Chief Investment Strategist, The Freeport Society
You know I’m a believer in AI.
I use it daily.
So, I’m not saying this as a skeptic or as a hater. But I think we’re starting to hit a brick wall of sorts.
OpenAI came out with their latest iteration of ChatGPT (GPT-5) earlier this month.
It didn’t go well.
Users are complaining that it’s not noticeably better than the last iteration.
It still hallucinates… badly. Even when you tell it not to.
And apparently it’s less friendly than before (although I don’t care how friendly it is or not… it’s AI!)
If you’re not familiar with what AI hallucinations are, remember how generative AI works. It doesn’t actually “think.” It imitates. It’s trained to give you an answer that looks like the correct one.
Often, that’s great!
But I asked it to do a simple time-value-of-money calculation… and it whiffed. It gave me a number that looked plausible but wasn’t accurate.
In other documented cases, it apparently insisted that “blueberry” had three bs.
It named imaginary books in a summer reading list.
There’s a long list…
I’m not going to throw ChatGPT or any other AI model under the bus. It just goes to show how quickly we become desensitized to how amazing the technology is.
But, the latest release is a far cry from the major step toward artificial general intelligence (AGI) that OpenAI founder Sam Altman promised.
And it was really expensive to build.
OpenAI doesn’t say what it costs to build and train each model, but we have reasonably accurate estimates based on GPU and power usage.
- GPT-2 was released in 2019 and cost several hundred thousand dollars to develop.
- GPT-3 was released in 2020 and cost somewhere between $5 million and $12 million.
- GPT-4 was released in 2023 and cost around $300 million.
- GPT-5 was released this year and was estimated to cost over $1 billion.
We can assume that Elon Musk’s Grok, Google’s Gemini, and the other major players in this space cost a similar amount.
And these are just the development costs, of course. I’ve said nothing about the ongoing running costs.
It’s entirely possible these estimates are off. But even if they’re overstated (or understated!) by half, the takeaway is the same. It costs a lot of money to build these models.
But thus far, it’s unclear if they’re making much money.
It’s even less clear if the rest of corporate America is seeing a return on their investment in rolling out AI into their businesses.
A report that the global consulting group McKinsey & Company recently published said that eight in 10 companies have reported using generative AI. But eight in 10 also reported “no significant bottom-line impact.”
And researchers at MIT published a report showing that 95% of the generative AI programs launched by companies failed to do the main thing they were intended for – ginning up more revenue.
Understandably, it’s hard to calculate any productivity gains resulting from AI in office environments.
But the suits want to at least see some measurable return. And they’re simply not.
(I’m not talking about AI and automation in warehouses and manufacturing plants. That’s a different animal.)
So, it came as little surprise this morning when the mainstream media began jabbering about the “AI Vibe Shift.”
A Large Correction May Be Looming
I’ve been sounding the alarm about growing unease in the markets. Now, a slew of big names are joining in the chorus, including JonesTrading’s chief market strategist Mike O’Rourke.
Hamish Douglass, former Magellan Co-Founder has warned that AI might disrupt markets so severely that up to half of the share index could collapse over the next half decade.
Vanguard’s Chief Economist, Joe Davis, says there’s a significant mismatch between AI valuations and what’s likely achievable.
Even OpenAI CEO Sam Altman is talking about bubbles.
As a result, traders are rushing to buy “disaster puts.” These are options that act as a kind of insurance for when the market drops.
So, is it time to jump ship?
Not necessarily.
Yes, caution is warranted. If you’re holding positions that have significant gains, raise your stop-loss levels or consider taking some profits off the table.
But don’t simply clear out your portfolio and rush to cash. There is no way to time to top so ride the momentum until it stops.
What you need to do urgently at this point is expand your investment strategies beyond simple buy and hold… and expand beyond just AI.
As I told our Freeport Investor readers last week, there is a great opportunity in Florida real estate right now. I don’t mean owning physical real estate. Rather, investing in companies who already own real estate and are making good decisions with how to monetize the land is the way to go.
And legendary tech investor, Luke Lango, has created a strategy that allows you to target “AI Income Events.”
He explains how it works in this briefing.
To life, liberty, and the pursuit of wealth.