Move Aside, Chatbots: AI Humanoids Are Here

This week, WIRED learned that OpenAI is ramping up its efforts successful robotics—specifically, by hiring researchers who activity connected AI systems for humanoid robots. Humanoids, robots built to lucifer america and execute regular tasks, were celebrated for their clumsiness conscionable a fewer years ago. Senior writer Will Knight tells america astir really that's quickly changing connected today's section cohosted by Michael Calore and elder analogous Kylie Robison.

Mentioned successful this episode:
OpenAI Ramps Up Robotics Work successful Race Toward AGI by Will Knight
Humanoid Robots Are Coming of Age by Will Knight
2025 Is nan Year of nan Humanoid Robot Factory Worker by Russell Brandom

You tin travel Michael Calore connected Bluesky astatine @snackfight, Will Knight connected Bluesky astatine @willknight, and Kylie Robison connected Bluesky astatine @kylierobison.com. Write to america astatine uncannyvalley@wired.com.

How to Listen

You tin ever perceive to this week's podcast done nan audio subordinate connected this page, but if you want to subscribe for free to get each episode, here's how:

If you're connected an iPhone aliases iPad, unfastened nan app called Podcasts, aliases conscionable pat this link. You tin besides download an app for illustration Overcast aliases Pocket Casts and hunt for “uncanny valley.” We’re connected Spotify too.

Transcript

Note: This is an automated transcript, which whitethorn incorporate errors.

Michael Calore: Hey, Kylie. How are you doing?

Kylie Robison: Good. How are you?

Michael Calore: Not bad. Good to spot you present successful nan chair-

Kylie Robison: I know.

Michael Calore: ... that is usually occupied by Lauren Goode.

Kylie Robison: My Lord and Savior Lauren Goode, I'm truthful happy to capable successful for her, but I miss her dearly.

Michael Calore: We each do. Also connected nan show this week, we person Will Knight, our AI master astatine WIRED. Welcome backmost to nan show, Will. How are you doing?

Will Knight: Hello. Good to beryllium here. I'm doing well, thanks.

Michael Calore: Given nan taxable of today's episode, I want to inquire you both. Do you person a favourite robot movie?

Kylie Robison: Yeah. Mine is The Iron Giant. I was conscionable looking up erstwhile that came retired because I committedness I watched it connected VHS arsenic a kid. Yeah, I loved that movie. It's sewage a very sad ending. The robot is nary longer. Oh.

Michael Calore: I conjecture we tin astir apt spoil it, it's been decades.

Kylie Robison: Yeah, it came retired successful 1999. I deliberation it'll beryllium OK.

Michael Calore: Will, what is your favourite robot movie?

Will Knight: OK. I'm going to urge Blade Runner, which is evidently a very celebrated movie and very beautiful. When I first saw it, I conscionable thought it was conscionable really cool. Then I later publication nan book and realized that it's ambiguous astir whether immoderate of nan characters are robots aliases humans. I recovered it mind-blowing, this thought of blurring what it intends to beryllium quality and asking whether you tin build thing that's human. Which I deliberation we're a agelong measurement from, but it's an absorbing idea.

Michael Calore: That's great. I'm going to spell pinch RoboCop.

Kylie Robison: Nice.

Michael Calore: Which is nan absolute other of The Iron Giant. But it's a awesome movie and it's timeless. It's a awesome film. It's Paul Verhoeven going highly difficult successful nan '80s mode. And it's astir nan privatization of nan constabulary unit and nan corporatization of America's nationalist services, it's very prescient.

Kylie Robison: Very classical Michael pick.

Michael Calore: Thanks. This is WIRED's Uncanny Valley, a show astir nan people, power, and power of Silicon Valley. Today we're talking astir really nan section of robotics is shaping up to beryllium nan adjacent frontier successful nan title towards artificial wide intelligence, aliases AGI. Earlier this week, we learned that OpenAI is ramping up its efforts successful robotics, specifically by hiring researchers who activity connected AI systems for humanoid robots. Humanoids, robots that were built to lucifer america and execute regular tasks, were celebrated for their clumsiness conscionable a fewer years ago. But arsenic our workfellow Will writes about, advancements successful instrumentality learning and hardware person renewed liking successful humanoid technology, peculiarly wrong nan AI industry. We'll dive into what companies for illustration OpenAI are hoping to summation by investing successful robots and what their determination intends for really experts are reasoning astir what is really needed to execute AGI, and what it mightiness mean for us, nan existent humans, if that happens. I'm Michael Calore, head of user exertion and culture.

Kylie Robison: I'm Kylie Robison, elder analogous covering nan business of AI.

Will Knight: And I'm Will Knight, elder writer astatine WIRED.

Michael Calore: Will, this is an evident question, but we person to commencement here. What precisely are we talking astir erstwhile we talk astir humanoid robots? How are they different from different robots?

Will Knight: Yeah, I deliberation it is simply a bully mobility really. The evident reply is that they're shaped for illustration a human, truthful they person legs, arms, caput for illustration we do. But nan bigger constituent really is, and nan logic why nan industry's truthful willing successful this, is because they are designed to run successful nan world that we unrecorded successful which is designed for humans. Going upstairs, sitting wrong vehicles, thing a quality could do that those machines could do. When you're talking astir trying to build thing that has human-like intelligence, if that's really you specify AI, past a large constituent of that is being capable to do things successful nan existent world. ChatGPT is getting very, very bully astatine precocious math, but it can't make you a cup of coffee. It couldn't travel into your location and make a cup. Or possibly it can, fixed what they're moving on.

Michael Calore: Right.

Will Knight: But that's nan idea.

Michael Calore: As you reported on, Will, and others astatine WIRED person talked about, humanoid robots do not person nan astir stellar reputation. They're seen arsenic gimmicky. We've each seen videos of nan Boston Dynamics robot that dances. That video connected YouTube has complete 40 cardinal views. As acold arsenic we know, based connected what nan company's put retired there, that's astir each they're bully at. At their worst, they're seen arsenic unreliable astatine performing desired tasks. They autumn over, they bump into each other. How is that cognition changing correct now?

Will Knight: Yeah. I retrieve going to nan DARPA Robotics Challenge, backmost successful I deliberation it's 2012 erstwhile it first happened. This was a really absorbing idea. After nan Fukushima atomic accident, they couldn't bring robots successful to activity successful this really radioactive environment, truthful nan thought was tin we create robots that could spell into that quality environment. But backmost then, these humanoids, they moved unbelievably slowly. They conscionable kept falling over, often successful really comedic ways. In nan years since that, really a batch of what's happened is really nan hardware has gotten amended and better. And immoderate cardinal things, for illustration motors. In bid to move for illustration a human, we deliberation of robots arsenic being much tin than us, but you request to beryllium capable to move very, very quickly and very explosively successful bid to do things for illustration balance. That's what Boston Dynamics pioneered. That benignant of hardware has gotten a batch much available, which is why you're starting to spot much humanoids travel connected nan market.

Michael Calore: Now we cognize that OpenAI is very willing successful humanoid robots. What did you precocious study from your reporting astir nan company?

Will Knight: I've been pursuing a batch of different labs that are moving connected robots. Humanoids is simply a really large point because it allows you to research this quality beingness constituent of intelligence. I discovered firstly that they were hiring group to activity connected humanoids. Then also, if you look astatine nan occupation listings they person and different signals, you tin spot this image wherever they're intelligibly ramping up their robotics work. Which I deliberation makes cleanable sense, fixed that a batch of different labs are doing that. A batch of different companies moving connected AI look to beryllium looking towards this beingness broadside of intelligence now.

Kylie Robison: Yeah. When I publication nan piece, my first thought was Yann LeCun's position wherever we're not going to scope AGI aliases AI that's smarter than each of america without that beingness intelligence bit, without it learning successful nan existent world, for illustration immoderate different carbon-based life form. It's absorbing that OpenAI is picking up their robotics activity erstwhile again, aft they shuttered successful backmost successful 2021. You merchandise that they were doing beautiful coagulated activity connected robotics, doing things for illustration processing an algorithm tin of solving a Rubik's cube utilizing a human-like hand. Now it seems for illustration OpenAI is not conscionable returning to robotics, but bringing that AI attraction to it, precisely what LeCun wanted. Back successful February, OpenAI parted ways pinch Figure AI, a startup focused connected creating humanoid robots. How do you deliberation soul OpenAI initiatives tin beryllium competitory successful comparison to companies that person been consistently processing humanoid robots? The full clip I'm reasoning astir nan robot connected nutrient hooks. I would beryllium remiss if I did not mention nan robot connected nutrient hooks.

Will Knight: What's nan robot connected nutrient hooks?

Kylie Robison: The humanoid robot connected nutrient hooks? I'm truthful shocked that you haven't seen this. Have you seen this, Michael?

Michael Calore: No.

Kylie Robison: Oh, my God. Do I conscionable person nan astir cursed algorithm connected satellite Earth? I really dream that nan listeners cognize what I'm talking about, and/or should spell hunt it up because it's nan astir horrifying thing. It almost feels for illustration a gimmick. Will, do you cognize what this is? Did you conscionable look it up?

Will Knight: They'll often person a humanoid connected a concatenation truthful it doesn't autumn complete and it goes berserk.

Kylie Robison: Yeah, yeah. And it loses it.

Will Knight: Is that nan one?

Kylie Robison: Yes.

Will Knight: OK, yeah.

Kylie Robison: And past it breaks because location was immoderate problem successful nan algorithm. Yeah, it goes berserk. It's horrifying, but astir apt a bully trading stunt. But I ideate OpenAI's not going to person humanoids connected nutrient hooks.

Will Knight: If they're processing nan hardware, they astir apt will person them connected something.

Kylie Robison: Incredible.

Will Knight: So they don't break themselves if they autumn over. They are going to beryllium competing pinch a batch of companies that are already putting videos of robots retired location and stuff. The advantages that they person is successful nan algorithms already. Large connection models, 1 of nan singular things astir them, is they person a astonishing magnitude of knowing of nan beingness world. But to get to nan adjacent step, whether they're processing hardware aliases conscionable doing nan algorithms, they're going to person to activity connected these systems that are amended capable to really understand really to run successful nan existent world. So really to move limbs and do things for illustration manipulation, which is 1 of nan really large unsolved challenges.

Michael Calore: Yeah. Some of nan large competitors, we should talk about. Kylie, you conscionable mentioned Figure. There's besides Agility Robotics, Apptronik, immoderate family names. Who are nan ones that we should deliberation about?

Will Knight: Yeah, location are a slew of companies doing humanoids now. You'll spot these videos successful your societal media. Agility is 1 of nan amended known ones, Apptronik. Boston Dynamics of people has pioneered this. Figure AI is very salient pinch immoderate of its demonstrations aliases videos. Of course, Tesla. Be remiss not to mention that. I deliberation Tesla is going to be, not conscionable because of Elon, but because they are really very steeped successful processing AI for nan beingness world done their cars, and besides done nan manufacturing that they do, intelligibly Elon seems to beryllium really focused connected that. Now I deliberation they will beryllium a really salient competitor. We should besides mention Unitree, which is this Chinese institution which is now nan biggest shaper of humanoid robots. I deliberation they conscionable revenge for their IPO. I sewage to spot a bunch of their robots. They're really debased cost. They're not ace blase I think, but they're very debased costs and very capable to do things for illustration dancing. I went to a convention successful China wherever they had them doing boxing matches, punching each different and doing kung-fu, which was rather impressive. But yeah, location is simply a increasing number of companies. I deliberation going is besides doing immoderate worldly pinch Apptronik, 1 of nan US firms.

Michael Calore: It feels for illustration nan extremity crippled for nan AI companies is to create humanoid robots that tin usability successful spot of a quality successful mill environments, successful shipping and receiving, and astir nan location doing adjuvant tasks. The robotics companies person nan aforesaid goals arsenic well. They want to beryllium capable to waste a robot that tin do this. But specifically pinch OpenAI, what are their superior motivations present extracurricular of conscionable nan title for AGI?

Will Knight: Well, I deliberation 1 of nan large things to mention is GPT-5, nan latest exemplary which was meant to beryllium a immense leap forward, benignant of flopped. I deliberation they, for illustration others, are looking for nan adjacent large thing, nan adjacent guidance successful AI. Developing a amended knowing of nan beingness world would mean building world models, alleged world models, is going to beryllium a very important point for AI. I deliberation that this could beryllium very important for user hardware, for illustration smart glasses aliases immoderate hardware they're moving on, because robots, by navigating nan beingness world, particularly if they person digits for illustration us, will interact pinch nan aforesaid world that we do. You do person nan imaginable to build models that are going to person a overmuch amended of nan knowing of nan world you're stepping through. I deliberation location could beryllium a bunch of commercialized applications location arsenic well.

Kylie Robison: Yeah. You antecedently brought up Elon Musk. Something I'd heard successful my reporting and I'd reported a information of this successful my newsletter is during a caller all-hands pinch X and xAI people, which are nan companies he besides owns, he said that Tesla and nan humanoid robot from there, he said that's their biggest advantage astatine xAI is having those robotics successful nan background. Which is not thing OpenAI has, they're building this from nan crushed up which is notoriously hard. But nan hype is very real, conscionable for illustration pinch nan AGI hype. Now we've moved connected to immoderate benignant of humanoid robot hype. Sam Altman, nan CEO of OpenAI, precocious told Bloomberg that group would soon beryllium stepping down nan thoroughfare and spot 7 robots locomotion past you doing this, aliases whatever. It's going to consciousness very sci-fi. He said that infinitesimal isn't very acold away. You reported successful your portion that it's a $5 cardinal manufacture correct now and could beryllium worthy five-trillion by 2050. What do you deliberation of this hype? Is it warranted?

Will Knight: Yeah, I deliberation it is really hyped. I deliberation 1 of nan problematic things pinch robots, you tin comparison to ChatGPT and ample connection models, is you tin do these demo videos that are really stunning and group people think, "Oh, nan robot tin do unthinkable things," if you spot it making a portion aliases loading nan dishwasher. But nan point is that often successful those situations, it mightiness beryllium teleoperated. It mightiness only activity erstwhile clip retired of 100. That is nan reality correct now. Getting these things to beryllium ace reliable successful immoderate situation I deliberation is an enormous, tremendous thing. Just look astatine self-driving cars. We're 20 years successful and they're conscionable operating successful immoderate situations now, immoderate roads. We're talking astir thing that's expected to spell into wholly unstructured quality world. I deliberation it's a batch further off.

Michael Calore: OK, that's a bully spot to return a break. We'll travel correct back. Welcome backmost to Uncanny Valley. Today we're talking astir Silicon Valley's renewed liking successful humanoid robots. We conscionable collapsed down why immoderate AI players, for illustration OpenAI and its competitors, are betting that these robots could beryllium nan cardinal to achieving AGI. But quality robots person besides been billed for years and years arsenic being nan early immunodeficiency that would thief america successful our regular tasks, and successful immoderate cases return complete our jobs for us. I want to cognize if that's still nan promise. If we are creating humanoid robots that tin do nan tasks of humans, really overmuch further do we person to spell for that to go existent reality?

Will Knight: There are a ton of robots that do activity that humans do. Through nan history of robotics, it's been nan lawsuit of expanding what they do from incredibly niche things. Manufacturing robots do these very, very precise things, but person zero flexibility. We started to spot successful caller decades much robots grow what they tin do. You spot mobile robots that tin navigate very elemental environments. In nan past fewer years, we've started to spot moreover immoderate humanoids being tested. Amazon has invested successful this company, Agility, which has a humanoid. They've been testing it successful immoderate of their facilities, having it move boxes around. For a institution for illustration Amazon, they person a really, really beardown history of deploying robots successful their ecommerce warehouses successful really valuable and useful ways. What they're doing pinch these humanoids I deliberation is exploring really they tin slot successful for humans successful immoderate situations. They person a very cyclical request for workers and sometimes codification a humanoid, fresh it and do conscionable a elemental point for illustration moving boxes from A to B. But ever pinch robots, nan lawsuit is really overmuch unstructured, unfamiliar worldly tin you get those to do? It's still limited, truthful you person to constrictive it down.

Kylie Robison: Yeah. You already mentioned Amazon. Some of these companies opportunity they want to reside labour shortages successful industries wherever much workers are needed. My first thought arsenic we're talking astir robots astatine Amazon, they don't person to pee successful cups and unionize.

Michael Calore: Yeah.

Kylie Robison: That's very bully for them I suppose.

Michael Calore: Very bully point.

Kylie Robison: They request much workers successful manufacturing, shipping, and logistics. The dream is that these humanoids will return complete nan much dangerous, little desirable jobs, too. Later this year, Boston Dynamics plans to put its all-electric humanoid robot Atlas to activity successful a Hyundai factory. The institution already has a dog-like robot, Spot, and a storage robot, Stretch. What a funny name.

Michael Calore: Nope.

Kylie Robison: Which is already deployed successful business sites. The Hyundai aviator will beryllium nan first clip their humanoid is utilized successful a commercialized manufacturing setting.

Michael Calore: Yeah. If you watch nan videos and nan promotional materials, spell to nan websites of these companies, each of them show robots connected nan mill floor.

Kylie Robison: Yeah.

Michael Calore: And they show them doing things for illustration putting nan bumper screen onto nan bumper.

Kylie Robison: Yeah.

Michael Calore: Or taking boxes disconnected of a palette and organizing it onto support abstraction based connected what's wrong nan box.

Kylie Robison: Yeah.

Michael Calore: Because they tin scan it pinch their robot eyes. Which sounds creepy, but that's besides really nan world useful now.

Kylie Robison: Cool set name, Robot Eyes.

Michael Calore: Indeed. The point that I deliberation a batch of group deliberation astir erstwhile they deliberation astir humanoid robots is nan robot butler. You do spot that imagery successful immoderate of nan promotional materials. Like Figure AI, who we talked about, they person a video of their robot folding nan laundry, watering nan plants, putting nan canine nutrient into nan canine nutrient bowl. I deliberation astir a early for illustration that wherever we person robots successful our location arsenic a batch further disconnected from nan manufacturing and nan storage environments. Will, I deliberation you've hinted astatine this by talking astir nan unstructured reality of life that we unrecorded in.

Will Knight: That's right. In immoderate of those demos, you person to inquire really overmuch programming, really overmuch mentation did they person to do to person it capable nan canine nutrient bowl? The point that humans do is you spell into an unfamiliar home, and tin you provender nan dog? You person to fig retired wherever nan vessel is, and you person to fig retired really to prime it up, which is thing that robots conscionable struggle with. Anything that's unstructured for illustration that, it's going to beryllium overmuch much challenging. I deliberation it is really important to carnivore successful mind erstwhile you're seeing those demos that nan accusation is that it tin do nan benignant of point that a quality does, but it's a batch much analyzable to do that successful nan existent world. This is why things for illustration manufacturing and ecommerce are going to beryllium overmuch much nan situation that you'll spot them travel into first.

Michael Calore: I deliberation erstwhile you talk astir really to programme a robot, nan aged measurement of programming it would beryllium you show it wherever nan vessel is, you show it really to prime up nan bowl, you show it wherever nan canine nutrient is, and you show it wherever nan vessel goes aft it's full. But successful nan world of AI robotic programming, past location each of those small micro-decisions being made astir grip and sensing touch, and nan perspective that you're expected to clasp nan container astatine truthful that nan canine nutrient tin travel retired of it. All of those things are variables that nan robot is going to person to fig retired connected its own, which is astir apt wherever nan instrumentality intelligence facet of it comes in. It seems to me, erstwhile these companies are talking about, "Yeah, we're hiring these group to build these things retired and move them into these machines that we tin usage successful our homes," those are nan sorts of targets that they're looking at, right?

Kylie Robison: I americium reasoning astir conscionable astir a decade ago, AI was figuring retired whether a canine was a canine and a feline was a cat, and now we're astatine this constituent wherever tin it moreover provender these animals feels strange.

Will Knight: That's specified a bully point. The point that is absorbing correct now is that, arsenic difficult arsenic doing a task for illustration that is, we're starting to spot these models which are much general. The thought is you're going to person thing that is yet thing for illustration ChatGPT for connection pinch nan beingness world, it'll beryllium truthful general. You are starting to spot these models that successful very constricted situations tin return what they've learned to do 1 thing, and fixed a caller unfamiliar task, do it successful a flimsy much reliable way. This is 1 of nan things that's creating that excitement. That is still a agelong measurement from figuring retired a wholly unfamiliar kitchen. But we're starting to spot nan beginnings of that, which is really breathtaking I think.

Kylie Robison: Yeah. All technologies person their ain challenges, particularly nan nascent ones for illustration ample connection models and robots. For ample connection models, there's hallucinations, sycophancy. I deliberation nan optimistic group that I talk to would reason humans do that, too. But sticking pinch nan technology, what are nan challenges for humanoids? What are you keeping your oculus on?

Michael Calore: Don't termination people.

Kylie Robison: Right. Number one.

Michael Calore: Do nary harm.

Kylie Robison: We could do a full conception connected that.

Will Knight: Yeah. Well, speaking of that, I've written immoderate stories astir how, if you return thing for illustration hallucinations successful connection models and past extrapolate to a exemplary operating successful nan beingness world, there's a batch much imaginable for it to spell wrong. I deliberation these things, location are going to beryllium a small spot of information issues, reliability issues. Generally, processing nan models that are going to beryllium much wide is not a judge thing. They don't cognize if you tin get capable information to do that successful a wide measurement because we don't person an internet's worthy of each beingness actions that humans do successful nan existent world. That is going to beryllium a existent challenge. I deliberation also, it's really important to retrieve that nan hardware is fundamentally different. We surely don't person hands that are tin of arsenic good power arsenic humans. The thought that you'll beryllium capable to do everything ... Some beingness things you'll beryllium capable to do measurement amended and that's ever been nan case. But location are definite things that nan hardware can't do arsenic well. It is absorbing to spot companies effort to innovate there. Tesla's doing immoderate absorbing stuff, arsenic are others.

Kylie Robison: Is it excessively spicy to talk astir robots and warfare? In a caller podcast pinch Tucker Carlson, Sam Altman said, "I don't really understand really nan subject is utilizing our models." Tucker really pressed. It's crazy. He's like, "Well, I don't person an service of humanoids." One clip I was doing root reach-outs and personification told me, "I can't talk, but you're nan first connected my database if I find retired Sam Altman has an service of humanoid robots astatine nan bottommost of Chase Center." Seriously, we are going toward this future, and we don't moreover understand really it's going to beryllium utilized successful nan military. It feels acold off, but it's concerning.

Michael Calore: Yeah.

Will Knight: I've done immoderate reporting astir AI pre-language models arsenic good arsenic existent worldly being utilized successful nan defense world. I deliberation 1 of nan things is that actually, those models person to beryllium utilized successful very constricted ways because they're truthful unreliable. The interest I conjecture is that group want to unreserved truthful overmuch and you person this pressure, this hype, that you commencement to crook what you're consenting to do. By and large, those systems are excessively unreliable to beryllium utilized successful a batch of contexts really.

Kylie Robison: Thank God.

Michael Calore: Yeah, convey God. All right. Well, convey you some for bringing your champion quality selves to this conversation. We're going to return different break and we'll travel correct back. Thank you some for a awesome conversation. It is now clip to spell to our very quality recommendations. Kylie, you're our impermanent successful nan chair successful nan studio, truthful you get to spell first.

Kylie Robison: Yay.

Michael Calore: What's your recommendation?

Kylie Robison: I americium addicted to a podcast called Armchair Expert pinch Dax Shepherd. They question and reply celebrities and sometimes experts successful fields. They did 1 connected hoarding, which was really interesting.

Michael Calore: Whoa.

Kylie Robison: One connected multidimensional; we unrecorded successful a simulation.

Michael Calore: Nope.

Kylie Robison: Very, very creepy stuff. They don't talk astir tech regularly. They are doing personage interviews. But precocious they've been talking astir AI a batch more. As a newsman who’s listened to this podcast for years, I'm conscionable screaming astatine my phone. At 1 point, they were trying to retrieve nan sanction of Yoshua Bengio. I'm washing my dishes going, "Yoshua Bengio! Yoshua Bengio!" Then this greeting erstwhile I was listening arsenic I was getting ready, they were talking astir mechanistic interpretability, but they didn't opportunity that. They were saying, "Didn't you perceive that AI models person this ain connection that we can't decipher and they tin talk to each other?" I'm erstwhile again screaming astatine my phone. I'm like, "Oh, my God, talk astir mechanistic interpretability. Who would attraction astir that but me?" I emotion that podcast and highly urge listening, aft you perceive to Uncanny Valley of course.

Michael Calore: Even pinch each of its flaws, it's still a bully listen?

Kylie Robison: It has galore flaws, galore flaws. I deliberation sometimes you dislike perceive to it. I don't cognize really to picture you conscionable violently disagree pinch immoderate of Dax's points. Otherwise, I don't deliberation I would perceive those points. I'm successful San Francisco. I'm successful nan AI bubble, and it's an absorbing constituent of view. But yes, pinch each of its flaws, it'd still recommend.

Michael Calore: Great. Yeah, it sounds for illustration you're learning a lot, too.

Kylie Robison: Yeah, that's true. I didn't cognize astir nan simulation theory, really heavy it went and who subscribes to it. Sam Altman successful Keach Hagey's book, he says that he benignant of believes successful it. I was like, "Oh, really nice, nan feline creating nan simulation believes he's God."

Michael Calore: Yeah. That's a full different topic.

Kylie Robison: Yes, it is. Will, what do you recommend?

Will Knight: I'm going to urge what I deliberation is simply a genuinely useful location robot. I person a cat, Leono. He's very bully but will bring successful various animals, dormant aliases alive, into our location overnight. Often including ample rabbits that past tally around.

Kylie Robison: Oh my God.

Will Knight: This feline flap has a camera and machine vision, and it tin show if your feline has a rat aliases thing successful it's mouth. It will say, "Contraband detected," and past artifact them from coming in, which I deliberation is simply a bully thing.

Michael Calore: When it sends you a notification that contraband is detected, do you get a photograph of what your feline has successful its mouth?

Will Knight: Yes. I deliberation you get a video provender of it trying to bring in. There's immoderate connected nan website, which are great. There's 1 that has a very ample frog successful its mouth.

Kylie Robison: You sewage to put a wanted poster pinch these.

Will Knight: Yes.

Kylie Robison: "Wanted: feline pinch frog successful mouth."

Will Knight: Yes. I should opportunity it's called OnlyCat. That's nan sanction of it. I'm really genuinely funny really she brings these rabbits into nan house.

Kylie Robison: Me, too.

Will Knight: I request to put a camera down there, because they're enormous. I consciousness for illustration she has to spell through, spell backmost and get it, and past propulsion it. It must be.

Kylie Robison: Wow. Free rabbit.

Michael Calore: She's really moving hard.

Kylie Robison: Paying her rent.

Will Knight: Yeah. Yeah. Over to you, Mike.

Michael Calore: OK. I would for illustration to urge merino wool T-shirts.

Kylie Robison: Why? I would ideate merino wool is rather sweaty, actually. I don't cognize why I deliberation that.

Michael Calore: No. It's nan opposite.

Kylie Robison: I see.

Michael Calore: We're successful mid-September present successful San Francisco, which is besides known arsenic existent summer.

Kylie Robison: Yeah.

Michael Calore: It's very hot, truthful I've been reasoning a batch astir thermal regulation. I'm besides doing a batch of activities, for illustration hiking and canoeing and doing things retired successful nan wilderness and getting sweaty. The merino wool T-shirt has been really great. A bunch of different group make them. My workfellow Scott Gilbertson, who useful connected nan WIRED reviews team, has been testing a bunch and he's written astir them. He has a buying guideline that gets continuously updated erstwhile a play pinch each nan champion picks. His favourite is nan Proof 72-hour merino wool T-shirt. Mine is nan 1 that Smartwool makes. Merino wool is very, very bully astatine wicking moisture distant from your assemblage truthful it really keeps you cooler. It keeps you from really emotion sweaty, which is nice. Even if you're not peculiarly sweating hard, it'll conscionable make you consciousness fresher. And also, it has earthy antimicrobial properties.

Kylie Robison: Whoa.

Michael Calore: It doesn't stink aft a mates of days, which is really important if you're camping.

Kylie Robison: Right.

Michael Calore: I've been wearing them connected runs. I've been wearing synthetic blends a lot, and I'm complete those, because those are starting to autumn apart. I'm investing successful merino wool. They're rather expensive. They tin costs 60, 70, $80 for a bully shirt. But you'll only request a couple-

Kylie Robison: Yeah.

Michael Calore: ... because you don't request to lavation them arsenic often. Yeah, they're great. It's an old-school innovation.

Kylie Robison: Yeah. It reminds maine that merino wool socks are recommended for Burning Man. That makes a batch much consciousness now.

Michael Calore: Is this wherever you show america that you went to Burning Man this year?

Kylie Robison: Oh, God. No, no, no, I never. I don't cognize what you're talking about.

Will Knight: I'm really wearing a merino wool T-shirt correct now.

Kylie Robison: Whoa.

Michael Calore: Are you?

Will Knight: I tin endorse. I've sewage an Icebreaker one-

Kylie Robison: #notsponsored.

Will Knight: ... which is very nice.

Michael Calore: Yeah. Yeah, this is not spon con. We're surviving nan existent life here. All right. Well, convey you some for being present and talking astir robots. We admit you very overmuch from nan bottoms of our acold mechanical hearts powered by sycophantic LLM models. Will, convey you for coming connected nan show.

Will Knight: Yeah. Thank you for having me. It's been great.

Michael Calore: Kylie, acknowledgment for being here.

Kylie Robison: Of course, happy to.

Michael Calore: Thank you for listening to Uncanny Valley. If you liked what you heard today, make judge to travel our show and complaint it connected your podcast app of choice. If you'd for illustration to get successful touch pinch america pinch immoderate questions, comments, aliases show suggestions, you tin constitute to america astatine uncannyvalley@WIRED.com. Today's show was produced by Adriana Tapia and Mark Lyda. Amar Lal astatine Macro Sound mixed this episode. Mark Lyda is our San Francisco workplace engineer. Kate Osborn is our executive producer. Katie Drummond is WIRED's world editorial director. Chris Bannon is Condé Nast's caput of world audio.