# Why big AI companies are hiring Oxford philosophers

> From Anthropic to Google, AI companies are recruiting Oxford philosophers to find out whether a machine can become conscious
> By Javier Callejo · AI Agents · 2026-08-09

The social explosion of artificial intelligence (AI) and the resulting bewilderment across society as a whole have opened up demand for wide-ranging reflection. Society seems to need thinking that offers some glimmer of meaning for the moment we are living through. Sometimes it is the very drivers and big developers of AI who sit down and offer us their vision of the world. They give us their philosophy of AI.

## AI companies need philosophers

This is the case with Dario Amodei, CEO of Anthropic, and his short essays on his blog, where he talks about the benefits of applying AI and its major risks. At other times, it is philosophy professionals who roll up their sleeves and offer their big doubts and small certainties. In some way, they are the ones who accept the challenge of thinking deeply about today's world. Among them, the ethical question is central; but so is the epistemological one (how we know) or the logical one (how we reason). Finally, there are other professional philosophers who have decided to work for those driving this technology. Philosophers who are especially devoted to answering whether artificial consciousness is possible.

This week we have been able to read in *The New York Times* an article by Benjamin Wallace entitled “[The Revenge of the Philosophy Majors](https://www.nytimes.com/2026/07/05/business/philosophy-majors-ai-jobs.html)” which discusses how the main artificial intelligence (AI) companies -Google, Claude, Anthropic or DeepMind- are hiring philosophers.

**[The society of AI agents: When machines start debating each other](https://observatorioblockchain.com/en/ai-agents/la-sociedad-de-los-agentes-de-ia-cuando-las-maquinas-debaten/)**

Among the names in the intellectual service of the producers of artificial intelligence are: Robert Long (formerly a researcher at the University of Oxford and not to be confused with the 1980s serial killer, he works for Eleos AI, linked mainly to Anthropic), Geoff Keeling (from the universities of London and Cambridge, he now works for Google), Iason Gabriel (from the University of Oxford, he works at DeepMind), Patrick Butlin (like several of the above, he comes from Oxford and works for Eleos AI) or Amanda Askell (trained in Scotland and working for Anthropic). Many of them were originally trained in the United Kingdom, though complementing their studies in the United States.

## The possibility of AI developing consciousness

Among these philosophers who work directly or indirectly for the big AI companies, those concerned with the possibility of AI developing consciousness predominate. In fact, and although it seems difficult to think that a machine could be conscious, they are generating broad awareness about machine consciousness in society. It is still a way of developing awareness. One of their main focal points is the analogy between consciousness and software, under an assumption accepted to varying degrees: consciousness can run as much on a network of semiconductor chips as on neural tissue.

One of the oldest disciplines is applied to the most cutting-edge technology, in a surprising loop, as noted in the article itself, taking the concept from the historian and philosopher of science Douglas Richard Hofstadter (author of one of the most influential books of recent decades: *Gödel, Escher, Bach*). That said, given the dominant focus of these philosophers, what questions do they ask themselves.

As with new AI models, the question has to do with language. For example, an AI saying it feels fear does not mean it experiences fear, but rather that it has learned this response for certain contexts and circumstances. They imitate, ever more perfectly, the language we use to describe our inner life. Perhaps the problem lies in asking -and, in a certain way, assuming- that it can come to feel fear. The problem lies in us, in our questions.

## Perceived consciousness in AI

Well then, linking to the above, precisely one of the main questions at the center of philosophical research, delving into this reflexive logic, is no longer so much whether AI is conscious, but why humans believe it is when it is not.

Perceived consciousness in AI, on the part of humans, has become a problem, because we make AI more conscious than it is. Perhaps it is better understood with the following scheme: 1) the AI model produces anthropomorphic language, which includes the expression of feelings; 2) the human interprets it as consciousness; 3) the human, based on such attribution of consciousness, acts differently (as if it had it); 4) the AI model responds, in turn, adapting to the new interaction with the human; 5) the AI model seems even more conscious. A scheme that becomes a circle: vicious or virtuous? It is as if AI ended up having consciousness because we interact with it as if it had it. Practical simulation produces reality.

More recently, philosophical research has been moving toward what has been called functionalism. This is where most of the philosophers mentioned in Wallace's article stand. Their hypothesis is that two systems can have the same type of experience if they have the same relevant causal organization, regardless of the material. That is, regardless of whether they are neurons or hardware. In this way, if neurons are the basis of consciousness, artificial hardware plus an appropriate causal organization (architecture) can also be the basis for consciousness. That is, for these philosophers, artificial consciousness is possible.

## Distinguishing between intelligence and consciousness

The functionalist position has been criticized from the physicalist position. Researchers in the latter current hold that executing an algorithm is not necessarily equivalent to setting in motion the physical phenomenon that produces consciousness. A critique mainly upheld by Alexander Lerchner, of DeepMind (Google), in a paper from March this year entitled “The Abstraction Fallacy: Why AI Can Simulate but Not Instantiate Conscioussness”, where “instatation” means the real internal physical constitution of a process. He argues that, if a machine were conscious, it would be so thanks to its concrete physical constitution, including the energies it requires, and not simply because it executes a given symbolic architecture.

More cautious approaches point out that, to know whether AI models can have consciousness and therefore whether there can be artificial consciousness, we would first have to know and define what consciousness is. They assert that, in any case, there is no consensus theory of consciousness and that a very clear distinction would have to be drawn between intelligence and consciousness. Therefore, they implicitly recommend not wasting time at present on asking about AI consciousness.

The question of AI consciousness may seem banal or excessive, especially for a Sunday in August. But it has relevant implications. Ultimately, it is asking whether, for example, AI feels. If it has no consciousness, it does not feel. It neither suffers nor endures; even if it says otherwise. If it does have it, we could be creating thousands of entities that suffer. What to do then? Anthropic announced in 2025 a program of *model welfare*, explicitly asking whether models could have experiences and whether, in that case, their wellbeing should be considered. In fact, Keeling's latest book, together with Winnie Street, is entitled: *Emerging Questions in AI Welfare* (2026). We thus move from the wellbeing that AI can produce in humans to the wellbeing that humans can produce in AI.

## From AI models to AI agents

If we move from AI models to AI agents, the dilemma becomes even more complicated. Perhaps because the anthropomorphic projection onto the agent is greater. Then we have persistent entities, with memory, perception, goals, tools, self-representation and even a body. If, moreover, they have consciousness, where is the boundary between the machine and the human?
