“On two occasions I have been asked [by members of Parliament], ‘Pray, Mr. Babbage, if you put into the machine the wrong figures, will the right answers come out?’ I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question.”
– Charles Babbage, Passages from the Life of a Philosopher (1864)
Just Press the AI Button
Do you add glue to your pizza to keep the cheese from slipping off?
This was the pizza making instruction provided by Google’s AI Overview tool.
When making a pizza: “You can also add about 1/8 cup of non-toxic glue to the sauce to give it more tackiness.”
The genesis of the artificial ‘intelligence’ was an 11-year-old Reddit comment from a joker with a username that’s too vulgar to reveal in these pages. Hint: it starts with the letter F.
Readers who’ve endeavored to learn a new language as an adult know that slang and humor are often the most difficult aspects to pick up and make sense of. AI tools appear to struggle with this too.
In this example, even the most dimwitted aspiring chef can easily discern that the direction to add glue to pizza sauce was obviously made in jest. No one would take this seriously. Not even a child.
Yet the AI bots didn’t get the joke. They thought it was valid, practical advice.
The promise of AI is that it can be better, smarter, and faster than humans. That, with enough graphics processing units (GPUs), people will no longer have to think. They can just let AI do the thinking for them so they can get on with the more fulfilling activities in life – like spacing out and watching TikTok videos all day.
Don’t want to do your taxes? Too lazy to write a research paper on 5th century monks? Want to know the top 10 biotech stocks with greatest moonshot potential? Want to harness AI to front run market wave patterns?
Don’t think. Just press the AI button.
Stitching Sentences
From what we can tell, generative AI uses a prompt to make a collage of search engine results. Then it stitches sentences together using common linking or transition words. The outputs are very mediocre.
Maybe you’ve noticed that Amazon now includes an AI generated review at the top of its customer reviews. It starts with the words “Customers say”. The content is a formulaic summary of reviews.
The first sentence says what customers like about the product. This is followed by a sentence that starts with ‘for example’ and lists several features and benefits. Then there’s another sentence on what customers ‘also like’. Then, there’s the transition statement, ‘that said,’ which is followed with general customer complaints.
The content reads like an amalgamated mismatch of nouns, verbs, and adjectives. There’s also a very synthetic and inauthentic – artificial – look and feel to it.
Amazon added these AI generated summary reviews as a solution to what it calls review fatigue. Too bad the AI generated review collages are a useless waste of time. Why rely on AI vomit when you can quickly read real reviews from real people?
The glue on pizza example, among others, illustrates the general shortcomings of AI generated text. AI, in its current state, lacks true understanding of context, satire, nuances and ultimately the quality and accuracy of its outputs.
Presumably, if not already, AI will derive content from previously generated AI content. Are you prepared for the jabberwocky? The output will be a deep layering cluster of gibberish and nonsense without Lewis Carroll’s charm.
Where’s the value?
Future’s So Bright
The supposed value of AI is not what it is today. Rather, it is the promise of what it can be tomorrow.
Bull markets love a story. They especially love a story that’s centered on new technology.
Automobiles, telephones, moving pictures, radio, and aviation, were the vehicles for the bull market of the roaring twenties. Anything dot-com is what enamored investors in the late-1990s.
These new technologies proved to be genuine. However, many of the companies that pioneered them quickly flamed out. And even those companies that lasted, like Radio Corporation America (RCA), had such a volatile share price boom and bust, that their investors were wiped out.
Have you ever heard of Pierce Arrow Motor Car Company? On January 12, 1928, its President, Myron E. Forbes, remarked that, “There will be no interruption of our permanent prosperity.”
Alas, within 24 months, permanent prosperity was severely interrupted. Within the decade, the luxury car maker was insolvent, and its assets were auctioned off.
Investors in the roaring twenties, as in the late-1990s, thought they were getting rich. They bid up share prices far above what corporate earnings could justify. Ultimately, the bubbles popped, and brutal bear markets followed.
The promise of AI has provided a fertile technology story that a bull market mania has bloomed from. Without question, it has been an exhilarating 21 months since OpenAI launched its ChatGPT chatbot.
Real technology companies with real products and services – like Microsoft, Google, and Apple – got into the game. They didn’t want to be left behind. It didn’t matter if the AI applications were any good. With enough GPUs they eventually would be.
How to Catch a Falling Knife
Here at the Economic Prism, we’re not dismissing AI. Whether we like it or not we believe AI applications will quickly become ubiquitous. They already are. We believe the value is in focused applications, not in designing AI bots to think and act like humans.
But that doesn’t mean investors shouldn’t strive to understand the value of what they’re placing their hard-earned investment capital into. Since the launch of ChatGPT, many investors, blinded by the bright lights of easy riches, carelessly placed their capital at risk.
NVIDIA, a company which supplies the cutting-edge GPUs that make AI processing possible, became the posterchild for the AI bull market mania.
Investors couldn’t get enough of it. On June 18, NVIDIA’s market capitalization topped $3.3 trillion. If NVIDIA were a nation, it would have been the sixth largest economy in the world.
Over the last 7 weeks some gas has come out of the technology sector. As of market close on August 8, and even with the market’s significant bounce, NVIDIA has a market capitalization of $2.6 trillion. By this, upwards of $700 billion of investor capital has been wiped away in less than two months’ time.
Similarly, NVIDIA’s share price, after peaking on June 18 at a closing price of $135.58, dropped to $104.97 – for a loss of 22.5 percent. Should you buy the dip?
We don’t doubt that NVIDIA is a great technology company. But, as investors, we must ask at what share price would it merit our capital?
Like Buffett, we want to buy wonderful companies at a fair price. And the AI story, as far as we can tell, has gotten well ahead of the AI reality. In other words, AI alone doesn’t justify NVIDIA’s peak valuation of $3.3 trillion.
Does $80 per share cut it? What about $60 per share?
Not for our money.
We’re looking for a 70 percent top to bottom decline before our interest is perked. This would put NVIDIA at an entry price of about $40 per share.
For perspective, accounting for the 10:1 stock split that occurred on June 7, this would put NVIDIA’s share price where it was on October 31, 2023 – last Halloween night.
At $40 per share, NVIDIA’s market capitalization would be $984 billion. By comparison, that’s well above Exxon Mobile’s market capitalization of $523 billion. Does this seem a bit lofty?
Maybe. But at that price, and assuming earnings haven’t significantly declined, we’re willing to take our chances and dollar cost average in as the share price bottoms out around $20.
In the interim, you can try your hand at catching a falling knife. With a little luck, you won’t sever your fingers.
[Editor’s note: It really is amazing how just a few simple contrary decisions can lead to life-changing wealth. And right now, at this very moment, I’m preparing to make a contrary decision once again. >> And I’d like to show you how you can too.]
Sincerely,
MN Gordon
for Economic Prism
15 years of ZIRP and trillions of printed money allows you to smell the dip buying frenzy from miles away, like on Monday when NASDAQ future contracts reversed from down 6% to 3%.
Falling knife ?
If it doesn’t fall in a straight line, then it doesn’t fall at all.
It is that simple.
Water cooled servers need water or Arctic/Antarctic plaement. There isn’t enough water to transfer heat away from AI servers for AI growth to continue given current tech. Similar to the EV trend, which requires oil/gas turbines to generate electricity and grid to transfer power, EV’s won’t make it either. EVs and AI are still novelties… interesting with possibilities but not enough to change the world.