Where is the wisdom we have lost in knowledge?
Where is the knowledge we have lost in information?
T.S. Eliot, Choruses from “The Rock”
Preamble. AI is sending academia (and society) out of equilibrium. We can’t pretend that nothing is happening, waiting for the next equilibrium to be attained; we can’t be sure of the long-term consequences. We need to take some bets for action here and now.
This charter is built on two pillars: the bet that theoretical physics will remain relevant; and the belief that we should always strive to become better persons (“ethic of virtues”).
1. On productivity vs competence. Until recently, a healthy level of productivity was well correlated with competence. AI destroyed this correlation and may even cause an anti-correlation: it is easy to churn out one paper a week, while almost certainly making incremental contributions and not even understanding them. We believe that competence is what we should aim at; so, we happily outsource tedious tasks that we already understand, but we try to assimilate new information and techniques. We bet that institutions will want competence and recognize fast enough that productivity is no longer a proxy for it. We are aware that we may lose job opportunities, but we should be proud to be rejected by a workplace where mere productivity is the criterion.
2. On quality vs speed. Some people have stopped reading papers: they rather feed them to the LLM and ask it to distill the main results. In turn, others may neglect clarity in their writing: the result will count for their tally anyway, and the LLM provides personalized explanations to the readers. We believe that spreading slop is not the right way to use our talents in the service of society, and commit to write papers that reflect our structured understanding. We bet again on competence: even if we may be scooped by people who post slop, in personal dealings and interviews we will be able to stand our grounds, and hopefully this will be appreciated.
3. On research topics. We work in theoretical physics with the preferential goal to describe reality (usually the outcome of experiments rather than bare natural phenomena); connections with information theory and other mathematical domains may happen as collateral effects. With AI providing many answers and generalisations, the focus is on asking the right questions. We believe that isolated strikes of genius may happen, but that it is far more frequent to find which questions are interesting and novel in discussion with other humans. We are aware that AI has increased the risk of being scooped by unscrupulous people, but we prefer this to withdrawing our trust from the many who deserve it. We also try to formulate objectives that are both ambitious and “decidable”, in the sense that their completion can be assessed (e.g. “propose how to cool down that system”, as opposed to “understand thermodynamics in the quantum regime”). This way, AI becomes a tool to achieve our objectives.
4. Opt-out strategy. Along the way, some may find that these changes destroy, perhaps not the whole field, but yes what they found attractive and compelling in theoretical physics. The classic advice by Ignatius of Loyola comes handy here: “In desolation, don’t change”. That is: don’t quit what you used to like in a fit of disappointment: rather, look for something else that enthuses you, and only then move on.