Years of job-loss warnings, billion-dollar announcements, and god-like claims about technology that often can’t reliably count to ten have quietly built up a head of public resentment. Tech critic Ed Zitron has been making this case for a while. In a recent episode of The Tech Report, he laid out exactly why the industry’s habit of overpromising and under-delivering is starting to catch up with it.
The AI Industry Has a People Problem
Ed Zitron has been saying this stuff for a while now, and it’s worth paying attention to. In a recent episode of The Tech Report, he sat down with host Isaac Pound and made a case that the AI industry isn’t just failing technically or financially – it’s actively alienating the people it claims to be helping.
Worth watching the full thing.
Here’s what stood out.
The Rhetoric Is the Problem
The AI industry has a habit of doing two things at once. First, it tells you that AI is going to take your job – not in a vague, distant future sort of way, but soon, specifically, and at scale. Sam Altman and Dario Amodei have both made public statements suggesting that vast swathes of white-collar work are on the way out. Fifty percent of jobs. “Real work” being automated away.
Second, it asks you to be grateful for it.
Zitron’s point is blunt: when ordinary people are struggling with mortgages, credit, healthcare costs, and job insecurity, and the people threatening their livelihoods are simultaneously raising hundreds of billions and giving each other grants, you don’t get to act surprised when people start resenting you.
That’s not a fringe reaction. That’s human nature.
The Fear Machine
There’s a specific kind of AI hype that Zitron takes apart here, and it’s the “AI Doomerism” angle. The idea that these models are potentially god-like, maybe dangerous, possibly the last invention humanity ever needs to make.
Anthropic published material suggesting one of its own models behaved in ways it found concerning – essentially framing it as almost too powerful to release comfortably. OpenAI has done similar things.
Zitron’s view is that this isn’t caution. It’s marketing.
If you say your product is terrifying, you’re also saying it’s extraordinarily powerful. And if it’s that powerful, why would you sell it to anyone except trusted, paying enterprise clients? It keeps the best access locked behind the biggest cheques.
The problem is, as Zitron points out, this language lands badly on people who are already struggling. Framing code as a sentient threat that might end human work isn’t a neutral communications decision. It does real psychological damage to real people.
Stargate and the Reality Gap
You probably heard about Stargate – the $500 billion data centre project announced with considerable fanfare earlier this year. The kind of announcement that ends up on the news, gets politicians excited, and makes it sound like the future is already being poured into concrete.
Zitron calls it a “reality distortion field.” And the evidence supports that reading.
Projects in Norway and the UK have been cancelled or stalled. The infrastructure hasn’t materialised anywhere near the scale the announcements implied. OpenAI doesn’t have the construction experience or the financial track record to deliver what was described. The Stargate announcement was a joint venture across multiple partners – but the headlines made it sound like a done deal.
Zitron’s suggestion is that these announcements serve a purpose that has nothing to do with building anything. They keep stock prices up. They help partners raise money against projects that exist on slides rather than sites.
The Subscription Model Is Broken
This is the part that hits closest to home for anyone using AI tools day to day.
The £20-a-month subscription you signed up for months ago is not the same product you’re using now. Rate limits have been cut. Models have been quietly changed. Downtime has increased. The service has got materially worse while the price has stayed the same.
Zitron describes the current pricing as heavily subsidised – companies selling access at a loss to build a user base, with no sustainable path to profitability at those rates. As capacity tightens, they quietly make the product worse rather than put prices up honestly.
His prediction is that flat-rate subscriptions for AI tools disappear eventually, replaced by token-based pricing that reflects actual compute costs. When that happens, most individuals and small businesses will have to decide whether these tools are actually worth what they really cost.
That’s going to be a reckoning for a lot of people who’ve built processes around these subscriptions.
What Zitron Actually Wants
It’s worth being clear that Zitron isn’t saying AI is useless or that the technology doesn’t exist. His argument is more specific than that.
Large language models are normal technology. Complex, yes. Useful in places, yes. But not magical, not sentient, not the last invention humanity will ever need. The industry’s habit of anthropomorphising these tools – giving them names, personalities, emotional language, the appearance of consciousness – is a deliberate choice that serves the companies, not the users.
His call is for regulators to step in, for the hype to stop, and for AI companies to talk about their products like the software tools they actually are.
That’s probably too much to hope for in the short term. But the cracks are showing, and this conversation is part of what’s pulling them apart.
Zitron’s a good talker and there’s a lot more in the episode than I’ve pulled out here.
What People Ask About AI
Is the AI industry really in trouble financially?
The current business model is under serious strain. Most AI subscriptions are sold at a loss, heavily subsidised to build user bases. As compute costs bite, companies are quietly reducing service quality rather than putting prices up. The flat-rate subscription model is unlikely to survive long-term.
Why do people dislike the AI industry so much?
A lot of it comes down to tone. Industry leaders have repeatedly suggested that AI will automate huge numbers of white-collar jobs, while simultaneously raising billions and celebrating their own success. When people are already struggling with the cost of living, that combination breeds real resentment.
What was wrong with the Stargate announcement?
The $500 billion Stargate project was announced as though construction was imminent, but the physical reality has been very different. Projects in Norway and the UK stalled or were cancelled quietly. Critics argue the announcements were designed to boost stock prices and help partners raise money, rather than reflect anything actually being built.
Are AI models actually getting worse?
Many users have noticed degraded performance, tighter rate limits, and more downtime from the same subscriptions they signed up for months ago. This lines up with what Zitron describes – companies quietly reducing service quality as they run up against the real cost of compute at scale.
Is AI actually dangerous or is that just hype?
Zitron’s position is that the “AI is dangerous” narrative is largely a marketing tool. Framing a model as potentially too powerful to release safely implies extraordinary capability and justifies restricting access to paying enterprise clients. Large language models are sophisticated software, not sentient systems.