Roman Yampolskiy: Most Replayed Moment: AI Safety Expert Predicts The Next 20 Years! Will It Really Take All Jobs?

The segment opens with the host asking about the pushback the guest must hear. The host says people feel "a huge amount of mental discomfort" when told that their job, their career, the thing they got a degree in, or "the thing they invested $100,000 into" will be taken away. For some, the host says, the result is cognitive dissonance, expressed in familiar lines: AI can't be creative, it will never be interested in my job, I'll be fine.

The guest says they test this by asking people in different occupations. An Uber driver, asked about self-driving cars, replies that no one knows the streets of New York or can navigate like they do. Professors say nobody can lecture like them. The guest calls this attitude "ridiculous" and argues that the question for drivers is no longer whether automation is possible: "we already have self-driving cars replacing drivers." In the guest's words, the real question is "how soon before you're fired."

The host backs this up with personal experience. In LA the day before, the host's car drove itself: the host typed in a destination and did not touch the steering wheel or brake pedals, "even if it's an hour long drive", with no intervention. The host notes one exception: they still end up parking the car themselves. The host also mentions Waymo, which is ordered by phone and arrives with no driver. On that basis the host says job loss for drivers is "potentially a matter of time". The host adds that driving may be the biggest occupation in the world, but qualifies it with "if I'm correct". The conversation offers no evidence for that figure.

Asked what drivers should retrain in, and on what timeline, the guest describes a "paradigm shift". The old advice assumed one job would be automated and workers could move to another. The guest's claim is broader:

if I'm telling you that all jobs will be automated, then there is no plan B, you cannot retrain.

The evidence offered is the recent history of career advice in computing. Two years ago, the guest says, people were told to learn to code, including artists who were told they could not make money. Then AI turned out to be able to code and kept getting better. The next advice was to become a prompt engineer, even to get a degree in it. But according to the guest, AI is "way better at designing prompts for AI than any human, so that's gone." Today's hardest skill, by the guest's account, is designing AI agents for practical applications, and the guest predicts that this too will be gone within a year or two. (The transcript renders the word as "gorgeous", apparently a transcription error.)

So the guest does not tell any occupation to learn a specific new skill. Instead they reframe the question as a collective one: if everyone loses their job, what do we do financially, who pays for us, what gives life meaning, and what do people do with an extra 60 to 80 hours a week?

When the host asks what is "around that corner", the guest says they have thought about it "a little bit". The guest treats the economics as the simpler part. Lots of free labor would mean lots of free wealth and abundance, so things that are unaffordable now would become "dirt cheap" and everyone's basic needs could be met. The guest attributes a more optimistic version to others: "some people say" everyone could have a very good existence, beyond basic needs. The guest gives no mechanism for how that wealth would be distributed.

The hard problem is what do you do with all that free time?

The guest's evidence here is an analogy to retirement. For many people their job is what gives life meaning, and the guest says we already see people feel lost when they retire, including early retirees. The guest grants that many people who hate their jobs would be very happy not working. The open questions are social: what happens when people are "chilling all day", and how that affects crime rates, pregnancy rates and other issues. The guest raises these as questions, not as predictions with data behind them. The guest ends with a claim about how unprepared governments are:

governments don't have programs prepared to deal with 99% unemployment.