Hacker Newshttps://xcancel.com/mkratsios47/status/2079933645888880708
Comments URL: https://news.ycombinator.com/item?id=49007610
Points: 226
# Comments: 583
Hacker News 讨论
226 points · 590 comments · 查看原帖
- himata4113
Does this matter? Distillation is not illegal by every definition of the word. There are millions of samples available on huggingface and models explicitely trained on output produced by fable. There has been no action taken against them. Another example is that it appears that the upper limit of what you can do is ultimately dependent on people working on the model, otherwise grok would be a LOT more competitive pre-cursor acquisition. And lastly, kimi architecture is vastly different than that of fable as it uses mechanisms developed by... kimi themselves. US AI labs are inspired by opensource advancements just as much as open source labs are inspired by traces from models such as fable. Claiming in any shape or form that fable disillation is one of the primary reasons why kimi k3 is so competitive is slandering the work of other labs that cooperatively push the open-source models forw
- throwa356262
Kimi K3 was released July 16, Fable ban was lifted on July 1 but access was still limited. How did Moonshot "distil" a huge model in such short time and still had time to run the benchmarks and do the usual release thingies? I think Anthropic is desperate to stop foreign competition and the administration is happy to help because they too are heavily invested in these companies
- madduci
So what is the issue here? Distilling is still fair, on the same level like Anthropic scraped copyright protected material for their training. So here robbers are blaming robbers? These claims are just pointless, everytime
- JumpCrisscross
"Samuel Slater (June 9, 1768 – April 21, 1835) was an early English-American industrialist known as the 'Father of the American Industrial Revolution', a phrase coined by Andrew Jackson, and the 'Father of the American Factory System'. In the United Kingdom, he was called 'Slater the Traitor' and 'Sam the Slate' because he brought British textile technology to the United States, modifying it for American use. He memorized the textile factory machinery designs as an apprentice to a pioneer in the British industry before migrating to the U.S. at the age of 21." https://en.wikipedia.org/wiki/Samuel_Slater
- sent-hil
Reminds of the quote by Bill Gates. > "Well, Steve [Jobs]… I think it’s more like we both had this rich neighbour named Xerox and I broke into his house to steal the TV set and found out that you had already stolen it." Source: https://www.goodreads.com/quotes/824084-well-steve-jobs-i-th...
- rustcleaner
Nice! I hope to see continued liberation of these locked up SOTA models. Cloud is a virtual prison, since other people's policies on what they think a user should and should not do cannot be [easily] bypassed, if enforced remotely on a cloud. All digital natives should be skeptical of cloud-hosted services or software. Think like an intelligence agency or sovereign: how are you going to get screwed by the cloud? Your data is fully accessible by the provider, and they can surveil your activities. You probably can't pirate it, so you are a slave in their rentier model. You could be prevented from doing something you want to do, because the provider disagrees philosophically or economically with your desire. You could be stripped of your information/data by a ban due to their policy enforcement system triggering. One should live by the maxim: you don't have the thing if you don't possess th
- nessex
Paying the list price Anthropic decided upon for outputs from an LLM, then using those outputs for your work? Or sourcing those outputs from others that paid for the service, and using those? If that's "distillation", it seems fine to me. And a lot like what most are doing with these same providers. Anthropic: training AI is "transformative", it's not copyright infringement if we don't re-transmit the copyrighted books we trained on Anthropic: training AI models on our outputs is stealing our secret sauce, outputs that could only be produced by us Anthropic: AI model outputs are unreliable and do not reflect Anthropic's views, we are not liable if they harm you (not exact quotes, they're "distilled") So Anthropic "distills" knowledge of others with reckless abandon, packages it up, sells it to you, claims it's your fault if anything bad happens but then also lobbies to treat you as a cri
- linkregister
Commenters are overlooking the significance of this information and posting emotional reactions based on perceptions of fairness or feelings of schadenfreude. The economic viability of Anthropic and OpenAI rely on their being able to charge more for model access than their R&D and inference costs. If the market price for SOTA model access drops below that level, then these businesses will have to decide whether to continue to lose money or to reduce spending on R&D. Moonshot's papers [1] claim that their training load was primarily from synthetic data and model self-teaching rather than RLHF and therefore keep their costs low. If Moonshot genuinely does not rely on human-led training, they will surpass US closed-source model providers. The United States government considers US supremacy in "AI" as a national security consideration. This announcement is noteworthy because it implies that