The conversation around AI at last year’s Davos was a mixture of hype and panic. Now, nobody’s panicking and hype has been replaced by a conversation about getting “ROI” on large language models. But if business leaders think it’s time to put the shock and awe of the “ChatGPT moment” behind them and get back to everyday business, they run the risk of being wrong again. Since last year, we’ve seen massive leaps in AI capability that prove out the so-called “scaling laws,” which say compute power is directly associated with AI capability. The world’s most valuable companies — which have the best window into AI capabilities — are about to spend hundreds of billions of dollars this year on new AI compute clusters that are going to require a rethinking of the entire US energy grid. This will make the cost of software creation negligible. If you thought software was “eating the world,” think again. It was only nibbling at it. We are about to enter an era where everything that can possibly be automated with software will be automated. Next, we’ll see a revolution in robotics. At that point, software will actually, physically, eat the world. For once, “Davos Man” may not be freaking out enough. That may be because global corporations know they will adapt, and poorer countries — also reasonably represented here at the WEF — see a chance to leapfrog older technology and improve living conditions, as happened during Web 2.0. Davos’ blindspot, as a wave of global elections have shown, is the middle class of developed countries, whose labor markets will be most dramatically changed by AI. The change may happen slowly. That would be bad for climate change, medicine, and other areas in which AI could benefit the world, but a good thing for labor. If it happens fast, Donald Trump’s inauguration will look like a quaint neoliberal ceremony. |