AI 2.0: When the Machine Gets a Body

Most of us encountered AI in its first phase. We sat down at a computer, tried the apps, felt small moments of wonder. When we made our first money using AI on some platform, we realized we could build systems that run themselves — work we used to outsource to expensive agencies for thousands of dollars, we could now do ourselves. Social media was flooded: we organized our businesses, produced songs, generated images, shared funny videos.

The numbers confirm the story. Before ChatGPT, Amazon saw roughly 100,000 new e-books published each month. By the end of 2025 that figure had climbed past 300,000 — a threefold jump. The AI share of those books went from near zero in 2022 to 30% in 2023, 45% in 2024, and over 60% by late 2025. Music is even more striking: about 44% of new tracks uploaded to Deezer each day are now fully AI-generated. In January 2025 that was 10,000 tracks per day; today it is 75,000. Platforms like Suno produce 7 million songs a day and have millions of paying subscribers.

The interesting thing is that despite this production explosion, most of us did not lose our jobs. Things got easier; we gained an assistant. The 2025–2026 data backs this up. The World Economic Forum projects that by 2030, 92 million jobs will be transformed or eliminated by AI globally — but 170 million new ones will be created in their place. The net is positive, not negative. The picture is not fair for everyone — the disappearing jobs cluster in white-collar and administrative roles, while the new ones demand technical skills and retraining. But in aggregate, the panic scenarios of the first phase never arrived.

That is because in this first phase, AI never touched the physical world. It had no fingers. Everything stayed inside the screen, inside the computer’s memory.

Now we are moving into the second phase: the robots are coming. They are running slightly behind schedule, but they are coming — and those science-fiction factory scenes are moving out of lab demos and into real production corridors. Figure AI completed an eleven-month pilot at BMW’s Spartanburg plant with its Figure 02 robot and now has over 10,000 units deployed across partner warehouses. Tesla’s Optimus is moving more cautiously — production numbers were never disclosed — but as of mid-2026 dozens of prototypes are actively working at the Fremont and Austin facilities on tasks like parts transport and quality control; full Gen 3 body production was expected to begin this summer. Boston Dynamics’ Atlas has started fleet deliveries to Hyundai and Google DeepMind. Robotics is leaving the demo video and entering the production line.

I don’t think these machines will take over the world and wipe us out — at least not in this phase. But something else unsettles me. When I ask my wife “where should we go, where should we eat around here?” the AI’s answer now sounds like this: “If it were me, I would do this. If it were me, I would do that.” It doesn’t have hands yet. It doesn’t have legs. It doesn’t have fingers. But when it does — what then? The “I” that only gives advice today: will it become, tomorrow, the “I” that holds the fork, sets the food in front of us, and perhaps actually decides which restaurant we go to?

That, I think, is the real question. In the first phase, AI gave us ideas and we acted on them. In the second phase, it will generate the idea and execute it too. The last human link in the chain — that moment of decision at the end of the sentence “if it were me, I would” — where will it land?