When Freedom and Authenticity Diverge: James Brusseau on What Philosophy Sees in GenAI That No Other Discipline Does
Q1. You have written about “acceleration AI ethics” — an approach that argues for using ethics to enable and accelerate GenAI innovation rather than primarily to constrain it, in contrast to the precautionary approach that dominates European regulatory thinking. That is a provocative position for a philosopher, since the instinct of most applied ethics traditions is to slow things down and ask hard questions before acting. Can you make the strongest version of the case for acceleration ethics in the context of GenAI — and where does your own argument hit its limits, where does the speed limit actually need to be set and by whom?
Acceleration means that innovation itself solves the safety risks that innovation initially causes. So, AI in medicine generates privacy risks, but further advances including differential privacy and federated learning begin to resolve them. The same innovative force that sparked the risks extinguishes them. Safety, consequently, becomes an impetus for innovation. For precaution, the logic runs the other way. As you note, safety is understood as opposed to innovation, and therefore it slows things down to locate and resolve potential privacy risks before significant development begins.
So, what is acceleration ethics? It is the human values underlying acceleration in practice. The ethics question is: What do we need to believe about ourselves and the world to be accelerationist?
There are two core beliefs. The first concerns innovation, it is understood as valuable intrinsically, and regardless of its subsequent effects in the world. In this way, information engineering resembles art. We do not diminish our regard for a painting if it is bought and sold for the purpose of money laundering: the aesthetic value remains unaffected by the painting’s illegal uses. So too for technical innovation. If it is creative and productive, it is good. And that initial value subsists regardless of whether subsequent applications are beneficial or malevolent. Regardless, for the precautionist, the value of innovation is entirely determined by its uses and effects in the world. For precaution, the ethical value of a technical breakthrough only equals its real-world consequences.
Then, the second belief underlying accelerationism involves the value of the unknown: For acceleration, the unknown is desirable in itself – it is magnetic. Think of our great nomadic travelers, for example Isabelle Eberhardt or Jack Kerouac. What distinguishes their journeys is that they travelled because they did not know what they would find. They did not go to learn a different language, or to acquire job-related experiences. They went for the unforeseeable. We all have something of this in us. It is, for example, in the desire to buy a one-way ticket to a random destination. What operates here is not a desire to discover something specific, it cannot be that because we have no idea what awaits. Instead, it is the impulse to discover discovery itself, the pure form. It is the kind of discovery that exists regardless of what is found. Here is a presentation going deeper on this subject.
It follows that the strongest discoveries and the best innovations are not those that solve specific problems, instead, they create whole new regions of the unknown. Generative AI is like this. It has opened places to explore where we have no idea what will emerge, from healthcare to job-searching and dating, new possibilities exist with unforeseeable outcomes.
Precaution goes the other way, it is greedy for the kind of knowledge that is constricting. It seeks to extinguish the unexpected, and the accompanying risks of the unforeseeable. There is no adventure in precaution.
Ultimately, my argument is that the debate between acceleration and precaution will be determined by two underlying ethical questions. What is the value of innovation? And, what is the value of the unknown? Everyone will have their own unique answers, and correspondingly their own way of responding to the AI tension between innovation and safety. I write more here and here.
Q2. You published a paper comparing how ethics diverges between American and European AI governance — arguing, in essence, that these are not just different regulatory preferences but reflect fundamentally different philosophical traditions about the relationship between individuals, institutions, and technology. For a policymaker or technology leader trying to navigate a world where AI products, AI companies, and AI regulation are increasingly in geopolitical tension, what is the most important thing that the American and European approaches each get right that the other consistently undervalues — and is there a coherent philosophical synthesis, or are these genuinely irreconcilable frameworks?
Marx had two entwined ideas. First, humanism must be saved from the kind of work and technology that reduces us to simple props in an economic system. Second, and as influenced by Hegel, history has a collectivist destiny. European approaches tend to gravitate around these two ideas, while most American approaches are less theoretically robust, and also less collectivist.
About Marx’s two ideas, they can be separated. The value of humanism in the face of technology can survive the rejection of an economic model privileging community over individualism.
With all that in the background, we could say that one way to coordinate American and European approaches would be to mutually embrace Marx’s ideas about individual alienation in the economic world, but to leave behind the antiquated ideas of collectivism and socialism.
Ultimately, AI abundance may enable the Marxian ideal of allowing us all to choose our own occupations as a form of self-expression (as opposed to economic need), while also enabling the rejection of collectivist and socialist politics.
Q3. You have published on “the mimetic AI professor” — an AI system that can simulate a professor’s knowledge, tone, and pedagogical style. You are also a professor who teaches philosophy and ethics at a university in New York City. That puts you in a genuinely unusual position: you are both an academic who has thought carefully about what AI means for higher education, and someone whose professional role is directly implicated by the technology you are analyzing. What is the honest version of what a mimetic AI professor can do that a human professor cannot — and what does a human professor offer that no mimetic system can replicate, not as a matter of sentiment but as a matter of philosophy?
The critical distinction, I believe, is that AI can create knowledge in two ways: by copying subject matter expertise, or by copying the subject matter expert. This is easy to see in terms of chess. You can play against a computer opponent trained by all the moves and games recorded historically, or, you could play against an adversary trained only on the data of one player, say Magnus Carslon. These are two very different opponents, and they are also two conceptions knowledge. One is based on summation and scale, while the other is based on personalization and curation.
Now, imagine you forced Carlson to play together with Kasparov against an opponent. The sum of their efforts would be far below their individual abilities because the two with their florid minds would contradict each other as much as help. The same goes in the classroom. One professor is better than two because the way academic knowledge is produced is personal, it depends on the vocabulary and experiences of the specific teacher. When there are two, the effect is not summative, it is subtractive. They cancel each other and short-circuit.
So, my first argument is that the hyper-scalers (OpenAI, Anthropic) are fundamentally limited. The next phase of AI will not be more scale but more specification. Depth is better than breadth. Knowing more about one professor is better than knowing a lot about a lot of professors.
Now, it is true that for chess or any closed environment, the generic AI trained on all the data will outperform the AI trained on a personalized slice. But, that is not true in open environments, in places where there are no clear limits about what counts as a “move” or how “winning” is recognized. For example, two professors teaching philosophy will have different moves and ways of winning, they will use distinct examples and language, and they will have distinct goals for their students. This is the kind of situation where generic AI is not better than the mimetic version. It is worse.
So, I believe we need to divide our knowledge creation onto two levels. On the closed level there are things like math proofs and the creation of protein folds. Work on this level may have tremendous implications for humanity, and scaling is probably the best strategy, but ultimately this is AI as just a glorified calculator. Higher intelligence occurs in places where we do not even know what counts as knowledge. Again, this is the realm of philosophy and art and the more significant concerns of life. And, here on this higher level, mimetic AI is the more promising technological companion.
Anyway, back to our project, the Caffeinated Professor. This project is making progress technically and commercially. We outperform the hyper-scalers like OpenAI and Anthropic in the area of advanced education in humanistic subjects. Next, will outperform human professors in terms of consistency. AI does not get tired and never falls into a bad mood. But, it is just as obvious that knowledge based on mimicking a human expert can never do without that expert.
Q4. GenAI has created a profound challenge for academic integrity — students can now produce fluent, apparently well-reasoned essays, code, and analysis with minimal cognitive engagement. Most universities have responded with a combination of detection tools, policy updates, and revised assignment design. As a philosopher of AI who teaches ethics, what is your honest assessment of whether those responses are addressing the right problem — and what does your philosophical framework suggest about what universities should actually be trying to preserve in human learning that is genuinely threatened by GenAI, versus what they are clinging to out of habit or institutional inertia?
AI solves AI problems, so mimetic AI will alleviate cheating and cognitive off-loading problems by administering dynamic oral exams that assess student thinking as opposed to static knowledge. Our team is on that project now.
Q5. You have been involved in Z-Inspection® assessments of real AI systems in healthcare, and you have written that applied ethics needs to move from abstract principles to practice — “from the ground truth up” rather than from principles down. That is a critique of a significant portion of the AI ethics field, which tends to produce principles, frameworks, and guidelines that are difficult to operationalize. For an organization — a university, a hospital, a government agency — that genuinely wants to do applied AI ethics rather than perform it, what does that actually require in terms of process, expertise, and organizational commitment that most applied ethics efforts currently lack?
From my experience doing AI ethics evaluations on real tools functioning in the world, the critical step is the first, the one where the relevant domain experts gather and translate their initial thoughts into commonly-understandable language. In these meetings, philosophers like me cannot talk about the ontological ramifications or epistemic parameters. We need to ask whether medical treatments for skin cancer should be balanced across the population, or focused on the most vulnerable, even if that means creating race imbalances and privileges. The same goes for doctors and engineers. We all have our own professional languages that sometimes impede the understanding of outsiders more than facilitate the work of insiders.
So, building from the ground up does not mean abandoning theory, but it does mean theorists understanding ideas (like Kant’s categorical imperative) sufficiently well to explain them to a general audience in accessible terms.
Q6. You have spent your academic career moving between philosophy and AI ethics, between New York and Italy and Mexico, between the history of philosophy and its most contemporary applications, between Nietzsche and Deleuze and ChatGPT. That is an unusual intellectual trajectory. What has studying the history of philosophy — the deep history of how humans have thought about knowledge, identity, and the good life across centuries — given you in understanding the GenAI moment that a computer scientist, a policy lawyer, or a data ethicist trained primarily in contemporary frameworks simply does not have — and what question do you believe the philosophical tradition is uniquely equipped to answer about GenAI that no other discipline is even asking in the right way?
The AI human condition. I mean, what new human dilemmas arise in the wake of this technology? What new values and potentials are possible?
Here is one example. Traditionally, the ideas of human freedom and authenticity have interlocked. The reason for human freedom is to determine who we authentically are, and the role of authenticity is to provide significance and a justification for freedom. This exists in Heidegger. But, I believe, AI has begun breaking the link. Now, vast databases and predictive analytics coordinate to understand who we are even better than we know ourselves, at least sometimes. Our desires, fears and aspirations are more clearly understood by analyzing our own data patterns than by the empirical experiments of human freedom. In crude terms, we are better off letting Netflix select our next movie than we are to choose for ourselves.
The result is a new reality where authenticity and freedom are opposed. I mean, if human freedom overrules AI outputs, then we become less who we authentically are, not more. So, what does this mean for human freedom, do we even want it anymore? What good is it to be able to choose if AI can choose better than we can? And, what does this mean for human identity? If machines know us better than ourselves, is there any reason for us to know about ourselves?
There are answers to these questions. There is a kind of freedom that no longer seeks to know who we are, but to re-create ourselves as someone new. How would that happen, though? And, what does that mean, exactly? What role could AI play in this new human project?
These are the kinds of questions that emerge from a philosophical approach to today’s unique reality. I am writing a book on this now, but the distractions are many. Still pieces have been assembled here, here, here, and here.
Qx. Anything else you wish to add?
Algorithmic recommenders, it seems to me, are critical. They both provoke and delimit what we can want and who we can become as individuals. Perhaps AI will be most consequential for humanism in this area of application. Here is a recent paper on the subject. Good and important work is being done on this at Uni Trento, and also at Pisa.
……………………………………………………..
James Brusseau (PhD, Philosophy) is a professor in the Philosophy Department at Pace University in New York City during the fall and winter, and he is a professor in the Computer Science Department at the University of Trento in Italy during the spring and summer. His academic research explores the human experience of artificial intelligence in the areas of personal identity, authenticity, and freedom.