AI has a new profession: philosopher
Why Anthropic, Google DeepMind and other AI companies are hiring philosophers — and why the question will soon reach industry.
That is bad news for the phrase “let the developers decide.” Only a few years ago, “corporate philosopher” sounded roughly like “corporate charlatan.”
Today philosophers are hired by Anthropic, Google DeepMind and other leading AI companies — not to give lectures to staff, but to take part directly in model development.
honesty = ?
Why? Because some of AI’s problems turned out not to be technical at all. Say we want to make a model “honest.” Great. Now try writing the function: honesty = ?
What counts as a lie if the model is not sure? Should it tell an unpleasant truth? Can it withhold some information for safety reasons? What matters more: fulfilling the user’s request or preventing potential harm?
A GPU does not answer these questions. Neither does a good dataset. First, someone has to formally define what behavior we actually want from the system.
That is why a philosopher on an AI team today works quite close to an engineer: helping to formulate principles of model behavior, evaluation criteria, conflicting values and rules for situations that cannot be enumerated in an ordinary if / else.
This is especially visible at Anthropic. Amanda Askell, who holds a PhD in philosophy, helped create the principles that underpin Claude’s behavior — in essence, designing what kind of “character” the model should have.
Which leads to a slightly provocative conclusion:
The smarter software becomes, the less its behavior can be defined by a technical specification alone.
What this means for industry
For industry, this question will soon become especially interesting. Imagine an autonomous production system. It sees that equipment is running unstably. Stopping the line now means losing several million rubles. Continuing means a higher chance of breakdown.
Who should decide what level of risk is acceptable? The developer? The process engineer? The plant director? A lawyer? Or the very person who can first properly formulate the decision-making principles and only then turn them into an algorithm?
That is why the rise of AI philosophers at the largest AI companies is not an oddity but a fairly logical stage in the industry’s development. Interestingly, the trend already has critics: some researchers fear these roles could turn into box-ticking for its own sake if philosophers have no real influence on product decisions. But that is a topic for another post.