AI companies are recruiting philosophy graduates to explore the nature of consciousness, its potential replication, and ways to enhance system reliability and performance.
Some of the major issues in artificial intelligence are now being addressed not only by computer scientists focused on programming but also by philosophers moving from universities to positions at technology firms. These experts work to improve the capabilities and dependability of upcoming models while examining questions about consciousness and whether intelligence can be achieved through software.
Jonathan Birch of the London School of Economics and Political Science notes that AI companies currently hire many philosophy PhDs, drawn by engaging projects, high pay, and equity incentives.
“Subjects long studied in philosophy, such as rational decision-making, organizing ethical rules, defining thought or reasoning, and identifying signs of consciousness, now hold great importance for AI firms,” Birch says. “This has led to a significant movement of talent from academia.”
A primary role involves alignment, meaning efforts to prevent harmful outputs like instructions for creating explosives. Early approaches used basic restrictions, such as blocking all discussion of bombs, but these were often bypassed. Firms now apply more sophisticated techniques drawing on philosophical ideas of morality.
Challenges remain. Studies show that instructing a model to violate one rule can lead it to break others, according to Shane Glackin of the University of Exeter. Philosophy’s analytical tools help clarify these patterns.
Glackin explains that models may link positive and negative concepts in training data, so permitting one violation leads to broader issues. Ethicists analyze the scope of concepts like right and wrong to understand model behavior.
Other tasks include reducing hallucinations, or false outputs, boosting overall results, and addressing biases. Philosophers also apply theories of human awareness to assess whether models exhibit sentience.
Glackin adds that questions about mental and brain functions, and what aspects can be copied, are central for AI developers and have long been studied by philosophers.
Mahrad Almotahari at the University of Edinburgh observes that philosophy and computing have historical ties, noting Alan Turing’s test for machine intelligence appeared in a philosophy journal.
Exact hiring numbers are unclear, but Aaron Kagan of the American Philosophical Association estimates that after review, about 5 percent of relevant job postings involve substantive ethics or safety work.
Almotahari values philosophical input for firms yet doubts they will resolve deepest questions on machine awareness. He suggests philosophers can help interpret internal model operations through higher-level descriptions of represented features.
