Will the first AGI be built mostly within the deep learning paradigm?
Get Ṁ600 play money
Sort by:

Is there any criteria to define “mostly”?

@capybara Having thought about that, it is quite hard to come up with very formal definitions. The “share” of deep learning must not be significantly lower than in current foundational and other frontier models. Adding Monte Carlo search is allowed. Adding different scaffolding is allowed. What is not allowed is the presence of a separate learning or “reasoning” algorithm/paradigm which is essential, according to the consensus at that time, for this AI. In ambiguous cases, I will consider it rather mostly deep learning than not. In very ambiguous cases, I will create a separate market to answer this question.