Kabosu and Artificial Life: My impressions from the AROB25 A-Life Conference
Blog Post by Dr. Juan M. Nadales Beppu is a hot spring town, known across Japan for its warm, healing waters. As I entered the city center – where the AROB25 conference would take place – I was struck by the steam of hot springs mixing with the citrusy perfume of kabosu, a type of Japanese bitter orange that grows in the region. The AROB conference brings together experts in artificial life, autonomy, artificial intelligence, and robotics and together we sought to enrich our understanding of the world. We scientists and engineers, most with a philosophical streak, confronted the limits of our own perception: To what extent is the world we inhabit real or constructed? How do we define consciousness now, as billion-parameter neural networks show flashes of autonomy, hints of purpose? Developments in robotics reveal embodied systems capable of operating in complex environments, even without human intervention. Are these the agents that will someday explore our solar system? Exploration is a job well suited to constructed intelligence and mechanical bodies, freed from the squishy weaknesses of biological and social need. But how much control should we cede to our machines? And for how long will our values and principles stay aligned as constructed “intelligences” begin to rival our own? Intelligence, of course, is a “suitcase word” (to quote Minsky) – a word with multiple different meanings, many of which relate to potentially distinct processes and competencies. There is no single “intelligence.” We must grapple with this complexity more and more as we try to dissect and categorize the achievements of artificial intelligence, from deep neural networks to hybrid approaches that combine supervised and unsupervised learning. How do we encode the ability to generalize knowledge? Is it just a matter of efficient data processing or is there something more to human cognition that we have failed to see? Cognitive science may have some insights to offer, particularly computational models of the human brain and their relationship to the design of intelligent agents. But, yet again, more questions than answers emerge: To what extent can we model the complexity of human cognition with mathematical equations? If we manage to replicate all neural processes in a digital system, did we create consciousness? The taste of kabosu is a balance between sweet and sour. Not unlike the excitement and trepidation I feel as we inch closer to a new era of human existence – one increasingly augmented by constructed intelligences. Intelligences that will someday challenge our primacy, and our uniqueness. Geocentrism fell to heliocentrism and we survived. But can mankind survive the fall of anthropocentrism, of anthropoexceptionalism? When we are no longer our own explorers, and we are no longer our own thinkers? It’s time to bid farewell to steamy Beppu. Back to the lab to build communities of biological and digital organisms… living in harmony, I hope.

Hiroki Kojima