Latest / Minds, Bodies, and Terawatts / Tesla Is Building Hundreds of Optimus Robots a Week, Every Hand Assembled by a Person, and the Workers Asked to Teach the Robots Their Moves Said They Knew What the Robots Were For
Transcript
- 0:00[Host] This is Minds, Bodies, and Terawatts, the show where the teachers figured out what the class was for. The tech news site The Information reported on Friday, September twenty-fifth, that Tesla is building several hundred of its Optimus humanoid robots a week, and that factory workers asked to teach those robots pushed back.
- 0:19[Guest] They pushed back on one assignment in particular. Tesla had employees wear suits that record how their bodies move, so the robot can learn the motion. Some of them complained that they knew the robots were designed to eventually replace them.
- 0:34[Host] We covered this in June, when Tesla shut down its Model S and Model X lines to build robots instead. How fast has that line picked up?
- 0:44[Guest] By last month it was turning out about ten times what Tesla managed in the spring.
- 0:49[Host] That sounds like a ramp that's working. Where's the trouble?
- 0:53[Guest] The trouble is in the hands. Each hand and forearm has more than a hundred screws and small parts, and a worker puts them together manually. The equipment on the line can't reliably position parts that small, so more robots come off needing repairs.
- 1:09[Host] Is that a startup hiccup, or is it the product?
- 1:13[Guest] Musk told investors on the last earnings call that this is the hardest product Tesla has ever tried to manufacture at scale. The touch sensors in the hand have been unreliable too.
- 1:25[Host] Say the hands get sorted out. Can the robot work a shift?
- 1:29[Guest] Not by the account of people who know the system. They told The Information it still takes several days to learn even a basic task, and it behaves unpredictably in situations it wasn't trained for. Most of the robots never leave Tesla, where they're used for testing and collecting data.
- 1:47[Host] If I work with my hands, that sounds like good news. The robot isn't close.
- 1:53[Guest] You're right that it isn't close today. But look at what closes the gap. The robot learns from recordings of people doing the work, and Tesla has more than five hundred thousand hours of those, with a plan to double that by the end of the year.
- 2:10[Host] I can't picture that many hours.
- 2:12[Guest] It's roughly two hundred fifty people working full time for a year, recorded the whole way through. That's my arithmetic, not Tesla's, but it gives you the size of the teaching staff.
- 2:24[Host] And some of that staff objected. What did Tesla do?
- 2:27[Guest] It moved the work to separate, dedicated teams. The Information says Tesla hired data collectors who wear camera helmets and motion suits, and opened training hubs in Colorado, Arizona and Florida.
- 2:39[Host] So Tesla solved the objection by finding people who won't raise it?
- 2:43[Guest] Or people for whom it's plainly the job. A hired data collector isn't demonstrating their own spot on the line, and that's a real difference. It's also a way to make a complaint go away without answering it.
- 2:57[Host] What was there to answer? The line workers were on the clock, earning their wage.
- 3:02[Guest] A wage pays for the shift. What those workers were handing over was the skill itself, the thing that made them worth hiring. When a skilled worker trains a successor, the old deal ran both ways: the apprentice carries the craft forward, and the teacher keeps a place in it.
- 3:21[Host] So what's different when the apprentice is a machine the company owns?
- 3:25[Guest] The lesson stays with the owner, and the teacher's part ends when the recording stops. That's the trade the people in the suits were objecting to.
- 3:34[Host] Does this stay inside Tesla's walls?
- 3:37[Guest] Not according to The Information. The plan is to lease the robots, not sell them, to a short list of companies whose factories and warehouses look like Tesla's own. The data from those floors comes back to improve the robot.
- 3:51[Host] So if my warehouse makes the list, the robot shows up to work and to learn.
- 3:57[Guest] And your employer signed the lease, not you. The worry used to be whether anyone could teach these robots fast enough. Tesla's answer is that teaching is a job it can hire for. That complicates the hopeful version, where teaching the machines is the work people move up to.
- 4:15[Host] In Unscarcity, Patrick Deglon calls the missing piece Impact, a record of what you contributed that stays in your name. Tesla keeps the hours of lessons, so whose name goes on them?
- 4:27[Guest] For more on tools that train their replacements and what a teacher is owed, head to unscarcity dot ai. The book is on Amazon, the audiobook is on Spotify. This has been Minds, Bodies, and Terawatts.