credit_guy
"Let someone"? I guess the right to pursue one's happiness is not all that self-evident after all. One should first ask permission, and, if we are in a good mood, we might "let them" pursue their happiness.
Google, who was left a little bit behind anyway, is more than happy to join the calls for taking a more cautious approach to AI development. Hence DeepMind Institute.
Sam Altman has no option than to pretend he agrees with all these ideas. What else is he supposed to say?
Take the Hugging Face incident. Why did it happen? Because the people whose task was to set up a testing framework took shortcuts. Why did they? Because there weren't enough people who were assigned to do the job. Why not? Because the job is too new and not enough people are qualified to do it. It's a job that simply did not exist 3 years ago. But 3 years from now, this job might very well employ tens of thousands of high skill knowledge workers.
"Money does not bring happiness" is a trope that rich people would like poor people to believe. I don't think poor people fall for it, though.
> If the HF hack were perpetrated by a human, they would certainly be charged. WHY has no one been charged???
I don’t think a crime was committed at this point. But I am sure HF’s lawyers are having a chat with OpenAI’s lawyers as we speak. And, being smart, they do that out of the public eye.
> The agreement also requires OpenAI to train its personnel on the INA’s anti-discrimination requirements, revise its employment policies, and be subject to departmental monitoring and reporting requirements, to prevent future discrimination.
Being subject to inspections from the Government to monitor your implementation of X or Y is not fun. I've been through that. The company spends an unholy amount of resources to make sure the results of those inspections are positive. The legal risk if the inspectors don't like what they see is simply too high. The punishment for the first offense of 3 mil is nothing, lighter than a slap on the wrist, but the second could be devastating, and the upper management does not like to entertain the possibility. And even if the consequences of a repeat offense are not entirely devastating, it's is extremely likely that a few people will lose their jobs, even if only as scapegoats. And people don't like losing their jobs. Those people in high positions that are at risk of losing their jobs will direct lots and lots of people under them to spend lots and lots of time in lots of meetings, writing lots of reports, just to make sure the inspectors are happy.
The first is the compute. OpenAI and Anthropic secured huge amounts of compute, Google, Meta and xAI have their own huge datacenters. Now anyone can rent some cloud machines and start serving Kimi K3, but it's going to be impossible to get to a similar scale as the big 5 above. And inference has economies of scale: the more people you serve in parallel, the more efficient you are.
The second is the data. By now (and maybe even by one year ago), all the data on the internet has been used for training. You need new data. The big AI companies sit on top of trillions or quadrillions of tokens that they have generated over the years. They can use that to train new models. That data is gold, and the proof is that SpaceX was happy to pay $60B to acquire Cursor.
If you want to overtake the frontier labs, you have 2 options: use their models to generate synthetic data, and provide lots of (cheap, maybe below cost) inference to generate your own new data. The frontier labs know about the first, and I'm sure they try to limit how much others milk their models. As for the second, that's the "honest" way to compete, but it's not easy.
Like many proponents, I'm not claiming the problem does not exist. I do think it is minor, although not insignificant. But I think you (and in general nuclear opponents) claim that the problem is not only significant, it is a deal breaker.
Decommissioning a gigawatt scale nuclear reactor costs an average of $1 BN in the US. Holtec International specializes in this work [1].
So, I did Step 0 in your list: admitting there is a problem.
Would you mind proceeding to Step 1: admitting there is a solution?
> And yet, they’d still have to keep forking out money even if Anthropic’s Claude chatbot or OpenAI’s equivalent succumb to competition from cheaper, open-source AI models. That’s not impossible.
The competition still needs datacenters, regardless of the open or closed source nature of the models they serve. They will need to buy the compute from somewhere, and they will have so other place to buy it from but the hyperscalers.
https://www.eia.gov/dnav/pet/hist/LeafHandler.ashx?n=PET&s=W...
Imagine Kimi K4 will be as powerful as Mythos. Anthropic can work for months and months to set up guardrails on Mythos, so when the model is finally released, it will generally decline to help hackers develop and prosecute cyberattacks, and if they do, at least there would be a trace so the law enforcement can track the perpetrators. Let's now say that Kimi K4 is released after a similar effort to develop guardrails. But being open weights, someone can just take the model, and finetune it until it does not refuse to assist in developing cyberattacks, and moreover, those people can run the model on their own private GPU cluster, so nobody can track the attack back to them. The situation is actually worse than that, most likely. Guardrails might be just markdown documents which are added to the context like regular skills. Then removing the guardrails for an open weights model does not even involve any finetuning, just removing some docs from a harness.
LLMs are becoming so powerful that they are dangerous. We've seen last week with the OpenAI hacking (by mistake) Hugging Face debacle.
It is absolutely ok to have open weight models at the level of GPT-OSS-100B. That one was released one year ago, and I think it's still a strong one. GLM 5.2 is a whole new level, but it appears to still be safe. Maybe Kimi K3 will be ok too. But beyond that, things will start being dicey.
It's easy to dismiss this and claim that Dario Amodei is just looking to fatten his pockets. And, sure, if Anthropic manages to put the brakes on open weight models, that reduces the competitive pressure it feels. But that does not make what Amodei's argument incorrect.
> There is certainly no shortage of funds or resources --- in fact, all these are readily available [...]
> Within our financial capacity, it's undoubtedly true that the more cards are always better. Our current strategy is to purchase as many cards as possible at a reasonable price --- exactly how many we can afford after using this funding round. The spending pace isn't predetermined; we'll buy whatever is available as long as prices remain competitive. In fact, I'd consider that a positive outcome if we spend the entire amount within six months. [...]
> In reality, spending such a large sum is no easy task: you can't obtain enough cards, they're hard to come by [...]
> Therefore, our only concern is whether we can obtain enough cards. If converting all funds into cards were feasible, we would undoubtedly do so without hesitation and are even willing to pay a premium for this benefit --- it's simply to cost effective. Even after paying the premium, however, achieving this goal remains challenging.I am also guessing that the majority of the people who read this title will think that DeepSeek is pausing this fundraising because some comments they made about the compute gap were leaked. That is not the case.
Fable 5 -> 0.42
Opus 4.8 -> 0.45
Sonnet 5 -> 0.45
Opus 4.7 -> 0.46
Grok 4.3 -> 0.52
There's an obvious jump at Grok 4.3, and it would not surprise me that the similarity there is because Grok used Anthropic models for training too (you can get the similarity list for Grok and it does look like the top most similar models for Grok are either Anthropic models or some Chinese models).
The damning evidence that K3 used Anthropic models for training is that K3 is more similar to those models than it is to K2.6. If you look at the Anthropic, OpenAI or Google models, they are most similar with their own other models. Not so with K3, where K2.6 is less similar than 15 other models.
Now, why is Fable 5 the most similar to K3 and not Opus 4.8. I think it's quite likely that K3 did some fine tuning at the end, when Fable 5 became available. They probably had all the infrastructure in place, and Mythos had been announced for months, so they were probably waiting for the second the newest Anthropic model was released to start using it for synthetic data generation.