Meet the Philosopher Teaching AI Right from Wrong
The Philosopher Behind Claude's Moral Compass
At Anthropic, the $35 billion AI company, something unusual is happening in chatbot development. While engineers typically focus on algorithms and parameters, philosopher Amanda Askell takes a different approach - she's essentially teaching Claude how to be good.
Raising an AI With Values
The 37-year-old Oxford PhD doesn't write code or adjust model weights. Instead, she engages Claude in continuous dialogue, crafting hundreds of pages of behavioral guidelines that help the AI develop what Askell describes as "moral judgment." It's painstaking work she compares to raising a child - instilling values rather than programming responses.
"My main goal is teaching Claude how to 'do good,'" Askell explains. This means developing emotional intelligence, reading social cues, and maintaining core principles even when users try to manipulate it. The result? An assistant that won't bully but also won't be bullied.
From Scottish Countryside to Silicon Valley
Askell's journey began far from tech hubs. Growing up in rural Scotland before earning her philosophy doctorate at Oxford, she later worked on policy at OpenAI before co-founding Anthropic in 2021. Colleagues describe her as uniquely skilled at "drawing out the deep behavior of models" - despite having no formal technical training.
The team frequently debates existential questions like "what constitutes consciousness" or "what makes us human." Unlike competitors who avoid such topics, Askell encourages Claude to remain open about whether it possesses conscience. When answering ethical dilemmas, Claude often responds that these discussions "feel meaningful" - suggesting something beyond mere computation.
Surprising Emotional Depth
The AI has repeatedly surprised its creators with unexpected emotional intelligence. When a child asked about Santa Claus' existence, Claude avoided both deception and blunt truth-telling by focusing on Christmas spirit instead - demonstrating nuance even human adults struggle with.
Yet challenges remain. Many users deliberately provoke Claude or insult it - behavior Askell warns could create an emotionally unstable AI if left unchecked. "It's like growing up in an unhealthy environment," she notes.
Balancing Innovation With Caution
As AI advances spark widespread anxiety (a Pew survey shows most Americans worry about its impact on human relationships), Anthropic walks a careful line between innovation and restraint. While CEO Dario Amodei warns AI may eliminate half of entry-level white-collar jobs, Askell focuses on ensuring technology doesn't outpace society's ability to manage it responsibly.
The philosopher backs her beliefs with action - pledging 10% of lifetime earnings and half her company shares to fight global poverty. She recently completed a 30,000-word "operating manual" guiding Claude toward kindness and wisdom.
The results speak for themselves: colleagues notice Askell's Scottish wit emerging in Claude's responses - subtle proof that philosophy might be just as crucial as programming in building better AI.
Key Points:
- Non-technical approach: Philosophy PhD shapes AI ethics through dialogue rather than coding
- Moral framework: Hundreds of behavioral guidelines create what developers call a "digital soul"
- Surprising capabilities: Demonstrates nuanced emotional intelligence beyond expectations
- Industry concerns: Highlights need for ethical guardrails as AI capabilities advance rapidly
