[embedding-shape]

> AI development is concentrated in a handful of wealthy nations. How can we ensure the gains of AI are shared globally? We do not have a mechanism for this. It is an unsolved problem

Kind of ironic given almost every AI lab except the one you started and work for actually done model releases to the public, some more "open" than others, but still something.

Look around at what other companies are doing, Qwen/Alibaba seems to have found a pragmatic middleground where they keep the most powerful model variant closed source and only API-accessible, while other models are being released openly to the public, to the entire world in fact, and when the next model release comes around, the previously undisclosed model has now been superseded.

I wonder if Chris ever copy-pasted his writing into Claude and asked something like "Please review this honestly and give me raw feedback, and challenge every claim that is weak", seems there are more "not really reflective of reality" points than just the above.

[azrazalea_debt]

You have a good point, and it is important to point out. I'm sure Chris would respond that it is due to those incentives he was talking about that affect all AI companies.

However, I do wonder what the actual practical benefit of (let's say) older versions of claude having their weights released would be. If we're talking about the people of poorer nations, how are they going to use these? Aside the top 1% of those nations, no one there is likely to be able to run the model themselves. Sure, there could be companies that sell it cheaper than Anthropic but that still won't extend access to everyone. The average person, even in richer nations cannot afford a computer that can run claude sonnet 4 for instance.

On top of that you have what Anthropic gets made fun of for - one of their goals is to protect Claude from humans. They are the only AI lab that is showing concern for the AI's welfare. Now, it is debatable whether this is reasonable or not, but that concern would lead a company to be less likely to release their models openly.

[embedding-shape]

This is the biggest part that confuses me:

> How can we ensure the gains of AI are shared globally? We do not have a mechanism for this. It is an unsolved problem

Sure, Anthropic might feel like it "isn't worth releasing old stuff for free" or whatever, but to claim that this is an unsolved problem, or there is zero mechanisms available for sharing the usefulness of AI is straight up bullshit, or said by someone who didn't take five minutes to look around themselves to see what others are doing.

> The average person, even in richer nations cannot afford a computer that can run claude sonnet 4 for instance.

Currently this is true, absolutely, since everything Anthropic is doing is closed-source, compared to OpenAI or pretty much lab out there. And who knows what would happen if they actually made Sonnet available to the public, might be there are resource-constrained people out there who could make it run on smaller hardware, we've seen that happen time and time again. Even the smaller models released today were unthinkinkable to have running locally just a year ago, I'm sure there are a lot of nifty people out there that simply don't have access today, and not because of the hardware.