I got a lot of feedback on my AI Bubble? : AI Skeptics Hard Look at the Value Equation<\/a>, and the reactions were fascinating. A lot of you read my analysis of the economics and put me in the “AI Skeptic” box. <\/p>\n\n\n\n That\u2019s a mistake. It misses the physics of what is actually happening.<\/p>\n\n\n\n While I agree with many of the skeptical points regarding the current financial bubble\u2014because let’s be honest, the math on some of these valuations is ‘no bueno’\u2014my long-term view aligns much more closely with Jim Keller.<\/p>\n\n\n\n\n\n\n\n If you don’t know Jim, you should fix that (I can’t tell you how many time’s I’ve sent people Jim’s talk on AI Hardware<\/a>). Jim is the CEO of Tenstorrent and a legendary chip architect. The guy understands the fundamental laws of computing better than almost anyone. He recently gave an interview<\/a> where he nailed the situation perfectly.<\/p>\n\n\n\n His take? The AI market is going to crash. It will go down. And then, it will go back up. Probably a few times.<\/p>\n\n\n\n His advice is simple: you have to “low pass filter all that stuff.”<\/strong> (how could you not fall in love with thinking like that!)<\/p>\n\n\n\n For those of you who didn’t waste your youth with an oscilloscope, a low pass filter is a circuit that removes high-frequency noise\u2014the static, the spikes, the frantic jitter\u2014and lets the deep, sustained signal come through.<\/p>\n\n\n\n Right now, the market is all high-frequency noise. It\u2019s frantic. It\u2019s people running around with their hair on fire because they think if they don’t integrate a chatbot by Tuesday and AI-wash their 10-K filing, their company dies. That is “incidental complexity.” It\u2019s a distraction.<\/p>\n\n\n\n Jim\u2019s point is that if you filter out that noise and look at the signal, the trajectory is undeniable: in 10 years, AI is going to dominate computing.<\/p>\n\n\n\n This aligns with the immutable law of technology adoption that Bill Gates drilled into us: We always overestimate the change that will occur in the next two years and underestimate the change that will occur in the next ten.<\/strong><\/p>\n\n\n\n If you only<\/strong> focus on the bubble bursting, you\u2019re going to miss the revolution. First – it might not burst. Second, if it does is is just the market’s way of resetting our calculations about timing and path to market. The companies with AI Underpants Gnomes’ <\/a>business models will die. The ones solving real problems with real physics will survive and become the giants of the next decade. Third, if it happens, there is going to be a lot of cheap AI capacity around that will fuel the next cycle on a much better foundation.<\/p>\n\n\n\n I used to tell people that with AI, we are at the “Beginning of the Beginning of the Beginning.”<\/p>\n\n\n\n But looking at the velocity of change, the dramatic improvements in capabilities (TPU, GPU, Networking, algorithms, etc.) , and the crazy drop in the cost of inferencing, I\u2019m updating my estimate.<\/p>\n\n\n\n We are now at the Beginning of the Beginning.<\/strong><\/p>\n\n\n\n Don’t half-ass your strategy by betting on the hype. And don’t half-ass it by betting on the crash. Look at the 10-year signal. That\u2019s where the truth is.<\/p>\n","protected":false},"excerpt":{"rendered":" I got a lot of feedback on my AI Bubble? : AI Skeptics Hard Look at the Value Equation, and the reactions were fascinating. A lot of you read my analysis of the economics and put me in the “AI … Continue reading