The AI for Good Global Summit 2018 was opened with a keynote speech by Sir Roger Penrose (Emeritus Rouse Ball Professor of Mathematics, University of Oxford). Based on his experience in physics, mathematics, and philosophy, Penrose addressed the question of Why Algorithmic Systems Possess No Understanding.
Artificial intelligence (AI) has advanced tremendously, and these developments have coincided with questions of whether – or when – AI will reach the level of human intelligence. Penrose compared AI to the cerebellum; the part of the brain that receives input from sensory systems and integrates them to fine-tune movement, coordination, precision, and timing, and he contrasted the cerebellum with the cerebrum, which initiates and coordinates activity in the body. Penrose explained that the relation between the cerebrum and cerebellum is akin to that between the programmer and the program.
According to Penrose, just like the cerebellum lacks an understanding of why it does what it does, computers are unlikely to encapsulate understanding or consciousness any time soon. The gap between algorithmic computation and general understanding is visible in quantum physics, and Penrose quoted the example of Schrödinger’s cat to highlight the discrepancy between computational outcomes (the cat is both alive and dead) and understanding (the cat is either alive or dead).
So how can we conceive consciousness and understanding? Penrose suggested that it must be rooted in physics, and may be explained by microtubules, which are tiny tubes located within brain neurons. According to this theory, the fine-scale activities of these microtubules form the building blocks for consciousness. In the absence of these biophysical elements, computers are unlikely to attain consciousness and understanding.
Penrose’s lecture was followed by a Q&A moderated by Mr Stephen Ibaraki, (Futurist and Social Entrepreneur) who started by asking Penrose about his thoughts on the term ‘AI’. Penrose explained that he felt particularly ‘nervous about the word intelligence’. Intelligence commonly requires understanding, and understanding commonly requires awareness. As AI devices are not aware, they are not intelligent in the normal use of the word; while such a system can achieve a lot, it ‘doesn’t seem to know what it’s doing’. He suggested that instead of AI, we could adopt the term ‘artificial cleverness’.
At the same time, Penrose explained that there is still a lot of room for AI to further develop. We can continue to use our human understanding to improve the algorithmic system, integrating missing ingredients, and transforming it into something that ‘goes beyond what you had before’. Yet, without the quantum processes of microtubules taking place in human brains, could computers ever mimic conscious brain activities? Penrose explained that we might, some day in the far future, be able to construct such protoconscious elements in a laboratory. However, this would raise many ethical problems that we are not ready to face.