Dario Amodei: Anthropic CEO on Claude, AGI & the Future of AI & Humanity | Lex Fridman Podcast #452

Another useful word in this context is “sycophancy,” meaning excessive flattery or insincere agreement. Amanda Askell of Anthropic has used it to describe a trait they try to suppress in Claude: ▶ this video

What I mean by Zeitgeist is this: once an event begins, it becomes unstoppable. The most classic and cliché examples include Galileo’s heliocentric theory and the Inquisition, or Martin Luther initiating the Protestant movement. Some ideas, once they start being built upon by certain individuals or institutions of that era, continue to develop in that direction if they achieve success. That’s why I say, "Zeitgeist predicts it this way." Researchers who have laid down important cornerstones in this field (e.g., Ilya Sutskever, Dario Amodei, etc.)[1, 2] suggest that this is bound to happen eventually, one way or another. Beyond that, most of the hardware developments, software optimizations, and academic papers being published right now are all focused on this field. Even when considering the enormous hype surrounding it, the development of this area will clearly continue unless there is a major bottleneck or the emergence of a bubble. Many people still approach such discussions sarcastically, labeling them as marketing or advertising gimmicks. However, as things stand, this seems to be the direction we are headed. [1] ▶ this video [2] link

It's been a bumper week for interesting podcast interviews with an AI theme! In addition to this, there are Lex Fridman's series of interviews with various key people from Anthropic [0], and a long discussion between Stephen Wolfram and Eliezer Yudkowsky on the theme of AI risk [1]. 0. ▶ this video 1. link

He has studied and influenced the real world. Here's Dario Amodei, the CEO of Anthropic, one of the leading AI companies, directly referencing him in an interview from this week: ▶ this video

> But I can tell the quality drops even when you do that Dario said in a recent interview that they never switch to a lower quality model in terms of something with different parameters during times of load. But he left room for interpretation on whether that means they could still use quantization or sparsity. And then additionally, his answer wasn't clear enough to know whether or not they use a lower depth of beam search or other cheaper sampling techniques. He said the only time you might get a different model itself is when they are A-B testing just before a new announced release. And I think he clarified this all applied to the webui and not just the API. (edit: I'm rate limited on hn, here's the source in reply to the below ▶ this video )