Are We at the End of Ai Progress— With Gary Marcus

In this episode of Big Technology Podcast, Alex Kantrowitz interviews Gary Marcus, a leading AI skeptic, about the current state and future of AI. Marcus argues that the scaling laws that once predicted exponential improvement in AI models with more compute and data are no longer holding true, leading to diminishing returns. He points out that despite significant investments, recent AI models haven't shown the dramatic improvements seen in earlier versions like GPT-4. Marcus expresses concerns about the black box nature of these models, their tendency to hallucinate, and the potential for misuse, especially with open-source AI. He suggests a shift towards neuro-symbolic AI, combining neural networks with classical AI approaches, and warns about the risks of relying on AI for critical thinking and the potential for AI companies to exploit user data for surveillance and targeted advertising.