Paper by Mehdi Bugallo:
Cognitivism, which has permeated society—as evidenced by the omnipresence of the terms “cognitive” and “cognition”—has perpetuated a traditional view of thought and intelligence as phenomena of inextricable complexity, and therefore phenomena that we can hardly imagine recreating artificially. This approach has prevented us from anticipating and continues to prevent us from understanding what is happening. Behaviorism, on the other hand, allows us to apprehend complexity through the simple processes from which it emerges and provides the framework for understanding current AI. According to this approach, here is what is essential to understand about psychology: the environment shapes the behavior of organisms via two processes, natural selection and associative learning; the first process structures the brain over generations, establishing a “pre-wiring” that provides the basis upon which the second process structures behaviors over the course of the individual’s life.
The idea of artificial neural networks functioning on associative principles is fundamentally simple, and it is not new (Geoffrey Hinton had been working on this idea for decades when he received the Nobel Prize in 2024). But for such a system to yield results, it needed to be able to integrate billions of parameters, something that was only possible with current graphics cards (GPUs); hence the sudden improvement in AI.
More here.
Enjoying the content on 3QD? Help keep us going by donating now.