"If physics is the description of the real world, mathematics is the best representation of God". My education is grounded on the assumption that the purest expression of intelligence is mathematical mastery. It is no surprise that, needing to prove the superiority of their progeniture, AI labs ferociously vanquished the highest mathematical summits in a record time as soon as these were within their reach.
Much has been written about the role of mathematicians in an AI future since OpenAI’s bulk proof dump. Very little however has been written about the questions raised by impressive AI performance on the useful benchmarking of human intelligence. If AI can now prove increasingly difficult mathematical conjectures in a few hours, is the most relevant sign of human intelligence mathematical fluency as defined by our education?
This essay is asking a simple question: should we mourn human lost intellectual prerogatives or exploit this new acquired freedom to develop further existing intellectual abilities undervalued in a world venerating knowledge. Furthermore, by identifying and developing AI complementary human intelligence capabilities, can we create a better human AI symbiotic coexistence?
Today’s education develops aspects of intelligence which are becoming quickly redundant with artificial intelligence capabilities, almost by design. However, today’s codified education is undervaluing aspects of human intelligence which are more difficult to understand and therefore to develop intentionally: intuition, gut feel, originality, critical personalities, sense of aesthetic, navigating uncertainty, soft skills difficult to quantify and difficult to codify. As AI is taking over knowledge-based expertise, we need to reconsider which of these human capabilities need to be developed now.
Julian Bradshaw’s recent “Relevance Realization as Human Comparative Advantage” points to a similar direction: cultivate human ability to recognize what matters and use it to leverage AI. Going one step further, we should ask how the advance of Ai could help us develop aspects of human intelligence that our existing way of life has undervalued.
The forced transformation of the world of mathematics provides a useful framework to think about the required human adaptation. Mathematics is an impressive edifice layered over centuries of beautiful minds contributions to the point that frontier research demands extraordinary specialization. Our understanding of what led to AI’s recent successes is still incomplete: greater proficiency within a specific field, combination of distant techniques impossible to fully comprehend for a small group of human intelligences, new mathematical ideas? The answers will provide important directions to explore where human interaction with AI will produce what neither would readily create alone.
Pure mathematics research illustrates well that within a given field, expertise guides investigation. However, across distant fields, the sense of a promising direction rather than precise knowledge can provide a new path to investigate.
Intuition, analogy, aesthetic appreciation and critical dissatisfaction become the critical intellectual capacities. Intuition can translate accumulated experience that has not yet become an explicit argument. It then becomes the vector to acquire necessary knowledge from other colleagues or from AI.
Poincaré describes mathematical invention as a combination of unconscious work and aesthetic selection. In his recent talk, Jean-Pierre Serre distinguished the pleasure of discovering mathematics from the pleasure of watching mathematics advance. These descriptions suggest that feeling the significance of an idea is part of the human intellectual process, alongside establishing its correctness.
Relevance realization provides the scaffolding to analyze how our concerns, understanding and sense of possibility develop together. An inquiry rarely begins with a complete menu of questions and settled ranking. Let’s call the corresponding research capacity navigation: recognizing worthwhile destinations, sensing connections, and revising direction as investigation changes the perceived landscape.
Herbert’s Dune guild navigators and their prescient foresight give a form to this ability: sensing a passage through an overwhelming landscape of possibilities. In Blindsight Watts imagines an intelligence capable of extraordinary problem solving without conscious experience, inquiring about the gap between solving a problem and understanding its significance. These authors help us recognize the difference between intellectual proficiency and the broader capacities involved in recognizing, questioning, and experiencing what matters.
AI is dramatically extending human ability to perceive larger landscapes, to a point never reached in human history, offering to the unexplored navigation intelligence an opportunity to play a role it never could play. Post AI learning should deliberately exercise the cycle of intuition, investigation and revision rather than knowledge-dominated orientation. AI is lowering the cost of intuition-based exploration, risks such as rejection that couldn’t be taken and required knowledge to eliminate are now worth taking and should be rewarded rather than discouraged. Curiosity, imagination, courage, tolerance for ambiguity, intellectual risk taking are some of human intelligence qualities which need to be combined with AI knowledge-based strength.
Nurturing human intelligence against the capabilities AI is rapidly acquiring is leading to a world where human and AI compete. We can start retooling human intelligence in the directions creating the right conditions for a constructive human AI symbiosis.
