In March 2026, Nvidia's CEO stood on stage and declared "the ChatGPT moment for physical AI is here." A few months later, Figure's humanoid robots were running logistics sequences at a BMW plant, receiving spoken instructions, and verbally confirming them before acting. Tesla's Optimus is doing battery cell sorting on the Fremont line. Boston Dynamics' electric Atlas is deployed at Hyundai.

None of this was science fiction eighteen months ago. All of it was science fiction eighty-four years ago — when a 22-year-old Isaac Asimov wrote it.

We're living in the early chapters

In 1942, Asimov introduced the Three Laws of Robotics in a short story called "Runaround," later collected in I, Robot:

1. A robot may not injure a human being, or, through inaction, allow a human being to come to harm.
2. A robot must obey orders given by human beings, except where such orders would conflict with the First Law.
3. A robot must protect its own existence, as long as such protection doesn't conflict with the First or Second Law.

The robots in I, Robot are narrow, task-bound machines — mining robots, factory robots, robots that do one job with hard behavioural limits baked in. That's roughly where we actually are right now: Optimus sorting battery cells, Figure sequencing warehouse logistics, robots doing bounded jobs under close supervision.

Later in the series — The Caves of Steel, The Naked Sun — Asimov introduces R. Daneel Olivaw: a fully autonomous humanoid robot, indistinguishable from a human, operating as an independent partner in a murder investigation. We are nowhere near that yet. But the ingredient that makes Daneel possible — genuine reasoning about ambiguous situations, not just executing pre-programmed tasks — is exactly what the new "Vision-Language-Action" models (like the one running Figure's robots) are the first real attempt at. We're not in Caves of Steel. We're in the opening chapters of I, Robot, and the sequel just got greenlit.

A personal aside

I can't write about this without saying it plainly: this is, in my opinion, the best book series I have ever read — better than Tolkien's Lord of the Rings. That's not just a stubborn personal opinion either. The original Foundation trilogy actually won the Hugo Award for "Best All-Time Series" in 1966 — beating Lord of the Rings for that exact prize.

Read the whole thing — Robots, Empire, and Foundation, which Asimov spent decades quietly stitching into one unified story — and you're covering roughly 20,000 years of future history in a single continuous arc. I still get goosebumps at the ending.

So why not just hardcode the Three Laws?

It's the obvious question. It's also one Asimov himself spent forty years answering — and the answer is: he tried, in fiction, and showed exactly why it doesn't work that simply.

Almost every I, Robot story is built around the laws breaking down at the edges. What counts as "harm" when the definition is ambiguous? What happens when a human splits a harmful instruction across several robots, so no single robot's action clearly violates the law? What happens when "obey orders" and "prevent harm" point in opposite directions and a robot has to weigh them? Asimov didn't write three tidy sentences that solved robot ethics. He wrote three sentences and then spent a career demonstrating every way they fail in contact with a messy, ambiguous real world.

That's the actual lesson for real AI development — and it's one the frontier labs have quietly already absorbed. Nobody serious is trying to hardcode three sentences and call it solved. What's actually happening — Anthropic's "constitutional AI," OpenAI's published model specifications — is a more honest descendant of Asimov's idea: not three universal laws, but an extensive, constantly revised set of principles, because "don't allow a human to come to harm" turns out to require the same thing Asimov's robots needed and didn't have: judgment, not just rules.

Which, if you've read anything else I write, is a very familiar conclusion wearing a different costume.

Thanks for reading!

If you liked it, please share and tell your friends to subscribe.

And if you want to browse the archive go here.

/Jonas

​
​