When Bernie Sanders and Donald Trump agree on something, that agreement deserves examination. Both are touting the idea that the federal government should have an ownership stake in major artificial intelligence (AI) companies. For Sanders this would be a way for the American people to share in the wealth generated by AI companies that those companies accumulated through the theft of other people's writing, art works, photos, videos and recordings. And it would give government a seat at the table as a representative of the public in deciding how AI companies will operate in the future. Just how Trump views the purpose of government ownership of a share of AI companies is not yet clear.
Problem is, AI companies are not really generating any wealth. Right now they are simply burning through billions of investor dollars that are heavily subsidizing users with no clear path to profitability. And no, the AI industry is not following the path of previous technologies such as the internet and the smartphone in which costs will rapidly fall as the technology matures and greater numbers of people adopt AI. Years into the AI boom the cost of providing its promised outputs is going up, again with no clear path to bringing the cost down. Adding users does not bring down unit costs because more users require more resources. It's the opposite of the software model where the cost of distributing a piece of already written software to the next customer is close to zero.
I warned last November that artificial intelligence (AI) companies were setting the stage for a bailout, one that will be the result of their deeply flawed AI models that won't remotely deliver what they are promising. AI critic Ed Zitron said in a recent interview that when he asks AI industry boosters to tell him what AI can actually do without using the word "will," they have little to offer. Zitron agreed that while search engines are using AI to provide more sophisticated outputs, those outputs have become more unreliable. "You have to check every single bloody thing. You can't rely on anything," he said.
As if to put a fine point on it, just two days after the Zitron interview, the Financial Times published a story about a report on the uses of AI put out by KPMG, one of the Big Four accounting firms, in which the KPMG made several false claims about corporate, governmental and nonprofit uses of AI. The false claims were apparently the product of AI "hallucinations," that is, stories concocted by AI that are simply made up or not factually correct. The report appears to have been withdrawn. The KPMG report was promoting its services to help its clients integrate AI into their organizations. Oops!
And to add insult to injury, the Ohio Business Roundtable, a well-connected business lobby, is trying to get the Ohio legislature to change eminent domain laws to allow AI companies to seize private land for energy projects needed by their data centers WITHOUT PAYING FOR IT FIRST! Why would supposedly wildly successful AI companies not be able to pay people for their seized land BEFORE occupying it and instead be allowed to wait months or years to pay? A big fight is brewing in Ohio and the Ohio Farm Bureau is leading the charge.
There is a lot more to say about why AI based on current models is likely to be a much smaller and less impactful technology than is being advertised. But the financial facts don't lie. If the government is lured into taking an ownership stake in AI companies, it will be walking into a trap. The first clue is that the CEO of the most prominent AI company, OpenAI, proposed the idea to Trump early last year. Why would the top executive of the best-known AI company offer to dilute his wealth and the wealth of his shareholders to give the government a share? Why would that CEO want the government looking over his shoulder as he runs his company? Unless he knows he is going to need a bailout.
And that is likely where we are headed for two reasons: 1) The AI industry has convinced lawmakers and the current administration that AI is somehow the future of technology and therefore the industry (as opposed to the technology) cannot fail and 2) the AI industry's capital expenditures are the main thing driving the economy and the stock market and a bust among major industry players (again, as opposed to the technology) would lead to too much pain in pocketbooks and portfolios.
Bernie Sanders wants the government to obtain its share of AI company ownership through a tax paid in shares of stocks giving the government 50 percent. How the Trump administration would finance such a purchase and how much of a stake it would take is not known. But no matter how such a government stake in the AI industry is achieved, buying it now would almost certainly be buying in at the top. Governments are actually very poor at making financial investment decisions, in part, because they are not set up for that. They make decisions for policy reasons that have little to do with shrewd market timing.
When the current AI bubble bursts, the federal government will be tempted to throw good money after bad to bail out its investment since the government has something private industry lacks, the power to tax and the ability to print money as needed. And, the government will be heavily lobbied by its new partners, AI company management, to proceed with the bailout even as there is no clear path to profitability.
Two clear facts suggest that government involvement in AI for now should be relegated to regulating it. First, the industry cannot explain how, using the current Large Language Model framework, it can both bring down costs and eliminate hallucinations (which make AI very dangerous for any task that would be autonomous, that is, operate without human intervention). Second, the income the industry would have to generate to justify its investments to date and future commitments require something like $2 trillion in NEW revenue over the next four years, according to Zitron.
The entire global software industry had revenues of about $719 billion in 2025. And that was for products that presumably work and have a marginal distribution cost close to zero. So far, according to one recent report, "80 percent of AI projects fail to deliver business value." And, "[o]nly 21% of S&P 500 companies can cite a measurable AI benefit, per Morgan Stanley, even as global AI spending is projected to hit $527 billion in 2026 capex [capital expenditures] alone."
Current AI models may offer at least some value to a wide range of users. Perhaps someone (not me) could make an argument for subsidizing AI as a public good, the way we do with public transportation. The alternative is to have businesses alone continue to deliver AI to various users. But that won't work unless users are willing to pay many times more than AI companies are currently charging for so-called pro or premium services ($100 a month for ChatGPT and Claude). Keep in mind that most users currently pay little to nothing to use AI. The much higher prices needed to make AI a profitable venture would almost certainly cause most users to abandon it since the value it creates is not worth that kind of money (perhaps many hundreds of dollars per month) either to individuals or most businesses—especially given the continuation of widespread errors and "hallucinations" which as Zitron points out means every output has to be thoroughly checked by an actual human.
Let the marketplace sort it out. If someone, somewhere can come up with an AI model that is profitable for businesses to sell to typical business customers and the broad public, let it emerge without government intervention. Right now the AI industry is simply looking for a government sugar daddy to pay the bills that its leaders foolishly incurred so that those leaders can have their personal wealth backstopped when the AI bubble bursts.
Kurt Cobb is a freelance writer and communications consultant who writes frequently about energy and environment. His work has appeared in The Christian Science Monitor, Resilience, Common Dreams, Naked Capitalism, Le Monde Diplomatique, Oilprice.com, OilVoice, TalkMarkets, Investing.com, Business Insider and many other places. He is the author of an oil-themed novel entitled Prelude and has a widely followed blog called Resource Insights. He can be contacted at kurtcobb2001@yahoo.com.
No comments:
Post a Comment