Why Simulation Fails for Dexterous Robots | Brian Potter
MTS
October 10, 2026
Brian Potter of Construction Physics and the Institute for Progress discusses the history and failure of plasma displays, why Vision-Language-Action (VLA) models and cheap hardware like UMI are changing robotics, and why prefab housing remains exceptionally difficult to scale without...
Brian Potter
The simulator is only good at certain things. It's very good to train a robot to move around. The reason that you can see robots good at walking and good at doing kung fu and stuff like that is because that sort of motion is relatively easy to train in simulation.
The reason that there's still crap at manipulating objects is because simulation doesn't work really as well for dexterous manipulation and having to model all these surface features and stuff like that. We're just really not as good at that.
SPEAKER_2
We are back with Brian Potter, who is the author of the excellent Construction Physics on Substack and also affiliated with Institute for Progress, Senior Infrastructure Fellow at the Institute for Progress. We love IFP, Platinum tier friends of the show. Brian, welcome.
Brian Potter
Thanks, it's nice to be here.
SPEAKER_2
Yeah, people say this to us a lot. They say long time listener, first time caller. Now it's my turn. I'm a long time reader, first time caller.
Brian Potter
Yeah.
SPEAKER_2
So tell us about your recent article, The Rise and Fall of the Plasma Screen. I remember I had a Plasma TV as a kid and now it's just totally washed.
Brian Potter
Yeah, it's super interesting because the technology, it begins way earlier than you think that it would.
You think that it's like this thing that comes along in the 2000s as a new way to build a TV screen. Actually starts in the 1960s. It was a way to build computer screens when it was like computer memory was like super, super expensive and it was like even just like drawing the data on the screen took like way more memory than like a computer could really justify. So the plasma screen was invented as this way to like require less computer memory because it has this like memory effect where only the stuff that changes needs to be redrawn other than like a normal computer screen that like everything needs to be drawn every single frame.
SPEAKER_2
So it's kind of like an OLED?
Brian Potter
Yeah, and so that's like the specific version reason that they invented this thing. And then like by the time they invent it, this reason has become pointless because memory has gotten so cheap that they don't need it for that reason anymore. So they kind of try to find other reasons for using it over like this, you know, the 70s and the 80s and it kind of keeps coming along. And then finally, and like, yeah, in like the 2000s, it like has its moments.
It's good enough. It's going to be for these great TVs. And, you know, here's the technology. But then LCD screen like quickly like surpasses it in terms of like cost and stuff like that. And then so it has this like really narrow moment of the sun. And now they basically don't make them at all. The technology is like completely dead.
SPEAKER_2
What is the difference between specifically between Plasma and LCD and OLED?
Brian Potter
Oh, gosh. So Plasma is like it's kind of like a big giant array of like a billion neon lights. You have this like little sort of glass box or cell or whatever. And you run an electrical with it as a gas in it, and like an electrical current, and you run some electrical current through it, and it like glows. And you put different mixes of gas, and they glow different colors. You make a bunch of these like really, really tiny things, and you can make each one its own little pixel, and you can make a screen out of that. An LED or an LCD screen is like, it works on like totally different principle. You have like what's called liquid crystals, which are these like liquid, like it's exactly what the name sounds like. It is a liquid that behaves like a crystal, and you can shine light through it. And by like running electric current through this liquid crystal, it'll like twist in a little way. And you can get, if you can make it selectively twist, and let light selectively pass through it. And again, you make an array of a huge number of these things, and you can let the light through selectively, and each one is a pixel, and you can make a screen out of that way.
OLED, LED, I think is basically, this is I'm less familiar with, it's like a light-emitting diode, just like a semiconductor basically thing. And you have a huge array of these little semiconductors, and each one lights up independently, and generate a screen that way. So each one uses like a fundamental.
Read the other 18 minutes of this episode here.
$15 gets you this episode and 99 more. Credits never expire. After checkout, one link brings you back here to read it.
Not what you expected? Full refund within 14 days.
Thousands of transcripts fetched by people building searchable podcast archives
Fetch transcripts with code
The same credits work from a script or an AI agent. The demo key returns a sample episode in full, no card needed:
curl -H "x-api-key: pt_demo" \
https://spoken.md/transcripts/1000651996090Markdown with the speakers named, for your notes, your knowledge base, or anything that makes HTTP calls.
From $0.10 per transcript. No subscription. Credits never expire. Prices exclude VAT, added at checkout for EU customers. Not what you expected? Email us within 14 days with 20 or fewer credits used and we refund the pack in full.
Using your own key:
curl -H "x-api-key: YOUR_KEY" \
https://spoken.md/transcripts/1000794666157