SDS 469: Learning Deep Learning Together — with Konrad Körding

General Intelligence

The pursuit of artificial general intelligence (AGI) presents significant challenges, particularly in the realm of symbolic representation. Konrad Körding highlights the difficulty of teaching neural networks to understand and manipulate symbols, which humans do naturally. He explains that while humans can easily imagine scenarios and understand causality, these tasks remain complex for AI systems.

We can't at the moment explain those things meaningfully to algorithms. This thing add an object to a scene, it's remarkably hard and similarly causal things like, if I have a phone and I drop it, what's going to happen?

--- Konrad Körding

Jon Krohn adds that understanding causality is crucial for AGI, as it allows systems to interact meaningfully with the world 1 2.

   

Reinforcement Learning

Deep reinforcement learning has achieved remarkable success, particularly in games like Go, where algorithms like Alphago and Alphazero have surpassed human capabilities. Konrad Körding explains that these systems combine intuitive position evaluation with strategic planning, allowing them to excel in complex environments. This blend of intuition and rule-based understanding is key to their success.

It combines a position evaluation, a position evaluation. Basically, you look at the bot, and you're like, that's great. I want to play this versus. This is a lost game.

--- Konrad Körding

Jon Krohn notes that these algorithms learn without human biases, offering new insights into game strategies 3 4.