George Hotz Talks About General Purpose Intelligence

In a recent conversation, George Hotz, the CEO of Comma.ai and a prominent figure in the artificial intelligence (AI) community, discussed the challenges of scaling up to general purpose intelligence. He argued that using cross-entropy loss on characters is not sufficient for achieving general intelligence. Hotz has been involved in AI research since his teenage years when he became famous for being the first person to unlock an iPhone. He has since become a leader in autonomous vehicle technology and was featured on the cover of Wired magazine in 2017. During his conversation with Lex Fridman, a research scientist at MIT who specializes in human-centered AI systems, Hotz made it clear that there are still many obstacles to overcome before we can achieve general purpose intelligence. “We’re not just gonna be able to scale up to GPT-12 and get general purpose intelligence, your loss function is just cross-entropy loss on the character, like that’s not the loss function of general intelligence," said Hotz. When asked if this was obvious to him, Hotz responded with a simple “yes”. His answer highlights how much progress needs to be made before we can achieve true artificial general intelligence (AGI). While current AI systems are capable of performing specific tasks such as playing chess or recognizing faces, they lack the ability to think abstractly and reason about complex situations like humans do. Hotz believes that progress will come from understanding how our brains work and developing algorithms that mimic those processes. This could involve using reinforcement learning or deep learning techniques but ultimately it will require an interdisciplinary approach combining neuroscience and computer science principles together with engineering solutions. This conversation between two experts provides insight into some of the challenges associated with developing AGI systems and demonstrates why it is so difficult for us to create machines that can think like humans do. It also shows why we need more people like George Hotz who are willing to push boundaries and explore new ideas if we want to make progress towards creating truly intelligent machines one day soon.