Many people assume the Doomer narrative about AI is more credible because the argument seems to run counter to the interests of those making it.
After all, how often do CEOs give lectures about how dangerous their products are? It seems to make zero sense. There must be something to it, or else they wouldn't be warning everyone by turning the volume up to 11. Right?
Wrong. Actually, all the doomerism, panic, and especially the claim that AI is becoming all-powerful and will change our lives and businesses help sell the idea that there must be more to AI than just chatbots. It's no coincidence that the AI DOOM! talk has spiked as Anthropic's IPO approaches.
Anthropic is seeking a $2 trillion valuation, blowing past what anyone would have guessed a year or two ago. If AI is just a useful tool people will generally pay $20 a month for, and programmers will use it with a $200/month subscription, the hype doesn't seem justified.
If it is so powerful and only needs tweaking to keep it safe, like nuclear power or chemical plants, the sky is the limit.
You may have noticed — I sure hope you have — that nobody in these companies is actually suggesting that AI development stop and the machines be turned off. Quite the opposite: trillions of dollars are being invested in new data centers to enable the rapid growth of AI, and a good chunk of the valuation in the stock market is driven by investments in the "hyperscalers" that will sell the computer power that makes AI run.
And China is going full speed ahead, meaning that AI development will continue, and if it is going to kill us all, it would just be a Chinese AI that did it. And probably quicker, too.
So what do AI companies gain from the doom talk, aside from inflating the sense that AI will be more revolutionary than the internet?
Simple: if you look at the economics of AI companies right now, they have a technology they correctly understand is very valuable in the aggregate, but it's also rapidly being dispersed to many companies, reducing their potential pricing power.
In business terms, they have no "moat" to protect them from competition. Chinese models can be downloaded for free for local AI, and they're getting pretty good, while vendors sell access to excellent, frontier-quality models at ridiculously low prices compared to the big guys. Claude and ChatGPT are still the best models, but the difference is small for almost every use case, and the time gap is narrowing to weeks rather than months or years for others to catch up.
You can build the best AI model, and two weeks later a competitor matched it or beat it. For most people, the differences are near zero. Edge cases exist, but not enough to matter.
Then there is the financial overextension of the companies. They make billions, which is not bad at all, but spend trillions, which is scary as hell. Many investors have noticed the math doesn't work, especially because their product is becoming a commodity. So much so that Microsoft, which went all in on ChatGPT and spent billions, is now basically indifferent to using other models or OpenAI's product.
How do you build a moat and stay at the top of the pile? How do you reduce the spend and stop the bleeding? And how does this relate to the upcoming IPOs?
Simple: if people buy the DOOM! scenario, AI companies will have the government build the moat, ban competitors as too dangerous to use, and the "slowdown" will reduce expenses and ease pressure to outpace competitors' progress, at least in the United States. China may still advance AI quickly, but in the US, top companies have a moat that keeps them in the market and keeps other competitors out.
That's an easier sell than "hey, we're becoming vendors for a commodity whose price is dropping like a rock; would you like to give us your money?"
Now I don't want to argue that AI doesn't come with problems. I could name many off the top of my head, and I'll do a deeper dive when I have the chance (hey, I'm a high-volume guy!).
But the same goes for smartphones, and we aren't uninventing them. They are too useful, and AI will be too. I am figuring out how to make it a tool, not a novelty, and the potential is high. It will not go away, and it will make money. But probably not enough money for the companies that went all in this early, given their financial commitments.
Intelligence will be built into more and more products, and most of it will be lower-tier, tailored to the specific needs of that application. Few things require the best models, and the almost-best models are proliferating at an amazing rate.
The model ChatGPT and Anthropic are trying to sell is "AI replaces people," and that is unlikely to happen anytime soon, if ever. What will likely happen is that tasks will be taken off people's plates. Less grunt work, more mind work. That is what happened in radiology. The proper metaphor for an AI is "tool," not "replacement" for humans.
Could AI doom the world? I suppose anything is possible. But if so, the top AI companies aren't acting as if they believe it. They want regulation, slower investment, a moat around their business, government regulations to limit their liability, and to be the #1 thing talked about. Dario Amodei, the CEO of Anthropic, has been warning about AI doom for years, but he also has built what is currently the smartest model on the market.
They aren't shutting themselves down. They are investing like they will be trillion-dollar companies years from now.
So take what they say about doom, or, for that matter, their financial prospects, with more than a little grain of salt.
Editor's Note: Do you enjoy HotAir's conservative reporting exposing the radical left and woke media? Support our work so we can keep bringing you the truth.
Join HotAir VIP and use promo code FIGHT to receive 60% off your membership.
Join the conversation as a VIP Member