FROM THE FRONTIER
Ex CTO of Yahoo: This is where AI is heading
Click here to watch the full conversation
As former CTO of Yahoo and founder of multiple successful exits, Ash Munshi has worked through almost every major tech wave since the 80s. He’s now CEO of Pepperdata, an AI data optimization company that lists two of the Fortune Five companies as customers. We sat down with him to get his take on where AI is headed.
Here are some of the key takeaways from our conversation:
AI doesn’t actually know how to learn. According to Munshi, this is one of the biggest problems with current LLMs, which are trained on massive datasets but can’t learn from experience the way humans do. Munshi believes that the real breakthrough will come when models observe, experiment, and learn from failure.
Human-in-the-loop is where the money is. Instead of chasing full automation, smart builders are working to ramp up human capabilities. “If you can suddenly 10x the engineer, your velocity increases and you can contribute more positively,” Munshi told us.
Companies are bleeding cash on AI infrastructure. Most companies size GPU resources for maximum load, but utilization tends to fluctuate. At Yahoo’s scale, for example, running at 80% efficiency would mean that about 200K machines are basically underutilized. Pepperdata pushes utilization to 95%, saving companies hundreds of millions annually.
China’s catching up faster than anyone wants to admit. “The lock we had on intellectual capital is becoming unlocked,” Munshi warned. He highlighted that Chinese-authored AI papers now exceed American ones, its universities rival top US schools, and it controls rare earth materials critical for chips and batteries.
via Superhuman