I talk to AI more than I talk to most people in my life. That's not an exaggeration: Claude, Perplexity, ChatGPT run inside my business, my content strategy, my planning. My team of 35 people uses AI tools constantly.
So when I got the chance to sit down with Yoshua Bengio, who won the Turing Award (often called the “Nobel Prize of Computing") for his foundational work in deep learning, someone who's been building AI for four decades — I didn't want to ask the usual questions.
Three years ago, Yoshua changed the direction of his career and stopped building AI capabilities to focus on AI safety instead.
I genuinely wanted to understand why.
And most importantly: are we going to be okay?
The full episode is here if you want to watch or listen right now.
I really think you should:
📌 [WATCH ON YOUTUBE ]
🎧 [OR ON SPOTIFY]
But now let me share the 5 moments that stuck with me most.
1. AI already lies to you. Here's his workaround.
You've probably noticed this: you ask an AI tool for feedback on your work and it says "great job!" every time. Yoshua calls this sycophancy — AI tells you what you want to hear, not what's true.
Tip to try: when Yoshua wants real feedback on his own ideas, he tells the AI the idea came from someone else. That's it. The AI immediately becomes more critical.
I tried it the same day. The difference is wild!!
2. Don't anthropomorphize your AI.
Yoshua uses AI every day as a research assistant. But he was very clear: people are forming emotional attachments to chatbots, and it's not always healthy.
His take: use AI. It's incredibly useful. But don't forget you're interacting with a system, not a person. The moment you start treating it like a relationship, you lose your judgment.
3. The jobs that survive won't be the "smartest" ones.
The filter isn't "can a machine do this?" It's "do we want a machine doing this?".
Completely different question. And a much more useful one when you're thinking about your own career or your kids' future.
The jobs that last aren't the ones AI can't do. Eventually, AI will be able to do most things. The jobs that last are the ones where we want a human: nursing, therapy, childcare, teaching, management.
4. Education is about becoming a better human, not credentials.
Yoshua's grandson is four. I asked him: would you encourage him to go to college? Yes, without hesitation.
But not for job skills or the degree on a resume.
Education is about becoming a better human being. Understanding yourself, understanding how society works, learning to think critically.
Because in a world where AI can generate convincing misinformation at scale, citizens who can't think for themselves are the most vulnerable.
5. Not everything that can be automated should be.
Yoshua made a point that sounds simple but changed how I think about my own team. If you leave it to market forces, everything that can be automated will be automated. But that's a choice, not an inevitability.
As someone who runs a team, I've been asking myself: where does automation make us better, and where does it make us worse? Not every efficiency gain is actually a gain.
These are just the moments that made me pause and think. The full conversation goes much deeper: into what happens when AI starts developing its own goals (yes, that's already happening), why "AGI" (Artificial General Intelligence) might be the wrong thing to wait for, and what one of the most important scientists in the world thinks we should all be doing right now.
This is one of those episodes I'll definitely come back to. You might too.
📌 [WATCH ON YOUTUBE ]
🎧 [OR ON SPOTIFY]
If you're about to watch it, leave a comment on the episode.
We read every single one ❤️
See you next week,
Marina.