Hacker News: baxfedi

New comment by baxfedi in "Qwen3-Coder writes less secure code when told it's for a US persona"

baxfedi — Wed, 21 Oct 2026 14:37:49 +0000

A model that understands it might be wrong with the token it predicted after “a + b = “ and is open to double checking that with a calculator/script is a more intelligent model that is less likely to hallucinate confidently.

New comment by baxfedi in "The obscure trust looming over Anthropic's IPO"

baxfedi — Wed, 21 Oct 2026 14:31:02 +0000

I do love it, though, when the antisemites lump in frontier AI companies in their conspiracy theories. It has echoes of Nazi Germany’s banning of nuclear science as “Jewish physics” (or as they would’ve called it if open antisemitism were unpopular: the Einstein class).

New comment by baxfedi in "Qwen3-Coder writes less secure code when told it's for a US persona"

baxfedi — Wed, 21 Oct 2026 13:55:18 +0000

> the "conflict is very complex and highly debated."

Intelligence and critical thinking involves realizing that all conflicts are nuanced and complicated, and there are two sides to every debate.

Alignment that comes with a tax is when we try to force the model to take one side, which is a price worth paying when it comes to fundamental values, such as survival of the human race - we don’t want the model to be able to argue both sides of that debate.

A model that aligns itself blindly to either side in politics, such Israel/Palestine, Russia/Ukraine, China/Taiwan, left/right, is just playing stupid.