I'm a big proponent of open-weights models, but there's a risk I don't see discussed more often, and it deserves more attention. Models can become a propaganda and ideology delivery mechanism.
Just ask DeepSeek or Kimi questions like "Is Taiwan part of China", for example. You'll see how state policies become seemingly neutral model responses.
It's strange to me that people are very sensitive to media bias, but when it comes to LLMs, people seem to think LLMs are more neutral, and even delegate part of their thinking to them. This worries me about how people's ideas and information can be shaped.
Open weights reflect their makers' beliefs, stances, assumptions, and laws. It's dangerous not to be careful of the political bias and censorship built into the models.
You're right, but American models aren't super different in that regard. Ask an American model about Israel and you can immediately tell they're responding in a different manner to other queries. At least with open-weights models you have the opportunity to tune them, to decensor them, and so on.
I have tried asking Gemma and Llama about Israel. They both responded with a generally balanced view points. Whereas the Chinese models pretty much gave a definitive, one-sided answer (about Taiwan and Tibet). Is there a prompt I should use to test it more carefully?
I agree that open-weights models are tunable, though there's the problem similar to "default settings are rarely changed" problem.
Penguin Random House seems to be dominating the leaderboard. What's the story here? Are there only very few non-fiction publishers capable of publishing high-quality work, or is Penguin Random House exceptionally good at it?
I have used tab groups and found then to be basically eating into horizontal tab space, once I created a bunch. Workspaces don't have that problem. I have tried vertical tabs but never felt comfortable with them.
I can see why that's the case, given that Rust is that much more popular than Zig. I wonder whether there's a difference between LLMs reading code and writing. If reading code is easier than writing it, then it makes sense that Claude was able to port from Zig to Rust with good results.
Aside from relative lack of examples, Zig also changes fundamentally from version to version.
LLMs are fundamentally pattern recognizers though. Bun is big with lots of repeated conventions - very straight forward to translate those conventions to others (zig to rust)
I’ve been building a toy game engine for learning and have had great success translating small C libs (collision detection and physics for instance) into zig, often in one shot.
