AI's trillion-dollar dilemma

The moon landing was cheaper?
I don't want you to think that I'm against artificial intelligence. I'm not. I understand it has a lot of potential. But I do think there is a reason why it's being promoted so heavily, and it is not always just about the technology's abilities. Which is resulting in some questionable applications of it.

The tech industry is plunging about $1 trillion a year into building out the hardware and services for AI. One trillion. A year. That is a gargantuan amount of money. "It would rank among the largest concentrated capital build-outs in economic history," ChatGPT told me. (See? I can use AI, too.) As a share of GDP, it is a larger expenditure than anything else in this nation's history apart from the Louisiana Purchase, according to The Wall Street Journal. Bigger than the railroads. Bigger than the interstates. Bigger than the moon program.

Now, those numbers are so big I don't think you can really wrap your head around them; at least, I can't. But what you can wrap your head around is this: $3 trillion worth of capital invested over the next few years will place an enormous amount of pressure on the borrowers to show a return on all that invested capital. Which is why, in my opinion, the industry is cramming AI into every nook and cranny of our lives that it can possibly be crammed into. They don't have a choice.

The latest example of this is Robinhood, our Joey Pizzolato reports, which announced it's starting two services that will use AI agents, one for trading and one for credit cards. This follows ChatGPT's announcement last week that it was starting a financial advice program for users of its Pro version. This is being seen by some as evidence that the entire banking industry should scramble to launch AI financial-advisor services, or automated spending and trading services. Or something. Anything.

Can we see what kind of financial advice ChatGPT hands out before declaring it the industry standard? It's not like these things have some sterling record of giving good advice in other instances. Relationships, for example. Even worse, there are multiple instances of AI bots advising its users to commit suicide.

What will happen when ChatGPT tells you to go all in on some stock or index fund that crashes to the ground? Will it just offer one of its patented "You're right, that was terrible advice" apologies? If you're a bank offering this service, do you think that will be acceptable to your clients?

The proper question is not, how can I get an AI program installed, like, yesterday to provide these services. It is, how can I best provide financial advice to my clients? Do I want to provide financial advice to my clients? Is that a valuable addition to my existing services? Is it worth the cost of providing it?

Good financial advice is valuable because there is so much terrible financial advice out there (and god only knows what's being fed into ChatGPT). But the answer isn't to outsource your decisions. It is to educate yourself, or your clients. You want good financial advice? Here, here is Warren Buffett's 2024 annual letter. Here is his 2023 annual letter. Here are all of his annual letters. Howard Marks of Oaktree Capital is another vaunted investor who also happens to be a good, clear writer. Here are all his investment memos going back to 1990. Michael Mauboussin is over at Morgan Stanley now, but has been on Wall Street for four decades and is one of the sharpest minds on the Street. Here's his work going back to the '90s. 

The reason so many people are financially illiterate isn't because good information isn't accessible, it's because they aren't encouraged to take the time to read and learn. It's because the markets have increasingly been treated as a casino. An AI program that takes over people's finances isn't going to change that. Maybe instead of offering a quick fix, banks should offer a reading list and a couple of good books on how to manage your finances and invest. That would actually be a lot cheaper than contracting out OpenAI to do the job, and maybe more effective. And encourage your clients to take the time to read rather than handing off life decisions to a bot whose creator is under pressure to produce an insane ROI because the hyperscalers are plowing a trillion annually into data centers.