Transcriptly

"ChatGPT Moment" for Robotics Is Coming. The Real Problem Isn't Intelligence | Stanford, Catie Cuan

EO18:51Added Aug 31, 2026

Catie Cuan, Stanford Roboticist, Robot Choreographer, and founder of ART Lab (AI Robot Technology), breaks down why the humanoid race led by 1X and Unitree m...

Watch on YouTube →
Contributed by 刘嘉琪

Transcript

Transcript format

  1. Intro

  2. 00:00It's a tremendous time to be working in
  3. 00:01robotics. It feels like every week or
  4. 00:04multiple times a week there's new
  5. 00:05announcements about amazing things
  6. 00:07happening. So, we're in this absolutely
  7. 00:10unbelievable moment [music] right now. I
  8. 00:13think the thing that we're missing in a
  9. 00:16lot of this is what is that human
  10. 00:18interaction piece look like? [music] The
  11. 00:21question though in terms of building
  12. 00:23humanoids for me is is this really what
  13. 00:26we want? Is this what literally my
  14. 00:29[music] 77year-old English is his third
  15. 00:31language dad wants? Even if maybe it can
  16. 00:35do some tasks. We've sort of locked
  17. 00:37ourselves into a slightly myopic vision
  18. 00:40of what our future with robots can look
  19. 00:42like. And I would love for that aperture
  20. 00:44[music] to be broadened.
  21. 00:46My name is Katie Quan. I am the founder
  22. 00:49and CEO of Art Lab. It stands for AI
  23. 00:52robot technology. I'm a robot
  24. 00:54choreographer, a roboticist, a [music]
  25. 00:56researcher, an entrepreneur, and an
  26. 00:58artist. My mission is to make nent
  27. 01:02technologies like AI and robots [music]
  28. 01:05more humanistic, whimsical, and widely
  29. 01:08accessible. And I do that through
  30. 01:10business, [music] research, and art. My
  31. 01:12definition of humanistic is human-
  32. 01:15centered. So thinking about how we can
  33. 01:17drive interactions and experiences with
  34. 01:20these machines that help humans feel
  35. 01:22more empowered and safe, but also help
  36. 01:25them live healthier, better lives.
  37. 01:36We're at the Stamford Robotics Center on
  38. 01:38the Stamford campus in Palo Alto,
  39. 01:40California. The idea was to bring
  40. 01:42together a variety [music] of
  41. 01:44disciplines and engineering backgrounds
  42. 01:47to work on flagship projects [music] in
  43. 01:50robotics like domestic robotics,
  44. 01:52industrial robotics, medical, underwater
  45. 01:55and [music] exploration or in my case
  46. 01:57education, art, culture, movement and
  47. 01:59dance.

  48. Billions of Robots Are Coming, Are We Ready?

  49. 02:05We're going to have billions of robots
  50. 02:07in our lifetimes. [music] It only took
  51. 02:1013 years for there to be a billion
  52. 02:12iPhones on Earth. We have so many people
  53. 02:15working in robotics right now. And so
  54. 02:17we're [music] going to see billions upon
  55. 02:19billions of these robots showing up in
  56. 02:20our everyday spaces, offices, hospitals,
  57. 02:24hotels, homes, caretaking facilities.
  58. 02:27When you take all of these robots out of
  59. 02:29these very niche, narrow vertical
  60. 02:31applications, and you give them a lot
  61. 02:33more intelligence and put them in situ
  62. 02:35with people, you have this huge
  63. 02:37opportunity. and also challenge which is
  64. 02:39how do you make humans and robots
  65. 02:41legible to one another and that's what
  66. 02:44my work is interested in. Human
  67. 02:46interaction is a grand challenge for
  68. 02:48robots. [music] It's been the case for
  69. 02:50the last several decades. Humans are
  70. 02:52really unpredictable. We [music] tend to
  71. 02:55do and say things and move in ways that
  72. 02:57you might not always expect. The
  73. 02:59problems that I'm interested in are how
  74. 03:01can we make robots more legible to
  75. 03:03people, but also more animating, more
  76. 03:05inspiring [music]
  77. 03:06to delight and bring awe into our lives.
  78. 03:10And so the thing that I have always been
  79. 03:12curious [music] about throughout my
  80. 03:13whole career is what kinds of
  81. 03:15interactions can we facilitate that both
  82. 03:19inspire and surprise us but also open
  83. 03:22the door into what [music] technology
  84. 03:24can be where it's not exclusively tool
  85. 03:28but it also is something [music] that
  86. 03:29you could socialize with or learn from.
  87. 03:32We are building a kind of model that we
  88. 03:34call a VLI a vision language interaction
  89. 03:36[music]
  90. 03:37model. This is the first kind of model
  91. 03:39that's existed which takes [music] the
  92. 03:42impetus of what it should be doing from
  93. 03:44humans in the environment. LLMs [music]
  94. 03:46are acting on human information and then
  95. 03:48maps that again to a series of robot
  96. 03:50actions but measures the success [music]
  97. 03:52or the failure of those robot actions on
  98. 03:54the basis of a human response. So we can
  99. 03:57tell is the human increasing in positive
  100. 03:59affect or decreasing was that what they
  101. 04:02wanted or not. And so we think this is
  102. 04:04the first start for robots that can
  103. 04:07really naturally and intuitively
  104. 04:09socialize and interact with people. And
  105. 04:11we want those interactions to be safe,
  106. 04:13legible, and clear. I grew up in the Bay
  107. 04:16Area from the time that I was really
  108. 04:18young. I always loved numbers and I
  109. 04:21always loved to dance. I loved math and
  110. 04:23I loved science, too. My dad's Cuban, so
  111. 04:26my whole family grew up social dancing.
  112. 04:28And I would say if you don't know how to
  113. 04:30dance, you're probably not Cuban. I
  114. 04:31always grew up with those two [music]
  115. 04:33very strong interests. And then at some
  116. 04:36point when I was getting towards the end
  117. 04:37of college, I had interned here at a big
  118. 04:39technology company in the Bay Area and I
  119. 04:42knew I really wanted to try what it was
  120. 04:44like to be a professional [music] dancer
  121. 04:46and that you only have a very small
  122. 04:48window in your early adulthood when you
  123. 04:50can go for it. I was a little scared. I
  124. 04:53didn't know anyone who had been [music]
  125. 04:54a professional dancer in my immediate
  126. 04:57sphere, my friends, my family. So, when
  127. 04:59I graduated from college, I wound up
  128. 05:01taking a job doing private equity
  129. 05:03consulting in New York. I realized
  130. 05:05really quickly [music] that wasn't for
  131. 05:06me, mainly because I had this childhood
  132. 05:09dream and I figured I'm only going to be
  133. 05:1121 once. [music] Um, and if I don't go
  134. 05:13for this now, I I'll regret it for the
  135. 05:15rest of my life. And so, I actually
  136. 05:17wound up leaving that job and then
  137. 05:19became a professional dancer in New York
  138. 05:21City. I had my own dance company for a
  139. 05:22long time. I danced for the Metropolitan
  140. 05:24Opera Ballet at Lincoln Center in New
  141. 05:26York, which was crazy experience. So,
  142. 05:29I'll never forget what it's like to bow
  143. 05:30on that stage. Around that time, my dad
  144. 05:33also got sick. He spent a while in the
  145. 05:35hospital. I think my dad seeing him be
  146. 05:39sick [music] in the hospital was really
  147. 05:41scary because here's somebody who you
  148. 05:43care about and you love so much and you
  149. 05:45want them to be healthy and safe. I
  150. 05:48think [music]
  151. 05:49the fact that he was surrounded by all
  152. 05:51these big machines that were supporting
  153. 05:53him, they were meant to help him
  154. 05:56literally survive. They were beeping and
  155. 05:59booping and had these proportions and
  156. 06:02colors and this sort of overall presence
  157. 06:05that I think was oppressive to him. And
  158. 06:09it was also opaque. He [music] didn't
  159. 06:11understand what they were doing. And
  160. 06:13there were very few opportunities where
  161. 06:16someone could explain it to him. And
  162. 06:18seeing him be that vulnerable, you know,
  163. 06:20it occurred to me that a lot of other
  164. 06:22people are in that same position all the
  165. 06:25time.
  166. 06:27One of the [music]
  167. 06:28projects I worked on when I was at
  168. 06:29Google was called Music Mode. When I
  169. 06:32first got there, you know, we had 200
  170. 06:34robots and they were all driving around
  171. 06:36Google buildings doing useful work. They
  172. 06:38were wiping tables, sorting trash,
  173. 06:41resetting conference rooms. They were
  174. 06:43doing valuable things [music] and I was
  175. 06:45there as an artist and residence and I
  176. 06:47remember going around and asking a few
  177. 06:50people who were software engineers you
  178. 06:51know what do you think about the robot
  179. 06:53and they were like h it's fine [music]
  180. 06:55you know it's a robot it was kind of
  181. 06:57cute or kind of cool initially and now
  182. 06:59it very slowly [music] does these
  183. 07:01different tasks and I'm thinking what a
  184. 07:02missed opportunity you know here's this
  185. 07:04amazing new flagship Vanguard technology
  186. 07:08showing up in [music] your lives and you
  187. 07:10feel a little bit ambivalent towards it
  188. 07:11I think for some people that would be a
  189. 07:13success. For me, it was like, huh, we
  190. 07:14could be doing a lot more. So, my first
  191. 07:16idea was to get the robots to play
  192. 07:19[music] music. I thought, why don't they
  193. 07:21sort some trash and wipe some tables and
  194. 07:23then they can go over here and play the
  195. 07:25cello? We worked on that for a few weeks
  196. 07:27[music] and then my friend and colleague
  197. 07:29Tom came up to me one day and he said,
  198. 07:31"You know, Katie, the robot generates a
  199. 07:33lot of data. [music] Instead of the
  200. 07:35robot playing music, what if the robot
  201. 07:39was the music? what if this data became
  202. 07:42the music? That was such a clever
  203. 07:44breakthrough [music] that he came up
  204. 07:46with because we then started working
  205. 07:47with a composer named Peter Van Stratton
  206. 07:49on this piece of software that we called
  207. 07:51music mode. And what we did was that
  208. 07:53every time a different part of the robot
  209. 07:55moved, [music]
  210. 07:56we would map it to a different sample of
  211. 07:59music. For example, me as the robot, I
  212. 08:01open and close my gripper, it goes cling
  213. 08:04cling. or [music] if I turn the torso of
  214. 08:07the robot, it would make this beautiful
  215. 08:09bass noise like boom boom boom. And
  216. 08:12Peter, [music] who was our composer, was
  217. 08:14such an amazing artist to work with
  218. 08:16because he understood [music] that you
  219. 08:18could move all these joints of the robot
  220. 08:20in any combination. So it was a real
  221. 08:22combinatorics problem. And so we worked
  222. 08:25on this for a couple months and then we
  223. 08:27shipped it on the robot this software
  224. 08:29called music mode. And we would just be
  225. 08:31able to say, "Hey robot, turn on music
  226. 08:32mode." And then as it's wiping the
  227. 08:35table, you just hear this symphony. Then
  228. 08:39I started getting emails from people at
  229. 08:42Google who I did not know that were
  230. 08:44like, "This is the most incredible thing
  231. 08:47I have ever seen." Like, I'm sitting at
  232. 08:49my desk crying because I never thought
  233. 08:53that a robot could be so beautiful. I
  234. 08:55never thought that a machine could make
  235. 08:57me feel this way. So we shipped music
  236. 08:59mode on every single one of the robots
  237. 09:00and and it was crazy. Like we would have
  238. 09:02nine robots that were sorting trash at
  239. 09:04the same time and you would go around
  240. 09:05and say, "Hey robot, turn on music
  241. 09:06mode." And just this beautiful symphony
  242. 09:09arose. [music]
  243. 09:10That for me was such a proof point that
  244. 09:12we need these experiences [music] with
  245. 09:14our technologies. We want to feel
  246. 09:17something. We want to feel engaged.
  247. 09:23We have been so influenced by fiction

  248. Break Out of the Humanoid Imagination

  249. 09:26and media. dozens, I mean, I [music]
  250. 09:28think at this point, of humanoid
  251. 09:30companies that are making robots that
  252. 09:31look like humans. And often the argument
  253. 09:33that you hear is, well, they're they're
  254. 09:35made for human spaces, so we need them
  255. 09:37to look like people. Rodney Brooks says
  256. 09:39it really well. He's a titan in our
  257. 09:41field, and he says, the closer that
  258. 09:42something looks to a person, the higher
  259. 09:45the expectations people have for it. And
  260. 09:47you can't help yourself. Your mirror
  261. 09:49neurons go off. you start having all
  262. 09:51these expectations that because it's
  263. 09:53bilaterally symmetric and it's roughly
  264. 09:55in the same proportions as a human, it
  265. 09:57should be able to do what a human does.
  266. 09:59The expectations that you set are so
  267. 10:01high and and I think also for people,
  268. 10:03you know, if I ask whether it's my
  269. 10:0677year-old dad or whether it's one of my
  270. 10:09nieces and nephews and I'm like, "Hey,
  271. 10:11is that the robot that you want?" I
  272. 10:13think there isn't enough expansive
  273. 10:15thinking into what robots can be. We've
  274. 10:18sort of locked ourselves. we meaning the
  275. 10:20the broader robotics field into a
  276. 10:22slightly myopic vision of what our
  277. 10:25future with robots can look like and I
  278. 10:27would love for that aperture to be
  279. 10:29broadened.
  280. 10:31There's this phrase called dirty, dull
  281. 10:33and dangerous. So for a long time robots
  282. 10:35were useful for tasks that were too
  283. 10:37dirty, dull or dangerous for people to
  284. 10:40do. And then I think there was appetite
  285. 10:42for robots to do other things that were
  286. 10:44not exclusively dirty, dull and
  287. 10:46dangerous jobs. It's just that I think
  288. 10:48we've still gotten locked into this
  289. 10:50paradigm of like robots are for utility.
  290. 10:53They need to alleviate [music]
  291. 10:55certain parts of physical labor in our
  292. 10:57lives when I think robots can do a lot
  293. 10:59more than that. The Paru robot was a
  294. 11:01very soft seal robot that was popular I
  295. 11:04think in the mid2010s [music] um that
  296. 11:06would just provide companionship for
  297. 11:08people who were ailing or lonely. There
  298. 11:10have been robots used for helping
  299. 11:13children with autism learn social
  300. 11:14skills. So, I think we have a lot of
  301. 11:17opportunities [music] for robots to show
  302. 11:19up in our lives and be beneficial for
  303. 11:21humans that are not exclusively [music]
  304. 11:23humanoids that do our dishes in our
  305. 11:25laundry. I've worked with a lot of
  306. 11:26different robots in my career. I've
  307. 11:28gotten to use Roombas in a dance
  308. 11:31project, small humanoid robots, drones,
  309. 11:35big industrial robot arms, [music] these
  310. 11:38mobile manipulators from Google. I've
  311. 11:40used quadripeds from Boston Dynamics.
  312. 11:42The very first experience I ever had
  313. 11:44dancing with a robot in Amy's lab in
  314. 11:462017 was transcendent. [music] It was
  315. 11:49the first time that I got to imbue
  316. 11:51myself inside of something and then also
  317. 11:53interact with it. [music] So I created
  318. 11:55this choreography for the machine and
  319. 11:57then it performed the choreography and
  320. 11:59then I got to dance with it and I was
  321. 12:01like [music] wo. I felt myself on this
  322. 12:04very long continuum of humans and our
  323. 12:07tools.
  324. 12:07[music]
  325. 12:07everything from early stone tools and
  326. 12:10fire to the wheel and the printing press
  327. 12:13and [music] the modern computer and
  328. 12:14these robots. And I thought, wow, this
  329. 12:17is part of definitionally what it means
  330. 12:19to be a human is to be in relationship
  331. 12:21with tools. Aristotle used to say we
  332. 12:23shouldn't be called homo sapiens. Homo
  333. 12:25sapiens, you know, wise man, intelligent
  334. 12:27man. There are a lot of species in the
  335. 12:29world that are intelligent. We should be
  336. 12:30called homoche. The thing that makes us
  337. 12:32unique is our relationship to our tools
  338. 12:34and our ability to make tools. What I've
  339. 12:36learned since then, oh my gosh, so much
  340. 12:38frustration, so much debugging, there's
  341. 12:41so many limitations. How my relationship
  342. 12:44has changed is I've become much more
  343. 12:46acutely aware of how amazing human
  344. 12:49beings are relative to all of the robots
  345. 12:52that I work with. They don't have the
  346. 12:54expressive potential that people have.
  347. 12:57Humans need to give ourselves a lot of
  348. 12:59credit. We can do a wide diversity of
  349. 13:02things. If you think about what was
  350. 13:04required to come into this building
  351. 13:06today, [music] you had to open a door.
  352. 13:08It was a door that you had never seen
  353. 13:10before with a handle that you couldn't
  354. 13:12have predicted. There are infinitely
  355. 13:14many different kinds of door handles
  356. 13:15with different friction [music] and
  357. 13:17torque required to get the door open.
  358. 13:18You had to walk down stairs that you'd
  359. 13:20never seen before in a space with a
  360. 13:23proportionate move, you know, a large
  361. 13:25object around, whether that's a backpack
  362. 13:26or something else that you're carrying.
  363. 13:28[music] You needed to then socialize and
  364. 13:31show up and make conversation and be
  365. 13:33acutely aware of timing. And to build
  366. 13:37the type of [music] dexterity and acuity
  367. 13:40and specificity that we have took not
  368. 13:44only your lifetime to evolve, but also
  369. 13:47the many generations before you that
  370. 13:49encoded all of that stuff in your DNA.
  371. 13:51We are really remarkable animals. I read
  372. 13:55a beautiful article yesterday about
  373. 13:58writing as a craft and how it's a
  374. 14:01hardearned craft in many cases that
  375. 14:04anything that's worth doing, [music]
  376. 14:07anything that is a craft in the richest
  377. 14:09sense of the word takes effort. I feel
  378. 14:12that way about many different [music]
  379. 14:14things, not only writing. That anything
  380. 14:16that's really worth doing or worth
  381. 14:19earning is hard. Otherwise, everybody
  382. 14:22would [music] do it. otherwise it
  383. 14:23wouldn't be considered worthy. I'm
  384. 14:26reminded again of this professor here at
  385. 14:28Stanford named Mark [music] Skyler Scott
  386. 14:30where he said, "My goal in my research
  387. 14:32is to 3D print a human heart and I think
  388. 14:35it's going to take me 25 years to do it.
  389. 14:38But it is a challenge that's worthy of
  390. 14:40your [music] life's attention because
  391. 14:42the impact at the end is so lifealtering
  392. 14:45literally that it will have been worthy
  393. 14:47of your life's [music] attention. I
  394. 14:49think we need things that are still
  395. 14:51challenging. Part of what I learned in
  396. 14:53my dance training [music] is that to
  397. 14:55really be at the top, to be at the
  398. 14:58highest levels, you need to [music] work
  399. 15:00hard. It is not easy. Otherwise,
  400. 15:02everyone else would do it. When you show
  401. 15:04up to an audition [music] and there's
  402. 15:05400 people there and they're going to
  403. 15:07hire four people, you better be good.
  404. 15:10Otherwise, everybody else in that room
  405. 15:12is good, [music] too, and you should go
  406. 15:14home and get better. And a thing I
  407. 15:16certainly don't [music] see myself
  408. 15:17delegating in the future is this pursuit
  409. 15:21of high impact challenging [music]
  410. 15:24problems with solutions that I want to
  411. 15:27build with people who inspire me.

  412. You Can Build Anything. The Question is Why

  413. 15:33I teach a class at Stamford called CS
  414. 15:36[music] 334 Robots and Arts. It is the
  415. 15:39first class that's cross-listed between
  416. 15:41computer science and theater arts and
  417. 15:43[music] performance studies. And the
  418. 15:45invitation was that they could use AI
  419. 15:48and robots to make [music] a creative
  420. 15:51statement with a political,
  421. 15:53philosophical, or social underpinning.
  422. 15:55[music] You couldn't just have a robot
  423. 15:57do something. You actually needed to
  424. 15:59make an assertion with that project
  425. 16:02about I believe the world [music] should
  426. 16:03be like this or this is something that
  427. 16:06inspires me about the world or here's
  428. 16:08something that I don't like and I would
  429. 16:09like to change it. [music] The very
  430. 16:10first day of the class, I was thinking
  431. 16:12about here are all of these amazing
  432. 16:14students at Stanford Nexus of
  433. 16:17technology, culture, [music] art, but
  434. 16:19we're in this moment that is so special
  435. 16:22when it seems that everything [music] is
  436. 16:24possible. People, researchers here at
  437. 16:26Stanford are 3D printing organs where
  438. 16:29you can take an autonomous car all
  439. 16:31around San Francisco or Austin or
  440. 16:33Phoenix, depending on where you are,
  441. 16:34where we also are using large language
  442. 16:36models to potentially [music] make
  443. 16:37sequences of DNA. so we could build new
  444. 16:40life forms. So, we're in this absolutely
  445. 16:43unbelievable moment right now. Well,
  446. 16:46ostensively this class is you're going
  447. 16:48to make a project. But really what this
  448. 16:50class is about [music] is when you are
  449. 16:52in a place of infinite opportunity to
  450. 16:55build so quickly, [music] easily, and
  451. 16:57seamlessly. The question that's
  452. 16:59interesting is not what I should build,
  453. 17:02but the question is why am I building
  454. 17:05[music] it? do I want it to say or to
  455. 17:08do? When you're in this space of
  456. 17:10infinite choice, you need to understand
  457. 17:12why you're making decisions, [music]
  458. 17:14where your values are. We all have a
  459. 17:16finite amount of resources that we can
  460. 17:18[music] dedicate to anything. We have
  461. 17:21our energy, our brilliance, our network,
  462. 17:24our relationships, our technical [music]
  463. 17:25skills. And we have this really precious
  464. 17:28thing which is our time. That is [music]
  465. 17:30the only resource that you cannot scale,
  466. 17:32you cannot get back, that is equal for
  467. 17:35everyone. does not matter who you are,
  468. 17:36where you are, your time [music] is
  469. 17:39tremendously valuable. The dedication of
  470. 17:42that time, it's one of the biggest
  471. 17:43choices that [music] you make in your
  472. 17:45life. And we need to know why we make
  473. 17:49choices because it's a reflection of the
  474. 17:52it's the spending of this really
  475. 17:54precious [music] resource which is our
  476. 17:55time and knowing why you're building and
  477. 17:58making what you're [music] doing, what
  478. 18:00the statement is that you wanted to
  479. 18:01make. So, I think the why for my
  480. 18:03students and and for everyone else, when
  481. 18:06it's authentic, when it's real, it will
  482. 18:08carry you through a lot. If you know why
  483. 18:10you're doing it, it can create less
  484. 18:12resilience in these challenging
  485. 18:15circumstances where you need to be able
  486. 18:16to overcome and do better, be better,
  487. 18:20work harder, learn faster, go farther
  488. 18:22than you thought. The why carries you
  489. 18:24through a lot of that.