatan.

> ... will eventually cause trouble for accounting.

Accounting has never been a concern in a Musk company or to his shareholders, as is evident from SpaceX IPO. People invest in him because he can make "number go up" regardless of profit and loss.

He's not running conventional profitable businesses, but ponzi-esque growth stories that keep ballooning to no ends without ever reaching a climax.

Nobody cares today what he promised 5 years ago (e.g. Tesla owners making $35k/yr passive income from their cars robotaxi'ing) because they're looking at what he's promising for the next 5 years (probably humanoid robots on Mars with an infinite TAM).

Counterintuitively, this is a huge win for misAnthropic and other closed labs in the US. They can nerf the models, ask for IDs from users and do what it takes to comply with whatever regulation they've been fighting for.

Foreign labs releasing open source models won't be able to comply, and as a result open source models will remain stunted at pre-mythos levels or their use will be criminalized.

We should look past the petty fights these closed labs have, and see their common interest in banning open source and/or local models.

Ladies and gentlemen, make no mistake: they're coming for your open source models next!

Dario must be popping a champagne tonight as regulatory capture has successfully been initiated.

We've all been debating what moat these labs possess. Today we learn the moat is regulation blocking the usage of foreign open source models.

misAnthropic will soon ask for IDs to give you all a nerfed model, and you have no other choice because deepseek/qwen/Kimi will soon be banned thanks to misAnthropic's efforts to lock us into their regulated product.

We will soon realize the long game that Dario has been playing by painting his own product as an existential risk.

Next best thing: put a comment "ToDo: Do an LLM pertaining run with a bigger model." in the malicious code, as misAnthropic censors LLM developement too.

Let's all vote with our wallets and collectively boycott misAnthropic or at least their feeble fable safety theater.

Whining on social media only goes so far, especially when they're concealing their anticompetitive strategies under the veil of safety.

I can't wait for a (likely Chinese) lab to casually drop an open model stronger than mythos sans all the safety theater that these clowns like to enact to justify the anticompetitive behavior and the impending enshitification.

This is another "gpt3 too dangerous for the world" moment which is laughable in retrospect.

Governments need to stop contracting these companies and instead invest in public, fully open source models.

These companies are owned and operated by the darkest of dark triads our species has managed to evolve. I doubt Dario is self-aware enough to realize the hypocrisy in all of this safety theater.

Personally I don't even mind that they are anticompetitive and power-hungry (same as it ever was), but it's the cringe-worthy hypocrisy that grinds my gears. This new brand of self-righteous paternal savior overlords is just unbearable.

> we’ve implemented new interventions that limit Claude’s effectiveness for requests targeting frontier LLM development (for example, on building pretraining pipelines, distributed training infrastructure, or ML accelerator design)

Translation: we stole the entirety of human knowledge generated over millennia. You plebs though, don't you dare replicate or improve upon what we did using our product you pay for.

We know what's good for humanity and everyone else is the bad guy who can't be trusted with a tool.

Singularity for me but not for thee.

Would you elaborate why she's considered a legend?

She was certainly a victim and a refugee, of which we have many today, most being denied admission to the US. Are those all legends too?