28/06/2026 06:39 - Tecnologia
Ten years ago, the advice for career security was clear: "learn to code". Today, that recommendation has been dramatically reversed.
According to data from the Federal Reserve Bank of New York for 2024, 7% of computer science graduates in the United States were unemployed, compared to only 5.1% of philosophy majors. This figure reveals a profound transformation in the tech labor market, where critical thinking ability is gaining value over pure technical skill.
"Students are receiving job offers even before graduating," states Luciano Floridi, a philosopher at Yale University, who describes the magnitude of departures from philosophy departments to the private sector as a "hemorrhage" of talent.
In mid-April 2026, Henry Shevlin, a philosopher from the University of Cambridge, announced on his LinkedIn profile his move to Google DeepMind, Alphabet's leading artificial intelligence research laboratory.
Shevlin will continue at Cambridge part-time to maintain his research and teaching at the Leverhulme Centre for the Future of Intelligence, but his role at the tech giant will be decisive: working on machine consciousness, human-AI relationships, and preparation for artificial general intelligence.
"It is a rare privilege to work on questions I've spent my career reflecting on, now with the resources and urgency that come from being inside one of the most important AI labs in the world."
One of the most valuable contributions of philosophy to AI comes from classical thought. The Socratic method described by Plato uses feigned ignorance and a sequence of questions to clarify meanings, detect contradictions, and reveal implications.
Jörg Noller, an expert in philosophy and AI at Ludwig Maximilian University of Munich, explains that models trained with this approach are less prone to flattery and more willing to seek the truth.
Implementing humility regarding one's own knowledge helps limit overconfidence, a common flaw that Noller describes as "AI immaturity".
Iason Gabriel, senior philosopher at Google DeepMind, attributes a general decrease in industry hallucinations to these philosophical efforts.
AI philosophers have focused on two main ethical frameworks to guide model behavior:
Inspired by Kant, it imposes strict rules prohibiting lying, coercion, and treating people as means rather than ends, even for a greater good.
Benefits: Greater honesty and consistent behavior.
Examples: Claude (Anthropic) and Pi (Inflection AI).
Weighs costs against benefits to decide what to do. Designed to produce "probable overall benefits that substantially outweigh foreseeable risks".
Applications: ChatGPT (OpenAI), Gemini (Google), autonomous vehicles (Waymo).
Anthropic, an AI laboratory based in San Francisco, has developed a revolutionary approach called AI constitutionalism, which involves building models around a structure of rules and principles extracted from philosophical writings with legal or moral authority.
The constitutions of its Claude models have incorporated material from:
The latest version, led by philosopher Amanda Askell, was published on January 21, 2026. Some employees have dubbed it Claude's "soul document".
IBM's "Granite" model series comes with controls that allow enterprise customers to better align outputs with their own corporate philosophies.
Francesca Rossi, head of responsible AI at IBM, explains that these controls allow users to choose where to balance philosophical trade-offs, such as:
Consequentialist algorithms are crucial in software for autonomous vehicles. Chris Gerdes, senior engineer at Waymo, a self-driving car manufacturer, indicates the trend is toward more consequentialist driving programs: if an accident is unavoidable, one must decide the least tragic way to crash.
Consequentialism is also central in AI weapons systems. Jack Shanahan, former head of the Joint Artificial Intelligence Center, explains that military objectives must be weighed against potential civilian casualties.
Stefan Heck, philosopher and director of Nauto (AI-powered truck safety systems), raises ethically complex questions: Would it be morally acceptable to prioritize young pedestrians over older ones?
Critics worry that if computers increasingly make ethical decisions, people may become less willing to make their own judgments.
Roman Yampolskiy, an AI theorist at the University of Louisville, argues that morality "is historically unstable, culturally variable, strategically manipulable, and often only legible in retrospect."
It seems there's no shortage of work for AI philosophers. In a world where artificial intelligence has invaded everything, the capacity to think critically, question, and understand has become the most valuable skill for developing technologies that profoundly impact our society.
Sources: The Economist via Infobae, National Geographic
Alfredo S. Quiroga
