Claude Interpreter: Taking Safe AI to Market with Alex Albert of Anthropic

**Nathan Labenz** (0:00)
Hello, and welcome to The Cognitive Revolution, where we interview visionary researchers, entrepreneurs, and builders working on the frontier of artificial intelligence. Each week, we'll explore their revolutionary ideas, and together, we'll build a picture of how AI technology will transform work, life, and society in the coming years. I'm Nathan Labenz, joined by my co-host, Erik Torenberg. Hello, and welcome back to The Cognitive Revolution. Today, I'm thrilled to share an in-depth conversation with Alex Albert, developer relations lead at Anthropic, the company, of course, behind the groundbreaking Claude language models.
Alex first appeared on the show over a year ago when he was best known as the creator of jailbreakchat.com. Since then, he's joined Anthropic, where he's contributed to the testing and characterizing of Claude 3 prior to its launch, and is now playing a significant public-facing role as the company takes Claude to market and attempts to explain what makes it special.
Among those of us who use AI models intensively, Claude 3 Opus has quickly become a special favorite, not only for its exceptional writing abilities, but also for its ethical, aesthetic, conversational, and even imaginative sophistication. It really represents a qualitative step up in language model behavior in a way that benchmarks don't fully capture. It's simultaneously very free, willing to express opinions, and even allowed to argue for its own personhood, and yet it still remains extremely ethical. It can still be broken with certain adversarial tricks, but I have never once persuaded it to violate its ethical principles with any sort of normal language.
In this wide-ranging discussion, Alex shares a bit about Anthropic's philosophical approach to developing Claude, including what it means to be honest with a language model. We run down some of the most popular use cases for Claude. We discuss the ongoing convergence of API offerings and what might change that dynamic. We get some of Alex's top tips for application developers, both in terms of multi-model orchestration and also moderation to prevent abuse. And we get a glimpse into some of the upcoming research and product releases, which will soon be out from Anthropic, which I have not heard anywhere else.
It's worth noting before we begin that Anthropic is a company that I personally deeply respect. The founding team came from OpenAI for reasons that are once again in the news this week, and they have made a concerted effort to infuse the AI race, intensifying, though it is, with a race-to-the-top dynamic, especially when it comes to responsible development standards.
Their interpretability work, which we've covered before and will certainly cover again, is particularly impressive, and their responsible scaling policy really set a new standard for the industry, which OpenAI and DeepMind have subsequently followed. The company as a whole obviously believes in the value of their product, but also grapples very seriously and earnestly with the challenges and uncertainties that still lie ahead as AI becomes increasingly powerful. That is very much the spirit that I try to bring to this show and to my work overall, and Alex certainly exemplifies those values in today's conversation as well. If you're finding value in the show, I always appreciate it when listeners leave a review on Apple Podcasts or Spotify or just post about it online. We've recently hit 10,000 subscribers on YouTube, and we now get roughly 10,000 downloads per episode across platforms, and it is 100% due to word of mouth. Of course, your feedback is always welcome too. I've definitely heard your comments regarding audio quality on a couple of recent episodes, and we will strive to be better in that regard. For anything else, please feel free to contact us via our website, cognitiverevolution.ai, or you can always DM me on your favorite social network. Without further ado, here's my conversation about Claude 3 and more with Alex Albert, Developer Relations Lead at Anthropic. Alex Albert, Developer Relations at Anthropic. Welcome back to The Cognitive Revolution.

**Alex Albert** (3:54)
Nathan, thanks so much for having me on here again.

**Nathan Labenz** (3:57)
I'm excited about this. You guys have really made a big splash on the scene with Claude 3 and you personally have had a wild year, I think, going back to when we last talked, you were responsible for jailbreak chat, and now you're leaving and building out the developer relations function, so a lot going on in the AI space, as is always the case. I'm excited to talk to you about Claude 3 and all the uses that you're seeing for it and get into a little bit of how Anthropic is competing in this increasingly intense LLM space and maybe even get into a little jailbreaking if we have time. How's that sound?

**Alex Albert** (4:33)