351–360 of 742 posts
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
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…
It is incredibly important to whether the US can maintain its AI lead. If foreign competition is closing the gap only by distillation, then the frontier labs can focus on preventing distillation and maintain their lead that way. US dominance is also important for approaches to safety, especially political approaches. If the frontier models are all US-based, safety might be tackled via internal US policy. If other cou…
I don’t think China’s necessarily any better, but I’d rather have the most powerful models be open rather than under the exclusive control of the US executive.
Earlier quoted context omitted.
While I agree on a moral level, I think there is a distinction to be made. Training a SOTA model takes a huge amount of resources and expertise so the people doing the training are adding a lot of value along the way. I think this is much less true for distillation (which is kind of the whole point). ed: to clarify, I totally agree that a huge chunk of the value in LLMs is coming from the source material. My point wa…
You can argue that reverse engineering anything is as hard if not harder than engineering something. I can’t imagine distillation is any different.
Earlier quoted context omitted.
Perhaps even more importantly, the current frontier LLM models are self-admittedly the product of enormous quantities of copyright infringement and even less savory inputs, so calling them out for distilling the fruit of that tainted tree reads as highly hypocritical at best.
While I agree on a moral level, I think there is a distinction to be made. Training a SOTA model takes a huge amount of resources and expertise so the people doing the training are adding a lot of value along the way. I think this is much less true for distillation (which is kind of the whole point). ed: to clarify, I totally agree that a huge chunk of the value in LLMs is coming from the source material. My point wa…
It's the most CS-major take ever!
[flagged]
Apparently we do, given that we've got govt officials wasting time, money, and effort on whining about this now.
Earlier quoted context omitted.
Stealing IP is how American industry got started. Goose: gander, pot: kettle. It is what built and sustains the movie and music industries. See: work for hire and 100+year copyright length The tech industry: see: copyright and patent assignment from discoverer to corporation. I know that corporations forcing me to assign patents and copyright to them was an incentive to take published works from "software practice an…
The things you listed are very far downstream of the start of American industry, which is in primary resource extraction and processing. Which is you know, what actually built the country. The media industry has always been materially irrelevant, and what we think of as the tech industry is extremely new. You're right to call out the nasty environment surrounding intellectual property in the US and the exploitation o…
Then there is Alexander Hamilton's advocacy for importing foreign technicians that bring back IP and reproduce it here in the states. Best of all was the patent act of 1793 which like with the literature copyright ignoring, let us citizens patent inventions from the other side of the pond.
The founding fathers definitely had the right idea on IP.
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…
It is incredibly important to whether the US can maintain its AI lead. If foreign competition is closing the gap only by distillation, then the frontier labs can focus on preventing distillation and maintain their lead that way. US dominance is also important for approaches to safety, especially political approaches. If the frontier models are all US-based, safety might be tackled via internal US policy. If other cou…
We already know it's false because you would have hundreds of competitors if it was that easy.
The reason why these Chinese labs are releasing good models is simpler, they have access to a tremendous pool of talented people.
Earlier quoted context omitted.
Perhaps even more importantly, the current frontier LLM models are self-admittedly the product of enormous quantities of copyright infringement and even less savory inputs, so calling them out for distilling the fruit of that tainted tree reads as highly hypocritical at best.
While I agree on a moral level, I think there is a distinction to be made. Training a SOTA model takes a huge amount of resources and expertise so the people doing the training are adding a lot of value along the way. I think this is much less true for distillation (which is kind of the whole point). ed: to clarify, I totally agree that a huge chunk of the value in LLMs is coming from the source material. My point wa…