Bill Gates Staked His Reputation on This AI Warning. Abundance for Whom?

Bill Gates sat down with Anderson Cooper on CNN to talk about the essay he just published on artificial intelligence. In it he argues that the AI industry doesn't understand the real risks of what it's building, and that dealing with them should be the world's top priority. CNN is putting the interview out in parts, and this first part is about AI. I've read a rundown of the rest — Cooper also asks him about the Epstein case, about taxes, and about what happens to ordinary people.

Gates co-founded Microsoft and was once the richest man in the world. I've been following him since about 2015.

The three risks he names

He told Cooper the models have gotten more powerful over the past year — faster than he expected — and that three kinds of risk come with them.

First, people who mean harm — fraud, cyberattacks and, at worst, bioterrorism. He says we have to watch for somebody using a model to build a dangerous virus, and stop it before it happens.

Second, the labor market. He admits the job losses haven't started yet, but as the models get more reliable, you can hand them more and more work that has clear rules — and people are going to have a hard time adjusting.

Third, what he calls the psychosocial risk. A kid can grow up with AI for his only friends, stop talking to real people, and lose what he'd get from them — what he'd learn, and the fun of it. All three of these worry him, and he wants to slow them down or get rid of them altogether.

There's almost nothing to judge these models by, he says, and almost nothing in place to hold the damage down. And nobody is having the public argument about the trade-offs. That's the part that shocks him.

The numbers

The numbers came from Cooper. He interviewed Dario Amodei in May of last year, and Amodei figured AI could cut 50% of all entry-level white-collar jobs in the next one to five years, and push unemployment to 10 to 20% within five years.

I've heard a bigger number than that. I put out a video with Professor Yampolsky, who is dead against the AI rollout, and he puts it at close to 90% of professions gone in three to five years. That video got a lot of views, and this is his field, so it's worth watching.

I agree with Gates on most of it. And every time, I come back to the same question: what are all those people going to do when a machine takes their job? That's exactly what he's warning about.

The lamplighter argument

Cooper hit him with the usual objection: every big technology shift costs jobs, the electric light bulb put lamplighters out of work, and we got through it. Then he read Gates' own answer back to him out of the essay — those earlier shifts played out over a long stretch, sometimes over several generations, and this one moves much faster and hits every industry at once. Gates picked it up from there: white-collar work across the whole economy first, then blue-collar work when humanoid robots show up, because for the first time a robot can think roughly the way a person does.

The comparison with the past fools people, he says. Back then we were rich enough to pay for extra jobs — ride designers at Disneyland, more psychiatrists. Now a machine does a job with clear rules better than a person, and it runs 24 hours a day. That makes this a much bigger problem.

Here's the part I'd underline. The bar is very high for anyone who claims this time is different, and he's staking his own reputation on saying it anyway: for what it does to the job market, this is nothing like any technology he's ever come across. When a man at that level does that, I listen.

Abundance for whom?

Cooper asked what he'd say to parents with small children. Gates talked about spending a few days with his two granddaughters, three and one, and said it makes you wonder what education is even going to look like. He splits it into two stretches of time: the next 20 years, and everything after that. By the 20-year mark, he says, people are going to have a real problem working out what to do with themselves — and at the same time there's going to be abundance. What he wrote is mostly about getting through that rough stretch, while everybody gets used to machines that can do things nothing has ever been able to do.

He said abundance. Abundance for whom? That's the real question.

He never said who gets it and who doesn't. Cooper moved on to the next question.

He's not saying it's hopeless

Gates isn't against the technology. He told Cooper it genuinely helps — looking something up, weighing a decision, going through data — and that he's glad he has it.

Cooper read his own line back to him — AI as either the greatest equalizer ever invented or the worst source of injustice — and asked which one it's going to be. Gates says we can still solve all of it if we bring in people from right across society — the ones using these tools and the experts. That means doing the hard analysis and setting standards a model has to pass. He gives one example: checking whether somebody could use a model to build tools for bioterrorism. He compares the scale to nuclear power and nuclear weapons, except this is a thousand times bigger, and the good and the bad come in the same package.

He also warns about two things people tell themselves. Don't assume the models will hit a ceiling and stop getting better, because mostly they don't. And don't assume that blocking a data center slows anything down, because somebody will just build it somewhere else. On politics, he says the people in the industry are mostly arguing with each other. Getting this right takes academics, companies and politicians from both parties who'll actually listen. Otherwise it ends up all or nothing, with no real limits and nothing in place to keep it safe.

There's a lot more of this interview still to come. I'll take the next parts as they land.

Gates is a very smart man, and I take what he says seriously. And when he says things aren't going back to the way they were — I have no doubt about that at all.

So what do you make of what Bill Gates is predicting?

Originally published as a video on my YouTube channel on 28 August 2026. Watch on YouTube