Hook
More breakout videos from this creator.
Here's where I think jobs are changing right now, in ways we won't see in the data until it's super obvious. First, many to one. First, many to one. So teams are shrinking through attrition and agents are producing the work and one person approves all of it. The job exists, it's just a different kind of job. Second, self-service. People are doing for themselves what they used to pay for professionals for. Legal documents, through to plumbing. Just yesterday someone broke a switch in our house and Dave used Gemini to help him replace it. So the professional doesn't lose their job, demand just stopped showing up. And the person doing it themselves doesn't know what they don't know. Third, the one-person company, one human plus agents doing what used to take an entire firm. So in the data this, like good news because business formation is up. Review, no compliance, no colleague to say that's wrong. One thing is in common here, each change removes the layer that catches a mistake. So the reviewer, the professional, the institution. We're in let it rip mode. So all three produce the same result, much more output, less checking. AI errors don't look real errors, they're well formatted, wrong used to look wrong. So I think we're up a huge stock of problems. Now who fixes all this? Fixing is a different from making. Something's broken when you have to work out the thing was supposed to do and you have to understand why it broke, which means having that causal model of the system, not just a picture of causal models. I've talked about this before. AI works from representations, it's pattern recognition not causal model. That's different. There's a deeper asymmetry as well. Things go right in a few ways and they go wrong in endless ways. So making means hitting that fixed target, fixing is a different open space all together. People learn to fix by doing the work, making things, getting them corrected by someone who's seen it before. All three of these cut that stage out. So mistakes are multiplying while the people who can fix them are getting scarcer. Can AI fix its own mistakes? Sometimes, if your break is a common one, a bug, lots of people have seen it, AI will fix it fast. AI fixes documented problems, but the undocumented ones are yours. Here's my claim. We're automating the making faster than we're keeping the fixing — and the gap will be the most expensive skill of the 2030s. Notice where you are, is there a review above you that you see getting thinner, less checking? Are your clients starting to self-serve? Are you working alone with agents and no one's checking? If any of this is true, your situation right now. And do this. Use AI to make things, that's what it's good at. But when breaks, work it out on your own. You have your own theory of why it broke before actually ask AI. That's how you build the skill that others are losing. I'm Helen from Artificiality. Don't forget to check our summit in Bend, Oregon, 22 to 24 October, two and a half days of figuring out a better future with AI, with our community of artists, entrepreneurs, builders and educators. All at the frontier of what it means to be human.