New Delhi: Those holding a degree in philosophy traditionally find jobs as professors or researchers in universities. But now AI seems to have opened the door to new opportunities for them. Big artificial intelligence firms are turning to philosophers to train their models better.
AI firms have always grappled with safety questions when it comes to training new models or improving old ones. But for deep behavioural alignment issues, companies need what have been dubbed as “ethics engineers”. And this is where philosophers come in.
These philosophers help translate human values into algorithm constraints and prevent the AI model from exploiting loopholes in the reward structure. According to The Percolator, a Substack, this phenomenon is called Goodhart’s Law, and philosophers are skilled at spotting such vulnerabilities “because their training focuses on the structural consistency of ethical frameworks”.
This is reshaping how AI firms hire. According to Federal Reserve Bank of New York figures, cited by The Economist, those graduating with a philosophy degree in the US are more likely to land a job than those who obtained a computer science degree. Luciano Floridi, a philosopher at Yale University, said that many philosophy students are hired by AI firms before they even graduate.
A report in The Week said that top AI firms such as Anthropic and OpenAI have hired “high-profile” philosophers to help them make their models better. Sam Altman also, in an interview last year, had said that OpenAI hired “hundreds of moral philosophers” while designing rules for how ChatGPT will behave.
While hiring philosophers to train AI models doesn’t seem alarming, critics have raised concerns about what they call “moral deskilling”, when humans no longer feel like making ethical calls because AI is doing the job for them.’
“Morality is historically unstable, culturally variable, strategically manipulable, and often only retrospectively legible,” said Roman Yampolskiy, an AI theoretician at the University of Louisville.