With great fanfare earlier this month, the pioneering artificial intelligence company OpenAI unveiled its latest model, GPT-5, an AI system so advanced that using it was said to be like conversing with a Ph.d. It landed with an unsettling thud.
Users who had grown accustomed to its chatty and supportive predecessor, GPT-4o, were put off by the more business-like GPT-5. It “honestly feels emotionally distant,” one wrote. Yes, it’s smarter and faster, many said in essence, but we’re just not feeling it.
This has large implications, starting with what it says about how much some users are being affected by artificial relationships. A lot of people have come to use his AI product “as a sort of therapist or life coach,” OpenAI CEO Sam Altman wrote in response to the criticism. GPT-5, in his description, sounds a little like a parent breaking up a teenage crush.
This fraught moment in the history of human-machine relations comes at a time when corporations are spending billions to keep up in the transition to AI, and growing impatient to see the bonanza profits that AI has been promised to bring. American business is in the middle of a romance with AI that in some ways mirrors the relationships some Americans have reportedly developed with their chatbots.
The economist Noah Smith, in his Substack Noahpinion, has unearthed as troubling a piece of data about the American economy as you will want to read. Over the past two quarters, capital spending for AI development — the money being spent on software development and, crucially, the data centers that AI runs on — has added more to the growth of the economy than all of consumer spending in that period.
That captures the enormity of the bet that is being placed on the promise of AI, before it has been clearly established how long it will take for that promise to be realized. Even if AI proves its worth in the long run, such a rapid shift raises the possibility that, in the meantime, it will become overbuilt and lead to the bursting of a huge financial bubble. The data point raises some questions as well about consumer spending, which has been generally viewed as a strong point of the economy.
What if the question becomes not when, but if, the promise of AI is fulfilled? The technology has already produced some amazing changes, and threatens to be the most disruptive force we’ve yet seen. With the passage of the One Big Beautiful Bill Act, the federal government has embarked on a rapid AI conversion. Yet a debate still rages about what AI really is.
In June, Apple published a research paper, which some have also described as a manifesto, titled “The Illusion of Thinking.” It argues that artificial intelligence programs, the older Large Language Models and the enhanced Large Reasoning Models aren’t really reasoning in the way humans do, but mimicking results from the tons of data they process.
Within a matter of days, “The Illusion of the Illusion of Thinking,” a paper challenging the evidence presented in the Apple paper, had appeared. For many of the uses AI is being put to, it doesn’t really matter if the program is really thinking, or just performing a super job of mix and match. But the question is critical to the future of AI, and therefore to the massive investment the country is making in it.
More so than anything Democrats and Republicans are arguing about right now, this could be the debate future generations will study for clues about where we were headed.
As data centers become more problematic, we’re seeing states show growing interest in other AI opportunities. Stanfield Gray, founder and CEO of DigSouth Tech Summit, recently proposed that a consortium of colleges and universities in South Carolina come together to build an open-source AI laboratory like those China is currently racing to construct, at a minimum cost of $100 million. Similar proposals shouldn’t be long popping up in Georgia.
You said “ That captures the enormity of the bet that is being placed on the promise of AI, before it has been clearly established how long it will take for that promise to be realized”
that need for clearly established security and safety is what separates leaders from followers in tech. Just like with Hollywood all the money and power is in SF, where they take risks and set the tone. Over and over.
There is a lot history on this site, but the connection between Nor cal and North GA has largely been forgotten.
today SF “gold mines” of tech have leaped over everyone, everywhere to the point without a gov intervention somehow no place will ever catch up. I heard something about being a top 5 this or they and realty is in the way of those words again.
In San Francisco I was running an experiment with a large language model, but on local hardware each iteration took 12 hours and crashed halfway. I connected to Hyperbolic Labs solutions with rentable H100 clusters and wow, the first three training runs of 1.2 M parameter models now take 45 minutes instead of half a day, and GPU costs dropped from $18/hr to $1.49/hr. Even test inference of 500,000 requests runs without lag. Have you ever tried running models like this on regular machines and realized how insane it is?