Why Masters of the AI Universe Lean Against What Works
We're Headed for a Fall
It’s clear to me that small AI projects outperform big ones. Limiting the input, and the program’s goals, can automate tasks that now take a lot of time. (Illustration from ChatGPT.)
An example is the Mayo Clinic’s StateViewer, which can identify 9 types of dementia from a single scan with 88% accuracy.
At the same time, it’s clear that more general AI projects aren’t working. The problem is basic, AI’s inability to create a coherent model of its world, which is what you need before you can perform creative tasks.
Yet the AI arms race continues to accelerate, with Meta offering salaries of $10 millionand Nvidia stock back at record highs.
The usual suspects on Wall Street are getting worried but you can’t tell it from my own IRA, whose balance has blown through previous records.
Just like in the 1990s we know the crash is out there, but we don’t know where. It’s like a fire that warms us from a distance but will burn us if we get too close. It’s also clear that this fire, once it starts, is going to burn everyone, badly. It will be more like the one in the 1920s than the one in the 1990s.
I’m taking some profits off the table, looking at bonds and international stocks as a replacement. But I’m also looking forward to the gains AI will bring regardless of what the market does.
Real AI Benefits
Big AI hasn’t crashed because small AI is showing big benefits. (Illustration, using the same prompts as used earlier, by Perplexity.)
I know many writers who use AI in their work. Even I can do research with it. Some will have AI do a first draft, then edit out the mistakes. Others will do their own first draft and take editing suggestions. Freelancers who embrace AI are doing well.
Unlike the Web, which was greeted warmly a generation ago because its benefits were obvious and its dangers less so, AI is almost universally hated. That’s because the billionaires behind the AI arms race have proven, on the whole, to be craven, greedy lunatics. Who knew in the 1990s that Jeff Bezos was such a jackwad?
An Administration following these bozos slavishly is obviously headed for trouble. The decision by Xi Jinping to discipline his tech sector beforehand starts to look wise. China is winning the AI race. This will continue so long as Americans use AI against the common interest. While the Internet did not deliver political upheaval, AI certainly will.
A lot of analysts seem confused over Nvidia. Even the optimists are focused entirely on the data center. I think Jensen Huang’s aim is different. He sees the limits of Big AI, but also sees that robots, with small AI capabilities, will deliver faster productivity gains into the next decade. ““We stopped thinking of ourselves as a chip company long ago,” Huang told his most recent shareholder meeting. But no one I read was listening to him.
What Next?
AI will be a bumpier road than the Internet was, because what’s at its heart, generative AI, isn’t nearly as solid as the Web was. But small AI can use all the computing capacity the Cloud Czars can create. The Czars, meanwhile, will transform from growth companies into utilities, the AT&Ts of the next generation.
As an investor, and as a thinking being, you want to be as far away from government control as you possibly can. That means you should learn what AI can do for you now and be prepared to fix AI tomorrow.