How does the AGI network learn?
There is more than one way to teach an AI!
Before 1980, the dominant method was to program precise rules, and the advantage was that we knew exactly why the AI behaved the way it did. But after 1980, a new neural-network machine learning approach took off and has evolved into the dominant way that almost every AI (e.g. ChatGPT) is trained today. Unfortunately, the modern approach lets AI figure out the details of learning by itself, resulting in humans having no precise idea why AI (e.g. ChatGPT) behaves as it does. AI behavior is hard to predict and potentially dangerous. By combining elements of the old school approach with the new approach it is possible to create safer, and better, AGI.
#agi #ai #aisafety #agi #artificialintelligence #llm @ElevateTechCA @mitaimlclub @iqstudios1