The job apocalypse. The idea that AI is about to take over most of our jobs.
"That's just not going to happen. We're just not good enough to make AI do everything a human does."
That's Andrew Ng, minutes into our conversation. I was happy to have him on the podcast last week. Not only because he's one of the biggest names in AI, but also the positive one.
Before we started. He came in expecting questions about building with AI. I brought a list of everything people are afraid of.
Andrew Ng co-founded Coursera, started Google Brain, and teaches AI at Stanford. 4.8 million people have taken his machine learning course since 2012. Last month Coursera put $100 million into his new company, built to teach one-to-one instead of one-to-many.
Every big economic shift brings huge opportunities. It also brings headlines like these:
AI is taking our jobs.
AI data centers are draining the water supply.
AI is the new nuclear weapon.
"Unfortunately, fear mongering works." And it is making people give up on AI.
Andrew said it was driven by "a handful of leading AI companies," protecting the billions they spent on training their giant models. If they succeed, everyone pays what he calls "a high toll" for AI. He was very open about it. I'll leave that part for you to watch.
I asked him to sort every AI fear into real, overhyped, and science fiction. Jobs went straight into the second pile.
🔴 A job is not one thing
Think of your job as a stack of tasks. AI is good at some of them.
"Maybe AI could do 30-40% of many jobs."
That figure comes from economists breaking real jobs down task by task, including Erik Brynjolfsson at Stanford and Andrew McAfee at MIT.
The tasks AI takes get cheap. The 60% it can't do becomes the scarce half, and scarce means expensive.
So who actually loses?
"People that use AI will replace people that don't use AI."
The engineers in trouble are the ones still writing code the way they wrote it in 2022. He says openings in that field are up right now, and every good engineer he knows is busier than ever.
Front-end and back-end developers became full stack. The marketing coordinator runs the full cycle. The recruiter runs the whole pipeline.
"Marketers, recruiters, HR professionals, operations specialists. If they learn to build with AI, they really just do much more, whatever the job role is."
Learning AI is essential, and on its own it doesn't get you there. You need the AI skills and the depth in your own craft.
And that craft is exactly the 60% AI can't get to, because almost none of it was ever written down. You watched the customer's face change when you named the price. Your manager told you what he actually cares about. That never reaches the model. Andrew calls it a context advantage, and nobody closes that gap in a few years.
🔴 How he hires
Every marketer on his team writes code. In interviews he asks candidates what they have built. One shipped a desktop app that crawls the web for related work before he writes an article. That is a marketer.
His advice, in his words: "learn AI, apply it, and go build something."
"Build something" may sound intimidating. Andrew gave a few examples, and not one of them is a startup.
🔴 Also in this episode
What are AI models terrible at? His word, and it's about learning.
Where the big opportunities are in 2026 for people who want to build.
How many years until AI can run a whole job by itself? He gave me a number.
Marina 💜
P.S. Know someone quietly worried about their job? Send them this episode, and tell them the 40% is the good news.
