Jimmy Ba, an xAI co-founder, was the final speaker of the day at the Cerebral Valley AI Summit, and he almost didn’t make it when his Tesla self-driving vehicle mistakenly steered onto the Bay Bridge.
Ba, who studied under famed AI researcher Geoffrey Hinton, reports directly to Elon Musk, and said he and the rest of the team look to Musk for guidance internally, and closely follow his tweets. “Every time the dude doesn’t post on X for more than 4 hours everyone gets nervous.”
He boasted about X, formerly Twitter, being one of the best sources of human-made, natural language content for training xAI’s models. He was also firm in the view that scaling laws remain very much in place, and worth pushing to the limit: “Why wouldn’t you want to have a model that’s 10x smarter?”
But he didn’t have very convincing answers when Eric challenged him on the “MechaHitler” debacle, when xAI’s Grok began spewing pro-Nazi content. He said the company’s goal was for Grok to find truth from “first principles,” but couldn’t explain how that squared with the reality that large language models are built from the corpus of words on the internet.
