Here's a sentence you probably didn't expect to read today: there is a person whose actual job is to give an AI a sense of morality. Her name is Amanda Askell, she has a philosophy PhD from NYU, and she might be doing one of the most quietly important jobs in tech right now.
As Fast Company reports, Askell joined Anthropic in 2021 after a stint at OpenAI, and she sits at the beating ethical heart of the company's efforts to make Claude - their flagship AI model - something more than a very confident search engine. We're talking about instilling actual ethical instincts into a machine. Which, depending on your mood, either sounds like the coolest philosophy gig ever or the setup to a very bad sci-fi movie.
Why this actually matters
Here's the thing most people miss about where AI is heading. We're moving fast out of the "chatbot that writes your emails" phase and into what the industry calls the "agentic" era - where AI models don't just answer questions, they complete tasks. Real tasks. With real consequences.
Think booking your flights, managing your calendar, making purchasing decisions, maybe one day doing things with significantly higher stakes than that. When an AI is just chatting, a weird or wrong answer is annoying. When an AI is acting - actually doing things in the world on your behalf - a weird or wrong decision becomes a genuinely different kind of problem.
That's precisely the gap Askell is trying to bridge. Not just making Claude polite or avoiding obvious disasters, but giving it something closer to a functional moral compass.
Philosophy majors, your time has come
There's a delicious irony here that deserves a moment of appreciation. For years, philosophy graduates have endured the "so what are you going to do with that?" conversation at every family dinner. Turns out the answer might be "shape the ethical framework of the most consequential technology of our generation." Not bad for a humanities degree.
Askell's work represents a growing recognition that building AI isn't purely an engineering problem. The hardest questions - what should an AI refuse to do? How should it weigh competing interests? What does it mean for a model to behave with integrity? - those are philosophy problems dressed up in Python.
The stakes are getting harder to ignore
As AI systems take on more autonomous roles, the question of whether they carry any ethical sensibility stops being abstract. It becomes very, very practical. Askell and the team at Anthropic are betting that getting this right from the inside - building ethics in rather than bolting safety on afterward - is the approach that actually holds up.
Whether they're right is one of the more interesting experiments running in tech today. No pressure, philosophy.
