normative

Alignment must cover continuously updating weights

Current alignment research mostly addresses how frozen weights behave in deployment, but labs need new work to guarantee that a continually updated model avoids jailbreaks, deceptive or evil personas, and backdoors injected by users.

normativepending

Speaker

Dwarkesh Patel

Evidence Quote

“how do you prevent users from injecting backdoors or some kind of malicious inclination into the base model?”

Source

Dwarkesh Patel Podcast - 8 Predictions for the Era of Continual Learning— Dwarkesh Patel Podcast
Created: 11/17/2026, 6:25:01 AM

Your notes

Loading notes...