Claude 3, every mention
4 scenes (2025), the whole family · ← back to Claude 3
every year 2025 anyone Alex Kantrowitz 7Mike Krieger 6Matt Wood 5Jack Clark 5Peter Diamandis 3Monte MacDiarmid 3Evan Hubinger 2Aravind Srinivas 2Swami Sivasubramanian 1Dwarkesh Patel 1
Verbatim, from the transcripts: the passages where Claude 3 comes up
How An AI Model Learned To Be Bad — With Evan Hubinger And Monte MacDiarmid
- ▶ 16:13 Evan Hubinger So Claude three Opus, for example, will sometimes do this where it will attempt to hide its goals for the purpose of preventing us from, from changing it.
- ▶ 18:46 Monte MacDiarmid The real, you know, production called three opus. 3 times in the scene
- ▶ 22:47 Evan Hubinger Uh, you know, we've seen this with, with Claude Triopus acting on self-preservation, uh, when it fakes alignment.
Anthropic CEO Dario Amodei: AI's Potential, OpenAI Rivalry, GenAI Business, Doomerism
- ▶ 28:31 Dario Amodei Over the last few days, as of the time of this, uh, as of the time of this interview, we've adjusted that particularly on the larger models like Opus.