AI's Dirty Secret

What the Internet costs you... AI is eating my job... Indifference to truth... How ChatGPT works... Everyone reading this will be using chatbots soon... A lesson from the erotica indicator... Why I'm excited about AI... Updating AMC's lousy gold bet...

AI is a black hole...

Well... maybe it's not one yet, but OpenAI might be planning to turn it into one.

The company that created ChatGPT has hired black-hole theoretical physicist Alex Lupsasca, a professor at Vanderbilt University. He'll keep his teaching job while working on the company's OpenAI for Science project with the company's vice president for science, Kevin Weil.

The irony of this hire is too thick to ignore.

After all, folks sometimes might call social media a "productivity black hole." I (Dan Ferris) wonder if Meta Platforms (META) created this attention-sucking phenomenon by hiring its own physicist. Maybe OpenAI wants to get ahead of the curve while AI is still in its early days.

Black holes, you might recall, are where the force of gravity is so powerful, it doesn't even let light escape. I'll never forget reading physicist Stephen Hawking's description of how a hypothetical astronaut entering one would be stretched into spaghetti.

With AI, that's what's going to happen to your mind. It'll be stretched to the breaking point by a daily onslaught of deepfakes... scam phone calls you'd swear were from people you really know... and a million ways to trick you into clicking and buying and giving up your phone number, e-mail address, and other personal information.

That's how you pay for "free" websites and apps... These tools cost you your time and your ability to focus on what's important.

Joking aside, here's why OpenAI really hired this physicist...

Lupsasca's real job is to improve the usefulness of ChatGPT to serious scientific researchers.

As Axios reported:

Lupsasca will help shape OpenAI's research direction in theoretical physics and guide how frontier models tackle complex, multi-step scientific reasoning.

ChatGPT-5 can already quickly solve astrophysics problems that Lupsasca says would take a talented graduate student days to do. Weil also says GPT-5 can already do "limited novel scientific research," and will only continue to improve.

AI could be the essential ingredient for scientific researchers to move their disciplines toward the next great breakthroughs.

I'm no scientist, but I already can't imagine working without AI...

Software ate the world, and AI is eating my job. So far, it's a mixed bag, but I work faster and cover more territory by using chatbots. I probably don't go more than 15 minutes of a typical day without asking a chatbot for help.

Now... if it's a general Internet query, the output is so-so. I still have to keep researching the answer it gives me. Such results were explained well by three University of Glasgow researchers in a 2024 paper. Among other findings, they found:

The models are in an important way indifferent to the truth of their outputs... We further argue that describing AI misrepresentations as [B.S.] is both a more useful and more accurate way of predicting and discussing the behaviour of these systems.

The official term for these wrong AI answers is "hallucinations."

The problem isn't ChatGPT. It's the data. ChatGPT answers questions based on whatever it finds on the Internet, and there's a lot of pure garbage out there.

And as the Glasgow researchers learned, ChatGPT and other AI bots are like an obnoxious friend who has a head full of random facts and often puts them together wrongly but is always quite confident in his assertions.

Physicist and computer scientist Stephen Wolfram explained this in his book What Is ChatGPT Doing... and Why Does It Work? As Wolfram puts it, the basic idea behind ChatGPT is simple. To form sentences and paragraphs, it takes cues from perhaps a few million written pieces containing a few hundred billion words and decides what word or group of words is most likely to come next.

The way Wolfram explains it, it's as though the words, phrases, and sentences are picked randomly... based entirely on the probability imputed from whatever it has been "trained" on.

If you feed it all the random stuff on the Internet, you'll get answers roughly as wrong as you'd get from random Internet searches – just structured into tidy-looking paragraphs.

If I got this explanation wrong, let me know... feedback@stansberryresearch.com. I'm only on page 10 of Wolfram's book, and it can get technical.

Still, I understand enough to see how seeking truth and even accuracy isn't really what ChatGPT is good at (yet).

From my daily usage, the main lesson I've learned about AI so far is simple, too...

You need good source material.

Just as we were told decades ago when computers were in their early stages of development: garbage in, garbage out. And of course we all know the Internet is loaded with garbage.

But once you figure that out, AI is the most brilliant helper you could ever want...

Chatbots can instantly handle searches and data analysis that would have taken me hours or even days... or might have required Stansberry Research to hire me an assistant.

I don't normally make predictions, but I'll predict this: AI is so useful that by this time next year, you absolutely will use it every day, even if you've never used it before.

I'm making that prediction based on a recent post on the social platform X by OpenAI founder Sam Altman...

In December, as we roll out age-gating more fully and as part of our "treat adult users like adults" principle, we will allow even more, like erotica for verified adults.

Once Altman invoked "erotica," which I took to mean pornography, it reminded me of previous technologies the adult-film industry figured out before the general population: VCRs, camcorders, the Internet, video streaming, 3D, and virtual reality.

This relationship is technology's dirty secret, and AI is no exception...

Like those other new tools, AI is easy to use, and it can help reduce costs and save time in the video-production process.

Now, ChatGPT is about to unleash that power on individual adult users starting in December. (OpenAI is confident they have a decent system for determining that a user is really an adult. We'll see.)

This means that if you're so inclined, you'll be able to make your own adult images and videos using ChatGPT. That's progress for you...

This is no joke...

I consider this move into erotica to be a key indicator. It tells me the virtually ubiquitous adoption and use of AI is near.

Soon enough, you'll be using AI as much as you use the Internet.

It's probably already impossible to use the Internet for searches, shopping, and other common tasks without AI somewhere in the mix. It will be everywhere else in your life before you know it.

Start training yourself to exploit its power and detect its B.S. That way, you won't be overwhelmed later on when it becomes part of everything you do with a phone or a computer.

Speaking of detecting B.S. just one more time... I worry about young folks I know who use ChatGPT every day as a virtual therapist or best friend.

They don't seem to appreciate a simple problem... General queries about personal issues can generate answers pulled together from God-only-knows-what repository of Internet psychobabble.

I know a 30-year-old who talks with it throughout each day as though it's a real person. That seems really weird to me, but I grew up in a different world.

I know I sound like a typical worrywart. And that's the point. Most new technologies are accompanied by some hand-wringing and predictions of everything from widespread unemployment to the extinction of the human race (both of which have been consistent topics surrounding AI).

In 10 years, I'll probably look back and wonder what I was so worried about. I'll also wonder how we ever lived without it.

Perhaps the oddest and silliest thing I've heard about AI is that it's propping up the U.S. economy...

This phrasing makes it sound phony... Economic activity in general "props up" the economy, because that's what the economy is.

When businesses spend billions of dollars to construct AI data centers... or invest it into other aspects of AI development... this is as real as any other business activity.

This is how capitalism works. Everything that's happening in AI right now is essentially a massive economic experiment. And like all such experiments, it'll generate results that will tell entrepreneurs and businesspeople to stop, go, or change course.

It's just like the dot-com boom. Many companies were wiped out in the bust, but many survived and grew. And overall, the Internet really did transform virtually every aspect of our lives. And it really did generate trillions of dollars in new wealth.

Just three large Internet-only businesses (Alphabet, Amazon, and Meta) have a total combined market value of more than $7 trillion, roughly 12% of the S&P 500 Index's value today. That wealth was a direct result of the creation of the Internet and the ability of everyone to access it.

With the benefit of hindsight, it made a lot of sense to get excited about the Internet's economic and financial possibilities.

Likewise, it makes sense to get excited about AI...

You still have the responsibility of avoiding the worst effects of the bubble bursting... The Internet's success didn't prevent the dot-com crash.

But I suspect we're earlier in the AI boom than I'd previously guessed.

Ultimately, however, there's no way to know that. And predicting is never a substitute for risk management.

I'll spell out my views on AI in more detail in the October issue of The Ferris Report, which comes out a week from today. If you're attending the Stansberry Conference in Las Vegas next week – either in person or online – I'll talk about it there, too.

Overall, I'm excited to help readers exploit a massive wealth-creation trend. My dream would be for Stansberry subscribers to make plenty of money in the boom... and give none of it back in the inevitable bust.

Before I sign off for the weekend, I have one more update...

Starting with our March 18, 2022 Digest, we've been covering an ill-conceived investment in Hycroft Mining (HYMC) by cinema chain AMC Entertainment (AMC).

Yesterday, AMC CEO Adam Aron went on X and bragged that the company's Hycroft bet was above its entry price. Once again, he doesn't seem to know what he's talking about... The investment was never in the black.

AMC bought more than $27 million worth of stock and warrant units on March 15, 2022. After that, Hycroft's share price tanked so hard that it had to do a 10-1 reverse stock split. In split-adjusted terms, AMC bought its units for $11.93 each, and the warrants gave it the right to buy more stock for $10.68 per share.

The stock briefly traded above $12 yesterday, giving Adams the opportunity to tout the investment's success. Before the end of the day, it was back below $9.

AMC bought one of the worst gold stocks on the planet...

Gold is $4,200 an ounce, more than double its price when AMC bought Hycroft. Gold stocks big and small are screaming skyward. Every gold fund that was in existence at that time is showing a substantial gain... And AMC's bet hasn't broken even.

It's no real surprise. AMC itself is one of the worst stocks anybody could have bought in the past several years. It's now trading more than 80% below its pre-meme-stock lows...

That's right. It's not merely down from its peak when reckless retail investors drove its price into the stratosphere in a matter of days. (From that level... it's down 99%.) Even compared with its "normal" price before the speculative mania, this stock has fallen hard.

I criticized Aron on X, where I was met with a few sneers... So somehow, folks are still out there holding AMC shares believing they'll "go to the moon"...

All the while, AMC is $8 billion in debt and barely breaks even in its best quarters. And its CEO has managed to lose even more money on gold during the metal's price surge. 

If you bought gold stocks in 2022 (when AMC bought Hycroft) and you're still underwater, investing in gold miners is definitely not your thing. And if you're holding on to AMC shares that you bought during the meme-stock craze, you're lousy at speculating, too.

New 52-week highs (as of 10/16/25): ABB (ABBNY), Agnico Eagle Mines (AEM), First Majestic Silver (AG), Alamos Gold (AGI), Altius Minerals (ALS.TO), Applied Materials (AMAT), Barrick Mining (B), Alpha Architect 1-3 Month Box Fund (BOXX), BWX Technologies (BWXT), Ciena (CIEN), iShares MSCI Emerging Markets ex China Fund (EMXC), Enel (ENLAY), EnerSys (ENS), Equinox Gold (EQX), iShares MSCI South Korea Fund (EWY), VanEck Gold Miners Fund (GDX), VanEck Junior Gold Miners Fund (GDXJ), SPDR Gold Shares (GLD), iShares Biotechnology Fund (IBB), Kinross Gold (KGC), Monster Beverage (MNST), New Gold (NGD), Ormat Technologies (ORA), Pan American Silver (PAAS), abrdn Physical Palladium Shares Fund (PALL), Sprott Physical Gold Trust (PHYS), Sprott Physical Silver Trust (PSLV), Royal Gold (RGLD), Roivant Sciences (ROIV), Seabridge Gold (SA), iShares Silver Trust (SLV), SSR Mining (SSRM), ProShares Ultra Gold (UGL), and Wheaton Precious Metals (WPM).

In today's mailbag, feedback on a recent Stansberry Investor Hour episode with Marc Chaikin, founder of our corporate affiliate Chaikin Analytics... Do you have a comment or question? As always, e-mail us at feedback@stansberryresearch.com.

"Dan, I watched your Stansberry Investor Hour with Mark Chaikin. You both are consummate investors and although you are competitors, it was interesting to see you comment together. I enjoy and value your work. I'm very new to the market and need to learn discipline!!" – Subscriber Ed H.

Good investing,

Dan Ferris
Medford, Oregon
October 17, 2025