IN PROGRESS! echoing LEAN methodology, many of these blogs are iterative and incomplete in nature :p Some are just paragraphs, for me to revist and keep cooking. With that said, I like to write in real time and draft in real time. Some are a sentence, some various paragraphs, some are just code i wrote, and some back and fourth human machine collaboration with GPT. I’ll be decompressing them slowly and fine tune them to my own vision ;)
To tell the story, I will divide AI into three periods: the early days, the symbolic era (Symbolic AI), and the natural computing and statistical era (Connectionist AI). Of course, distinct categorization is not always the case. Theres been ongoing pursuits of hybrid models (e.g. Neurosymbolic AI) that uses both symbolic and connectionism to design intelligent systems.
Let me introduce why such pursuit exist, and how, we, “tool-making” species, always somehow end up reverse engineering nature to our own designs. We sought inspiration from birds for airplanes, from fish for submarines, and from nature's dynamics for trains. In a similar vein, we now seek to emulate and understand the essence of human intelligence.
Intelligence and the pursuit of Flight
Formalizing Artificial Intelligence, there exists paradigms of AI, or methodologies to which people sought to achieve “intelligence” artificially. Each methodology (Connectionism, Evolutionary, Symbolic) has their own methods to how we can emulate human intelligence (purely symbolic and logical in nature, prediction = intelligence, or darwinian’s algorithm of natural selection). Each has their own benefits and drawbacks, and now - most modern AI systems and researches are merging symbolic, logical, prediction, and evolution (i.e, reinforcement learning) into one!