Kontxt Kontxt @kontxt
In the latest episode of the Dwarkesh Podcast, Andrej Karpathy discusses the current state and future of artificial general intelligence (AGI), emphasizing that it is still a decade away. He critiques reinforcement learning for its inefficiencies and highlights the challenges faced by large language models (LLMs) in terms of continual learning and cognitive deficits. Karpathy also shares insights from his experience in self-driving technology, explaining the extensive iterative work needed to ensure safety and competence in real-world applications. He concludes by expressing a desire to enhance education through innovative methods, ideally creating a system where learning becomes engaging and effective.