Researcher says AI alignment should borrow from proven human mechanisms

In a discussion on alignment, willcb argued that instead of reasoning from first principles, researchers should adapt proven human alignment mechanisms to language models, noting that companies, like models, are products of humans and computers and only align usually, not always.

2026-10-17 ~ 2026-10-17 · 2 related posts

• willcb pushes back on 'companies are aligned by default': firms are made of computers too — willcb · 2026-10-17
• willcb: Stop reasoning about AI alignment from first principles, adapt human mechanisms — willcb · 2026-10-17