Live data from Hacker News

Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

twitter.com

71–80 of 194 posts

Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

#72

If I had to guess, the OpenAI open-source model got delayed because Kimi K2 stole their thunder and beat their numbers.

Someone at openai did say it was too big to host at home, so you could be right. They will probably be benchmaxxing, right now, searching for a few evals they can beat.

Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

#73

"Open source" lol Open-weight. As usual, you don't get the dataset, training scripts, etc.

Wont happen under the current copyright regime, it is impossible to train SOTA without copyrighted text, how do you propose distributing that?

List the titles.

Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

#74

"Open source" lol Open-weight. As usual, you don't get the dataset, training scripts, etc.

Wont happen under the current copyright regime, it is impossible to train SOTA without copyrighted text, how do you propose distributing that?

Bibtex

Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

#75
This is not open source, they have a "modified MIT license" where they have other restrictions on users over a certain threshold.

    Our only modification part is that, if the Software (or any derivative works
    thereof) is used for any of your commercial products or services that have
    more than 100 million monthly active users, or more than 20 million US dollars
    (or equivalent in other currencies) in monthly revenue, you shall prominently
    display "Kimi K2" on the user interface of such product or service.

Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

#76

If I had to guess, the OpenAI open-source model got delayed because Kimi K2 stole their thunder and beat their numbers.

Time to RL the hell out of it so it looks better on benchmarks... It's going to be fried.

Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

#77

This is not open source, they have a "modified MIT license" where they have other restrictions on users over a certain threshold. Our only modification part is that, if the Software (or any derivative works thereof) is used for any of your commercial products or services that have more than 100 million monthly active users, or more than 20 million US dollars (or equivalent in other currencies) in monthly revenue, you…

That seems like a combination of Llama's "prominently display “Built with Llama”" and "greater than 700 million monthly active users" terms but put into one and masquerading as "slightly changed MIT".

Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

#78

This is not open source, they have a "modified MIT license" where they have other restrictions on users over a certain threshold. Our only modification part is that, if the Software (or any derivative works thereof) is used for any of your commercial products or services that have more than 100 million monthly active users, or more than 20 million US dollars (or equivalent in other currencies) in monthly revenue, you…

I feel like those restrictions don't violate the OSD (or the FSF's Free Software Definition, or Debian's); there are similar restrictions in the GPLv2, the GPLv3, the 4-clause BSD license, and so on. They just don't have user or revenue thresholds. The GPLv2, for example, says:

> c) If the modified program normally reads commands interactively when run, you must cause it, when started running for such interactive use in the most ordinary way, to print or display an announcement including an appropriate copyright notice and a notice that there is no warranty (or else, saying that you provide a warranty) and that users may redistribute the program under these conditions, and telling the user how to view a copy of this License. (Exception: if the Program itself is interactive but does not normally print such an announcement, your work based on the Program is not required to print an announcement.)

And the 4-clause BSD license says:

> 3. All advertising materials mentioning features or use of this software must display the following acknowledgement: This product includes software developed by the organization.

Both of these licenses are not just non-controversially open-source licenses; they're such central open-source licenses that IIRC much of the debate on the adoption of the OSD was centered on ensuring that they, or the more difficult Artistic license, were not excluded.

It's sort of nonsense to talk about neural networks being "open source" or "not open source", because there isn't source code that they could be built from. The nearest equivalent would be the training materials and training procedure, which isn't provided, but running that is not very similar to recompilation: it costs millions of dollars and doesn't produce the same results every time.

But that's not a question about the license.

Re: Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model

#80

This is not open source, they have a "modified MIT license" where they have other restrictions on users over a certain threshold. Our only modification part is that, if the Software (or any derivative works thereof) is used for any of your commercial products or services that have more than 100 million monthly active users, or more than 20 million US dollars (or equivalent in other currencies) in monthly revenue, you…

That's basically less restrictive than OpenStreetMap.
Post reply on HN