Intelligence and Curiosity
May 5, 2023
In the previous article on intelligence and memory, I posed the question of our pressing need to find an alternative to memory as an appropriate yardstick for measuring intelligence. I also asked the important moral and legal question about living in a world where nothing is ever forgotten.
Looking at the first of these issues, the nature of intelligence, we must move past the beguiling impressions created by interactive AI to consider what essential qualities are most needed by humanity in the future. If information storage and recall are no longer attributes of intelligence unique to humanity, what else might serve as a reliable indicator of human intelligence?
If we take the current most visible and perhaps controversial open use of AI, the generative, pre-trained origins of interactions between machine and human point to their largely scripted nature. While the script may be hidden from the human participant, the interactions essentially mimic human communication based on existing models, rather than creating genuinely novel conversations. As such, we face an interesting prospect: are AI-human interactions just a technologically clever way of talking to ourselves?
Human interaction, as a part of the shared human experience, is most frequently framed around the question-and-answer dynamic. We learn because we encounter a gap in our knowledge, and we ask someone to ‘fill the gap’; the interaction facilitates the construction of new knowledge. In fact, the posing of novel questions to which there are no current answers serves as a powerful driving force in human development. Our native curiosity compels us to focus our attention on the unknown.
So, what might we say of questions asked of and answered by AIs? Who is asking and why? Are all questions or problems equal or equally important or interesting? And to whom? In a virtual universe of infinite questions, how are the nonsensical, impossible, or trivial questions eliminated? How are potentially meaningful questions generated ab initio – from the beginning? Can an AI genuinely generate new (novel) questions? Or do we build ever-bigger artificial systems based on the ‘Infinite Monkey Theorem*’ and embrace the absurd notion of an infinite number of random text-generating systems eventually asking useful questions? Even systems cleverly designed on inductive principles to draw from ‘data-in-depth’ to respond to questions cannot induce from what ‘isn’t’, they can only work with what ‘is’. A system may extrapolate and generate accordingly, extending output along a line of inquiry or structure anticipated by the system designer, but to generate a truly novel question we must return to the probability trap of the ‘infinite monkeys’.
Perhaps one avenue for consideration is the nature of novel questions and that of curiosity itself. For example, an element missing in manufactured demonstrations of ‘intelligence’ is the ability of the artefact to experience genuine curiosity about its existence. Unlike animals and AI, humans are natively, innately, incurably curious. Unlike animal curiosity (‘Curiosity killed the cat’, after all!), the extent to which human curiosity shapes and empowers action is evident in many of the milestones of history. In fact, neither animals nor machines in controlled experimental settings have shown any inclination to express any curiosity about why the question exists in the first place. This would likely not be the case with a human subject.
Humans are not programmed systems. There is no infinite storage or recall. Memories are often incomplete, skewed by experience or emotion, or flawed. Humans habitually form subjective judgements based on individual emotions, experience, world views, intuition, and imagination. Unlike an AI, a human subject would not see all questions as ‘equal’. Based on feelings, whim, intuition, or sheer imagination, we may decide that some questions are more interesting, or ‘intriguing’ than others.
Perhaps most tellingly, beyond curiosity about why things are the way they are, humanity has an amazing capacity to be curious about why things are ‘not’. Extrapolating or inferring data from the known into the unknown, or negative, ‘blank’ spaces of human knowledge, can only be anticipated or programmed to an extent. Exploring the unknown, the truly novel emerges, springs, or spontaneously combusts in spaces where no precedents exist. Without ‘rhyme or reason’, we impose highly personalized elements into our exploration of the unknown, sometimes without result, sometimes with truly inspiring results. Insight and imagination may not necessarily have any antecedents.
To avoid a future that is completely derivative, we must call on educators and parents to encourage, support, and inspire curiosity and creativity. Making time and space to nurture curiosity and harness creative energy in the education of children remains a major challenge.
*The ‘Infinite Monkey Theorem’ is based on the notion that an infinite number of monkeys randomly hitting keys on an infinite number of typewriters would eventually replicate a script for the Shakespearean play Hamlet.
Dr. Malcolm Pritchard
Head of School
