Home › Hard problems › Physical AI, robots and autonomy

Physical AI, robots and autonomy

Proving an autonomous system is safe

Solo: possibleOne person can test a narrow slice of it.

every situation the robot could meetthe scenarios you testedred: rare failuresnobody testedDrive it?far too many milesto prove it safeTest scenarios?nobody agrees whichones are enough
You cannot drive enough miles to prove a machine safe, and nobody agrees which test scenarios cover the rare failures that matter.

Why it is still open

Driving enough miles to prove safety is impossible, so you need scenario testing, and nobody agrees which scenarios are enough.

An experiment one person can run in 8 weeks

Generate scenarios for one narrow case (a warehouse robot crossing), run them in simulation, and measure how many rare failures random tests find against targeted tests.

Verdict: Solo: possible

The same "faced world" idea as AgentTwin, applied to machines.

The solo test behind the verdict

  1. Can it be measured? There is a number that shows you did better, such as an error rate, a catch rate or a cost per answer.
  2. Is it cheap to try? It runs on a laptop, free open models and open data, or under about ₹50,000 of rented GPU time.
  3. Is there a public yardstick? A benchmark or dataset others use, so your result can be compared and believed.
  4. Would a small win get noticed? A lab, a funder or a customer would care about a 10% improvement, not only a full solution.

Four yeses means Solo: strong. Two or three means Solo: possible. Fewer means Follow, don't build.

PreviousThe gap between simulation and realityNextFlying and moving without GPS

All 79 hard problems · Working on this, or know of a result we missed? basant@cleandataengineering.ai