Beware the Frothy AI Frenzy

Photo: Cfoto/Zuma Press

Artificial intelligence continues both to astound and confound. Google’s recent launch of its new Gemini AI tools was a mess, producing images of Founding Fathers and Nazi soldiers as people of color. When asked if Elon Musk or Adolf Hitler had a more negative effect on society, Gemini responded that it was “difficult to say.” Google pulled the product over “inaccuracies.” AI is the next wave of computing and human interface but it’s definitely out over its skis.

But Gemini’s problems say more about Google than they do about AI. A Google “AI Ethics & Compliance Advisor” is on video talking about “deliberate steps to ensure that the advanced technologies we develop and deploy lead to a positive impact on individuals and society.” This is DEI-speak. No wonder Gemini’s output is screwy.

AI finds patterns of patterns in data and then uses a technique called “reinforcement learning from human feedback.” Humans help train and fine-tune large language models. Some humans, like “ethics & compliance” folks, have a heavier hand than others in tuning models to their liking. Google, with 190,000 employees, controls 90% of search. So how did this flaw make it into the marketplace? A bloated bureaucracy. Alex Kantrowitz at the Wrap quotes a member of Google’s Trust and Safety Team: “Organizationally at this place, it’s impossible to navigate and understand who’s in rooms and who owns things. Maybe that’s by design so that nobody can ever get in trouble for failure.” Business 101: Hiding failure limits success.

Advertisement

There’s a long history of dominant players missing the next big market because of bloat and hubris.

Xerox

,

IBM

, Digital Equipment, Compaq,

AT&T

, AOL, Netscape, Yahoo. I can go on. The Federal Trade Commission should Sharpie this on employee foreheads: There’s no need to break up technology companies, as they will diminish or fail all by themselves.

In the 1990s,

Cisco

owned most of the router market. During the dot-com explosion, investors believed a long-run rebuild of telecom infrastructure was imminent to build out the internet. In 2000 the company’s stock peaked at a $555 billion value. The internet got built all right, but Cisco shares crashed almost 90%. It has taken the company 24 years to crawl back to its $200 billion value today.

Change takes a toll.

Microsoft

is now a data-center company, more utility than innovator. Notice how Microsoft lets OpenAI make all the mistakes and hallucinations with ChatGPT and stays comfortably distant from problems.

Amazon

masquerades as an online retailer but is really also a data-center company. Plus, the tech giants’ quest for the next platform is treacherous.

Facebook

controls social media but has lost $42 billion investing in the still-nascent metaverse.

Apple

owns the smartphone market, which saw unit sales peak in 2018, and it lost $10 billion over 10 years not building its Titan electric car. Good thing.

Tesla

is stuck in the Biden administration-created EV glut. Instead, Apple thinks we’ll all wear expensive augmented-reality ski goggles, like bumblebee eyes. The Apple Bee look is beyond dorky.

Rounding out the Magnificent Seven stocks driving today’s market is

Nvidia

, which momentum-obsessed investors drove past $2.3 trillion in value last week. Nvidia CEO Jensen Huang talks of computing going from retrieval to generative, which investors believe will require a long-run overhaul of data centers to handle AI. All true. But that premise about an overhaul also was true for Cisco.

Follow the money. Microsoft invested $13 billion in OpenAI for just under half the company—the for-profit entity anyway—to help develop and roll out ChatGPT. But much of that was funny money—investment not in cash but in credits for Microsoft’s Azure data centers. AOL used to do this logrolling from its balance sheet to reported revenue by forcing startups to buy its ads in exchange for investment. That didn’t end well.

Microsoft then spends real money on Nvidia H100 processors to speed up AI calculations. Nvidia makes the H100 with chipmaker

TSMC

, for an estimated cost of $3,300 apiece. Because of shortages, Nvidia can sell each processor for $30,000 or more. TSMC, in turn, buys Extreme Ultraviolet lithography machines, also in short supply, from ASML in the Netherlands.

Microsoft and Facebook are reportedly Nvidia’s biggest customers, buying these processors last year with a frenzy. You can see why Nvidia makes 62% operating margins and its stock has headed toward Jupiter. But for how much longer?

Fat margins attract competition, including AMD, Intel,

Qualcomm

and others. Google has its own AI chips. So does Microsoft. It’s impossible to predict how fast additional fabricating capacity comes on, but soon there will be more AI chips than snow in Lake Tahoe last week.

Meanwhile, the Bubblicious bat signals are flashing red again: Bitcoin,

Carvana

, Super Micro are all flying—even the meme-screamer Reddit is readying its IPO. Momentum investors can be relentless, so AI is up, up and away. But remember, while markets continue to grow, valuations peak and leadership often changes. Be careful out there.

Wonder Land: Progressive policy decisions on—among other things—crime, higher education and migrants are hurting Americans. And President Biden gets the blame. Images: AFP/Getty Images/Reuters/Bloomberg News Composite: Mark Kelly

Advertisement

Copyright ©2024 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8

Appeared in the March 11, 2024, print edition as 'Beware the Frothy AI Frenzy'.