I was wrong about being wrong about AI

The speaker revises their view on AI, emphasizing that human intelligence is not merely an algorithm but an emergent property of billions of years of evolutionary processes, making it inseparable from living beings. They argue that true artificial general intelligence cannot be achieved through shortcuts or abstractions, as it requires retracing the entire natural evolutionary journey that produced human intelligence.

In the video, the speaker reflects on their evolving perspective about artificial intelligence (AI). Initially, they considered intelligence as an algorithm that humans run, akin to an app on an iPhone, which approximates reality through updating internal weights based on real-world interactions. This led to the question of whether computers could eventually replicate or surpass human intelligence by drawing similar approximations if given direct access to reality. However, the speaker emphasizes that their views are not fixed but part of an ongoing process of questioning and exploration, inviting viewers to join them on this journey rather than expecting definitive answers.

The speaker then clarifies their revised understanding: humans are not running intelligence as an algorithm; rather, humans themselves are the intelligence algorithm. They frame intelligence from a secular, materialistic standpoint, explaining it as a product of a long evolutionary process—autopoiesis—where simple self-sustaining systems combined and evolved over billions of years. This natural search process, which led to human intelligence, was not goal-directed but emerged as a solution for survival and subsistence. The speaker expresses awe at this expansive evolutionary search and suggests that replicating intelligence requires retracing this entire process from the very beginning.

They argue that intelligence cannot be reverse-engineered or abstracted as a standalone entity because it is not a designed system but an emergent property of complex chemical and physical interactions. Intelligence, as humans experience it, is a unique outcome of this evolutionary history, and attempts to model it mathematically or algorithmically are merely approximations—maps rather than the actual terrain. The speaker stresses that intelligence, as a concept, is an arbitrary label assigned to certain human behaviors rather than an independent phenomenon that can be directly recreated.

Consequently, the speaker concludes that there is no straightforward path to achieving artificial general intelligence (AGI). This is not due to any mystical or spiritual uniqueness of humans but because true intelligence requires undergoing the extensive evolutionary search process that produced humans. Current AI systems are described as “hacky shortcuts” that mimic human language and behavior based on training data, lacking the depth and continuity of natural intelligence. While AI has practical uses and is “cool,” the speaker remains skeptical about the existence or attainability of genuine AGI.

In closing, the speaker emphasizes that humans are the embodiment of intelligence, inseparable from it, and that intelligence cannot exist independently of living creatures. Computers and AI, being abstract and virtual constructs, cannot become creatures and thus cannot possess true intelligence. The video ends with a philosophical reflection on the uniqueness of human intelligence as the product of billions of years of natural processes, inviting viewers to appreciate this perspective while continuing to explore the complex landscape of AI and intelligence.