The Future of AI: What Google Research Says Is Coming Next — Transcript

Full transcript

1. 0:00People won't need to know how to code
2. 0:01anymore.
3. 0:02>> Hype. I think people need to learn
4. 0:04everything.
5. 0:04>> Are we going to need fewer of them?
6. 0:06>> Oh, I think we're going to need more of
7. 0:07them. The work from Google took two
8. 0:10people and they generated 200,000
9. 0:13different experiments.
10. 0:15>> This is Lizzy Dorfman, principal PM at
11. 0:18Google Research. While we're worried
12. 0:20about what AI might replace, scientists
13. 0:23are using it to make discoveries that
14. 0:25weren't possible before. I asked her
15. 0:27what AI is actually changing about the
16. 0:30work, which skills still matter, and
17. 0:33what she would study if she were 22
18. 0:36today.
19. 0:36>> I tell the computer what to do and then
20. 0:37I go to sleep.
21. 0:38>> First of all, thank you for being here.
22. 0:40I'm very excited to talk to you.
23. 0:42>> Likewise.
24. 0:43>> So, you started your career in public
25. 0:44health. When did you first realize that
26. 0:46AI was going to change your field?
27. 0:48>> So, um I studied a interesting field
28. 0:50called public health genetics. I would
29. 0:52say that I had a realization that AI was
30. 0:55going to change the field, I'm going to
31. 0:56say around 2016. Um this was right after
32. 1:00I finished graduate school. I had just
33. 1:02come back to Google and we were in the
34. 1:03deep learning heyday, which most people,
35. 1:06general public, might remember as how we
36. 1:09first differentiated pictures of cats
37. 1:11from dogs on the internet or then we had
38. 1:13pictures of dogs from pictures of
39. 1:15blueberry muffins that looked a lot like
40. 1:16dogs.
41. 1:17Um but it was this really powerful
42. 1:19technology and I was working with these
43. 1:21brilliant scientists, Mark DePristo and
44. 1:23Ryan Poplin, who were genome scientists.
45. 1:27And they said, "What if we try to apply
46. 1:29deep learning to a common problem in
47. 1:31genetics?" Uh it was the first time that
48. 1:32anyone had really done that. We didn't
49. 1:34think about using deep learning for that
50. 1:35type of data. And so we did it. We made
51. 1:37this tool. We actually still maintain
52. 1:39this tool today. It's called Deep
53. 1:40Variant. This was a common problem.
54. 1:42There were existing tools, but we said,
55. 1:43"Can we make a tool that uses deep
56. 1:45learning?" We did. We entered it in an
57. 1:47FDA competition and won highest
58. 1:49accuracy. We published it in Nature
59. 1:51Biotechnology. We released it open
60. 1:53source. But the aha moment, that's all
61. 1:55context, the aha moment was when I
62. 1:57realized that all of the existing tools
63. 2:00that people had that had been using sort
64. 2:03of very hand curated code, any little
65. 2:06change that somebody made to how they
66. 2:09prepared the DNA, how they sequenced it,
67. 2:11what type of instrument they were
68. 2:12making, meant that they had to make a
69. 2:14whole new tool. But with deep learning,
70. 2:15all we needed was example data. And
71. 2:18there's a lot of different types of
72. 2:19sequencing instruments, there's a lot of
73. 2:20different ways we can prepare DNA, and
74. 2:22it was this aha moment that well,
75. 2:24clearly we're going to do it this way
76. 2:26instead of this way going forward.
77. 2:28>> There's a lot of AI usage in science
78. 2:31we're not really aware about. Do we feel
79. 2:33like we're actually on the verge of
80. 2:35major scientific breakthrough, or do you
81. 2:37think AI is more hype?
82. 2:39>> I think we are absolutely on the verge
83. 2:41of a major scientific breakthrough. Um I
84. 2:43will say it's also true that real
85. 2:47breakthroughs and transformation are
86. 2:49deeply entangled right now with lots of
87. 2:52hype and lots of hope. Um but it's
88. 2:54absolutely the case that we are seeing a
89. 2:56major major change in the pace of
90. 2:59scientific discovery.
91. 3:00>> I have a little rapid-fire game prepared
92. 3:02for you.
93. 3:03>> Great.
94. 3:04>> I'm going to give you three cards. We
95. 3:06have
96. 3:07hype,
97. 3:08real, and too early. I'm going to read
98. 3:10out some common predictions about AI in
99. 3:13science, and you'll tell me if you think
100. 3:15it's hype, or too early, or real. Okay,
101. 3:18are you ready?
102. 3:18>> I'm ready.
103. 3:19>> It'll be a rapid-fire. So, AI will help
104. 3:22cure cancer.
105. 3:23>> Hype. Hype because this doesn't mean
106. 3:25anything.
107. 3:26>> Oh.
108. 3:26>> Curing cancer.
109. 3:27>> Oh, okay. We'll we'll talk more about it
110. 3:30in a bit, but AI will predict the
111. 3:32weather accurately.
112. 3:33>> Already happening.
113. 3:34>> Oh, okay. AI will discover things humans
114. 3:37couldn't find on our own.
115. 3:40>> Already happening. I'm going to
116. 3:41emphasize the word couldn't is different
117. 3:43from the word wouldn't.
118. 3:46>> People won't need to know how to code
119. 3:48anymore.
120. 3:48>> Hype. I think people need to learn
121. 3:50everything, and I'm happy to explain
122. 3:51why.
123. 3:52>> AI can win a Nobel Prize.
124. 3:54>> Um not under current guidelines for who
125. 3:56is eligible to win a Nobel Prize.
126. 3:59>> And the last one, AI will replace
127. 4:00scientists.
128. 4:01>> No, I don't believe it will.
129. 4:02>> Wow, okay. Let's talk about it. So, cure
130. 4:05cancer. This is a personal question for
131. 4:07me because my dad recently had cancer.
132. 4:09Luckily, he's he's recovered now.
133. 4:11>> that.
134. 4:12>> But it it was a complicated process of
135. 4:14both like medicine,
136. 4:17um chemotherapy, and surgery. So, like
137. 4:19all but the robot did the surgery for
138. 4:21him.
139. 4:21>> Amazing.
140. 4:22>> Yeah, so we we are getting closer to um
141. 4:26solving cancer
142. 4:27>> Mhm.
143. 4:28>> in various different ways.
144. 4:29>> So, this is a great illustration of how
145. 4:32diverse the specific conditions are when
146. 4:35we say cancer. And it's in its a very
147. 4:38good thing that some cancers are already
148. 4:41curable. And there's progress that we've
149. 4:43been making in different types of
150. 4:45treatments. And this is wonderful. And
151. 4:48this is part of why it's exciting for me
152. 4:49to get to do science is to think about
153. 4:51these real consequential problems and
154. 4:53how to advance the frontier. But there's
155. 4:55a lot of cancers that we don't have good
156. 4:57treatments for. So, I'm so glad that
157. 4:59your dad was able to get treated and had
158. 5:00a good outcome.
159. 5:01>> Thank you. Thank you. Yeah, I'm I'm
160. 5:03excited for getting older now that we
161. 5:06can do more.
162. 5:07>> Yeah.
163. 5:07>> So, the weather, AI can already predict
164. 5:10weather accurately, you said?
165. 5:11>> Yes, in fact, AI has been making really
166. 5:15exciting progress in weather prediction
167. 5:17and for nowcasting, which is like what's
168. 5:18happening right now, but also near-term
169. 5:20weather forecasting. So, when I open up
170. 5:23my phone and said, "What should I wear
171. 5:24today?" because how cold is it going to
172. 5:26be and is it going to rain?
173. 5:28A lot of that already uses lots of AI.
174. 5:31But I do want to speak to
175. 5:33a different facet of what we think of as
176. 5:35weather, which is that we're starting to
177. 5:36see more you know, severe weather
178. 5:39events. Right? We might have a tropical
179. 5:41storm or
180. 5:43you know, a major snowstorm or a
181. 5:44blizzard. Or we have we have a rainstorm
182. 5:46that results in a flood. And this is
183. 5:48actually a really exciting set of places
184. 5:51where AI is also helping. At Google, we
185. 5:54have
186. 5:55a lot of work that we do. We call it
187. 5:57crisis resilience. Our goal is that
188. 5:59nobody is surprised by a natural
189. 6:01disaster. You might have received one if
190. 6:03you were in an area that had a risk of a
191. 6:05flood or an earthquake. I don't know if
192. 6:06you've gotten an earthquake alert on
193. 6:07your phone before. Um but AI is is part
194. 6:10of how we're able to more accurately
195. 6:13understand what's coming and help people
196. 6:15know so that they can take action and
197. 6:17prepare.
198. 6:17>> That's really exciting. I I can't wait
199. 6:20to not have to worry about weather.
200. 6:22One Let's do one more. So, about coding,
201. 6:26you think that would you mark that one
202. 6:28as a hype that people won't need to code
203. 6:30anymore? Let's talk more about that.
204. 6:32>> So, I'll I'll give you um
205. 6:34to explain my rationale there. So, when
206. 6:36I when I went to graduate school, when
207. 6:38you are when you're doing a doctorate,
208. 6:40you're like, "I'm going to make I'm
209. 6:41going to discover some new facts. What
210. 6:43are the new facts that I'm going to
211. 6:44discover and get to contribute to the
212. 6:46world's knowledge?" And I decided that I
213. 6:48wanted to actually do a clinical study
214. 6:50from start to finish. I wanted to pick
215. 6:52the problem. I wanted to write the study
216. 6:53design. I wanted to find funding. I
217. 6:55wanted to get it approved by the
218. 6:57Institutional Review Board. I wanted to
219. 6:59make the study materials. I wanted to
220. 7:00recruit patients and take their samples
221. 7:02and analyze them in the lab and then
222. 7:03analyze the data and write it up and get
223. 7:05it published. And I'm I'm kind of
224. 7:07shocked in retrospect that I was able to
225. 7:09finish all of that. Um I wanted to do
226. 7:11all of those things not because I
227. 7:13expected that I was going to be doing
228. 7:14all of those things, but it was
229. 7:16important for me to understand
230. 7:18what is the process that it takes to go
231. 7:21from idea to results. And I think that
232. 7:23we still need to have a fundamental
233. 7:25understanding of what's happening um
234. 7:28because that's where opportunities are
235. 7:30to to sort of come up with new ideas.
236. 7:32It's also where you think about what
237. 7:33might have gone wrong that would have
238. 7:35implications for is this real or or is
239. 7:37there a problem here? Um so, I think
240. 7:39that some fundamental basic
241. 7:41understanding of of coding and of math
242. 7:44is absolutely still a part of what's
243. 7:45important for us to be able to get the
244. 7:47most out of these technologies and do
245. 7:49the most good for the world.
246. 7:50>> Definitely agree with that. I think
247. 7:52foundation is foundation. It's going to
248. 7:55last beyond AI. Yeah. Okay, so we talked
249. 7:57about various different hype or real
250. 7:59things with AI. What is the coolest
251. 8:02thing that you have personally seen AI
252. 8:04help scientists discover?
253. 8:06>> Can I share two? Is that
254. 8:08Okay. How how to choose so many of my
255. 8:10favorite discoveries? So, I'm going to
256. 8:12share some really specific examples that
257. 8:13I had the chance to be involved in. Um,
258. 8:15so I I love the brain.
259. 8:17I think everybody loves the brain. Like
260. 8:19it's what makes us human and
261. 8:21you know, individuals and and and
262. 8:23ourselves.
263. 8:24Um, and
264. 8:26my dad's a neurologist. So, I grew up
265. 8:27with brains. My dad brought a human
266. 8:29brain in for show and tell when I first
267. 8:31studied human biology in fifth grade.
268. 8:33Um, but he said to me like he brought it
269. 8:35home. He was like, "We don't understand
270. 8:37how this works."
271. 8:38And I and I over the course of my career
272. 8:40like learned it's true. We don't
273. 8:42actually understand like a lot of
274. 8:43fundamentals about how our brains work.
275. 8:45My team has for the last decade actually
276. 8:48been doing a really cool type of
277. 8:49neuroscience which involves mapping
278. 8:51every single cell and every single
279. 8:53cellular connection in the brain. And
280. 8:56um, couple of years ago we had the
281. 8:58chance. This was in partnership with
282. 9:00Harvard a professor named Jeff Lichtman
283. 9:02there to map a piece of human brain.
284. 9:06Every single cell and every single
285. 9:07cellular connection in a piece of human
286. 9:08brain. And the coolest part, so AI was
287. 9:11essential to making that happen. But the
288. 9:13coolest part was that that discovered
289. 9:16that there are structures in our brains
290. 9:18that we had never seen before and that
291. 9:20we have no idea what they do if they do
292. 9:22anything at all. And I think that's
293. 9:24really cool because we didn't even know
294. 9:25that was there before. And AI helped us
295. 9:28to understand what was there which lets
296. 9:30us then think about, "Okay, let's let's
297. 9:32study and try to figure out what that
298. 9:33does." So, that's that's example one.
299. 9:35Uh, example two is going to be something
300. 9:36a little more recent. So, um, my team
301. 9:39for the last couple of years has been
302. 9:40trying to think about how to build LLM
303. 9:44AI based technologies that can help to
304. 9:46accelerate scientific discoveries and
305. 9:49enable scientists to be much much more
306. 9:51productive. And one of the capabilities
307. 9:53that we focused on is the very very
308. 9:55laborious process of creating very
309. 9:57specialized scientific code to do
310. 10:00scientific experiments. And so just I
311. 10:02think two weeks ago,
312. 10:04uh, the CDC officially named forecasts
313. 10:07that our team had been making the most
314. 10:09accurate for forecasting flu
315. 10:11hospitalizations for the 2025-2026 flu
316. 10:14season. Um which is really gratifying
317. 10:16because hospitalizations from flu are
318. 10:18important and forecasting in advance is
319. 10:21what like hospitals and public health
320. 10:23departments use to get in front of
321. 10:26expected capacity needs, to mobilize
322. 10:29resources, to to do vaccination
323. 10:32campaigns. It's It's a It's a
324. 10:33consequential application. But here's
325. 10:35what's really cool about it is that the
326. 10:37work from Google took two people and
327. 10:41they generated 200,000
328. 10:44different experiments to ultimately get
329. 10:47that highest accuracy. Um two people,
330. 10:50Zara Shamsi and Sarah Martinson. Sarah
331. 10:52Martinson's a graduate student. Neither
332. 10:54of them is an epidemiologist. So we see
333. 10:57here
334. 10:58just how powerful at enabling this
335. 11:02acceleration of capabilities, smaller
336. 11:05teams are so much more productive.
337. 11:08>> So I have a follow-up question for that.
338. 11:10So it sounds like AI is definitely
339. 11:12helping us. So is it that it's helping
340. 11:15us work faster, discover things faster
341. 11:18or is it enabling us from doing
342. 11:21something completely different that we
343. 11:22couldn't have done before without AI?
344. 11:24>> Yeah. So we're we are absolutely seeing
345. 11:28dramatic acceleration of the things that
346. 11:31we have already known how to do.
347. 11:34Full stop. And I think that there is
348. 11:36actually a huge amount of opportunity
349. 11:38for discovery in the things that we knew
350. 11:40how to do but we just didn't have the
351. 11:44time or the resources to do it. So those
352. 11:46200,000 different experiments if we
353. 11:49weren't using this type of AI system
354. 11:51that was helping with coding um
355. 11:54you might have been able to do
356. 11:5610
357. 11:57over the course of a matter of months
358. 11:59because it would take you a week or two
359. 12:01to do any given one. So doing 200,000
360. 12:04just would have been infeasible. But AI
361. 12:06lets us parallelize. It lets actually
362. 12:08Zara said this quote to me which I
363. 12:10thought was She was was "I tell the
364. 12:10computer what to do and then I go to
365. 12:12sleep And then I wake up in the morning
366. 12:14and I see results.
367. 12:15>> It's like having a thousand research
368. 12:16students for working for you.
369. 12:18>> Exactly. I think that's a perfect
370. 12:19analogy, actually. This idea of AI being
371. 12:21able to let us step outside of the
372. 12:24regime of science that we even could
373. 12:27contemplate how to do. I think that this
374. 12:29is a little bit more, we will see how
375. 12:32this plays out. Some people get very
376. 12:34excited about that. I personally just
377. 12:36look at the total pace of discovery and
378. 12:37I think that there's there's so much to
379. 12:39learn just by speeding up some of the
380. 12:40things that we already knew how to do.
381. 12:42>> That's really exciting. I was a software
382. 12:43engineering manager at Meta, so for me,
383. 12:46when I'm coding with AI, it feels like
384. 12:48I'm managing a hundred really brilliant
385. 12:50engineers, and they're working for me
386. 12:52while I sleep.
387. 12:53>> Yes. So, it's very reliable analogy.
388. 12:55Yeah.
389. 12:55>> So, AI is making all this progress. Do
390. 12:58you think that will make scientists more
391. 12:59valuable, or are we going to need fewer
392. 13:01of them?
393. 13:02>> Oh, I think we're going to need more of
394. 13:03them. I think scientists are heroes, and
395. 13:06my career has been focused on helping
396. 13:09them and enabling them, and and, you
397. 13:11know, my my goal uh for the near future
398. 13:13is like bringing them joy as much as
399. 13:16possible as we go about this
400. 13:18transformation of some of the tools and
401. 13:20ways of working that's being brought
402. 13:22about by AI. But, we will always need
403. 13:25scientists, and what's most important
404. 13:27for me is that people who might be the
405. 13:30next generation of scientists, so maybe
406. 13:32they're high school students today, or
407. 13:34they're in college today, are able to
408. 13:37appreciate that we need more graduate
409. 13:39students and more scientists, not fewer.
410. 13:41>> That reminds me of this um paradox. When
411. 13:44they discovered uh steam engines, they
412. 13:47thought, "Oh, we're not going to need
413. 13:48more coal anymore." But, they actually
414. 13:50needed a lot more because they learned
415. 13:52that people ride the trains a lot more.
416. 13:54Yes. So, similarly, AI is making our
417. 13:58scientific uh discoveries much faster,
418. 14:00but it doesn't eliminate scientists. It
419. 14:02makes scientists even more valuable
420. 14:04because we want to just do more
421. 14:06research.
422. 14:06>> Absolutely. The We have the We've been
423. 14:08really really constrained in just how
424. 14:12much research it's possible to do.
425. 14:13You're constrained by your time. So,
426. 14:16this acceleration and the ability for
427. 14:18certain things I use the term like
428. 14:20drudgery and schlep. Like scientists,
429. 14:23many of them, that that's a that's a
430. 14:24broad term that encompasses a lot of
431. 14:26different specializations and the work
432. 14:27can look different, but there's like a
433. 14:28lot of like debugging of code. There's a
434. 14:30lot of like pouring over old
435. 14:33documentation. There's a lot of like
436. 14:35reading an old paper and trying to
437. 14:37figure out if it's possible to reproduce
438. 14:39their work. That's what I as a product
439. 14:41manager think about is like what do you
440. 14:43want to
441. 14:44not have to spend as much time doing?
442. 14:46And if you had more time, what would you
443. 14:48want to spend more of your time doing?
444. 14:50Um and so, I really think about AI as
445. 14:53this complimentary uh system that
446. 14:56enables scientists to achieve more.
447. 14:58>> I love that framing. I think it can
448. 15:00apply to a lot of different fields. AI
449. 15:03is enabling us to do a lot more work,
450. 15:06not taking away our ability to do our
451. 15:09jobs. We're going to play a little game.
452. 15:11I have these six topics ready for you. I
453. 15:15have them [music] in a random order. You
454. 15:17will rank them in the order that you
455. 15:20think is most likely, like biggest
456. 15:22impact on our everyday lives to the
457. 15:25least impact.
458. 15:26>> Okay.
459. 15:29Okay, that's my order.
460. 15:30>> This is interesting. Um
461. 15:33Tell us more. You said, okay, first
462. 15:35ranking was medicine, number two is
463. 15:38climate, energy, robotics, the brain, we
464. 15:42talked a lot about brain, and then
465. 15:44space.
466. 15:44>> Yes.
467. 15:45>> So, tell us more. Why is medicine the
468. 15:47biggest impact?
469. 15:48>> to make clear that my ordering was
470. 15:51particularly influenced by use You said
471. 15:53on our everyday lives.
472. 15:55>> Yes.
473. 15:56>> So, this is for everyday people, what
474. 15:59they might see and appreciate and
475. 16:01understand about where AI progress is
476. 16:03happening. So, that doesn't mean that
477. 16:04we're not going to see major AI progress
478. 16:06in all of these different spaces. I
479. 16:08think that medicine is actually head and
480. 16:10shoulders above the other five items on
481. 16:13this list. Um there's so much
482. 16:15opportunity
483. 16:17uh for individuals and patients and
484. 16:19families and also healthcare providers.
485. 16:22I know I can speak from personal example
486. 16:24that I use Gemini and AI to ask a lot of
487. 16:28questions about health.
488. 16:30I, you know, for better or worse there
489. 16:31have been health issues going on in my
490. 16:33own family and
491. 16:35I I come from a family of doctors, so I
492. 16:37have people to ask too, but to be able
493. 16:39to ask very specific questions. And yes,
494. 16:42I will go and I will read the, you know,
495. 16:45the the website from
496. 16:47a Cleveland Clinic or other reputable
497. 16:49sources, um but to be able to ask this
498. 16:51is exactly what's going on for me or
499. 16:53tell me exactly what this medication
500. 16:56is used to treat for or what is this set
501. 16:59of symptoms or what are the what are the
502. 17:01risks, what are the warning signs that I
503. 17:03should watch out for of this is maybe
504. 17:04something I should call my doctor about.
505. 17:07I use AI extensively to help support
506. 17:10this and its accessibility, its
507. 17:12immediacy,
508. 17:13how personalized it is for me, I think
509. 17:16this is representative of a lot of
510. 17:17users. But we're also, you know, seeing
511. 17:20a lot of innovation in AI. It is how we
512. 17:22will build the capacity to be able to
513. 17:25support more of the needs that patients
514. 17:27have in triaging and in diagnosis and in
515. 17:29follow-up care. And so it's a really,
516. 17:31really exciting space where lots of
517. 17:33people are doing lots of great work.
518. 17:34We've been doing health-related work
519. 17:36with AI at Google for most of the
520. 17:38history of the company, actually. So,
521. 17:40some of the other ones, oh, did you have
522. 17:42a follow-up question?
523. 17:42>> Yeah, I was just curious,
524. 17:44do you have any that you can share like
525. 17:46about health?
526. 17:47>> So, there is really fantastic work that
527. 17:52we know we have been doing in
528. 17:55different types of medical imaging for
529. 17:57example. So, so the ability to say if I
530. 18:01got a picture of my eye that was taken
531. 18:04at the ophthalmologist,
532. 18:05if I had a chest x-ray that was done, if
533. 18:09I had a mammography that was done, it's
534. 18:12increasingly the case that there is
535. 18:13assistance that comes from AI in helping
536. 18:17a doctor to determine is there something
537. 18:19here that might be problematic that is
538. 18:22worth looking at. Um and this is good
539. 18:23because because those are important um
540. 18:27and uh sort of time-consuming. Like
541. 18:29there's long long long queues. So that's
542. 18:31an application in medicine where we've
543. 18:32been seeing AI uh effective for a long
544. 18:34time. Long enough that not only can we
545. 18:38do the research question of hey, can an
546. 18:40AI read this? But we've been able to
547. 18:42also start to do the clinical studies
548. 18:44where you say okay, not only from a I
549. 18:46wrote a paper about it, but can we show
550. 18:48in practice that this is able to help
551. 18:51improve the overall accuracy. We're able
552. 18:54to put this into into a delivery system.
553. 18:57Um it's not just a research concept.
554. 18:59It's real-world impact. So in the
555. 19:01medical imaging space, we're actually
556. 19:02pretty far along. But I also think
557. 19:04there's a lot of really cool stuff
558. 19:06that's been happening in more um sort of
559. 19:08consumer-facing. We had a we had a paper
560. 19:11um that came out in The Lancet actually
561. 19:13in the last week that was focused on
562. 19:15conversational medical AI uh and a tool
563. 19:18called Amie. Um and I think we're going
564. 19:20to see be seeing a lot more of this. And
565. 19:22I want to emphasize here from my
566. 19:24perspective, like people are still at
567. 19:26the core of this. The patient and the
568. 19:29doctor and how AI can be helping both of
569. 19:32them to understand more, to do more, um
570. 19:37and to and to you know, get the right
571. 19:38decision, get access to the right care,
572. 19:41and hopefully get a get a better outcome
573. 19:43um the first time.
574. 19:44>> That's that's really interesting because
575. 19:45I can relate because when I was talking
576. 19:48to the doctors about my dad's cancer, I
577. 19:50could read up from AI. AI was educating
578. 19:53me so I can have a better conversation
579. 19:55with the doctors. So
580. 19:57>> that's a really really important thing
581. 19:58is is that preparation. I have this I
582. 20:01have precious time. Doctors are scarce
583. 20:03sometimes um or you have a limited
584. 20:05amount of time with them.
585. 20:06And I always make a list of questions.
586. 20:08And I feel like I've also I've also done
587. 20:09this exact same thing you did or I've
588. 20:11helped my husband do the exact same
589. 20:12thing you did, which is like make sure
590. 20:14you ask them these five things. But it
591. 20:15feels like it's very informed. You can
592. 20:16get them you can get you can ensure that
593. 20:18you understand the questions that you
594. 20:20want to be asking.
595. 20:22>> Tell us more about robotics in your
596. 20:23household.
597. 20:24>> Well, I like many people have a robot
598. 20:26vacuum cleaner, um but I also have a
599. 20:28robot arm,
600. 20:29uh a cobot, and it's my husband's, has a
601. 20:31name, his name is J Robot Choppenheimer.
602. 20:34Um and uh and he loves it and uh we
603. 20:38definitely sometimes control it with AI.
604. 20:40Uh and one day, like, could we make it
605. 20:42make a latte for us and make latte art
606. 20:44on top of that latte? Or could it make
607. 20:47um you know, a a mocktail and mix
608. 20:49different things together? Or we do lots
609. 20:51of fun stuff. Actually, my nephew, I
610. 20:52have a nephew who is 12. We He came over
611. 20:55uh and he's learning to code, and so we
612. 20:58made the robot arm do spirograph art. Uh
613. 21:00which was really cool uh that he was
614. 21:02sort of learning about and what was
615. 21:04great was that he said please and thank
616. 21:06you um to all the chatbots that he was
617. 21:08having help him with his coding. But,
618. 21:10the the opportunities for creativity, I
619. 21:12think, are kind of endless. And so, when
620. 21:14you have people who are uh in
621. 21:16really enjoy robots as my household
622. 21:18does.
623. 21:18>> So, what what are the robot arms like?
624. 21:21Do you have two arms that you can
625. 21:22control?
626. 21:23>> Um we only have one.
627. 21:25>> [laughter]
628. 21:25>> We'd love to get two.
629. 21:28Um yes, and so they're they're mounted
630. 21:29on a base, and they come with sort of
631. 21:31different um you know, there's sort of
632. 21:33different numbers of joints that allow
633. 21:34them to have different ranges of motion.
634. 21:36You can get different attachments to the
635. 21:37end of them. But, what was most recent
636. 21:40was to actually control the robot arm
637. 21:43with a chatbot to say, "Oh, uh it was it
638. 21:45was it was actually holding It's holding
639. 21:47a kitchen utensil at the time.
640. 21:49>> [laughter]
641. 21:49>> That's why it's called Choppen
642. 21:50>> Chop.
643. 21:51>> Choppenheimer.
644. 21:52>> Choppenheimer.
645. 21:52>> Exactly. Um but, could you make it
646. 21:54dance? I think that these are these are
647. 21:55opportunities for for sort of fun and
648. 21:57creativity and exploration and play.
649. 21:59>> You know, uh that actually reminds me,
650. 22:00my friend recently hired her first robot
651. 22:04cleaner.
652. 22:05>> Tell me more.
653. 22:06>> Um it was teleoperated. She said it did
654. 22:10a really good job cleaning the toilet.
655. 22:12>> That's unexpected. I I can't wait for
656. 22:14robots to
657. 22:16do all the cleaning for me.
658. 22:17>> I think you're not alone.
659. 22:18>> Among the list, is there a topic do you
660. 22:21think people are getting too excited
661. 22:23about too early?
662. 22:24>> Well, I actually think that medicine is
663. 22:27both where we're going to see the most
664. 22:29progress and also where we see the most
665. 22:31hype.
666. 22:31>> [laughter]
667. 22:32>> So, so this is where I just I think that
668. 22:34the details matter. It's really good
669. 22:37that we have like a good understanding
670. 22:39of ways that people definitively support
671. 22:43and believe that improvement would be
672. 22:45welcome and helpful and is necessary.
673. 22:47But I think that we have to be careful
674. 22:49when we're talking about people who are
675. 22:51sick about you know, not misrepresenting
676. 22:56where we do see results and or see a
677. 22:59path forward and and have high
678. 23:00confidence that in 1 2 5 years, this is
679. 23:04where we'll be.
680. 23:05>> That makes a lot of sense. It's
681. 23:06literally life-changing, so it's an
682. 23:08important topic. So, we talked about a
683. 23:10lot of really exciting opportunities for
684. 23:13the future. Now, if you were I know
685. 23:15you've been working in tech for 20
686. 23:16years, but if you go back in time and
687. 23:19you were 22 again starting out your
688. 23:21career, what are some things that you
689. 23:23would study?
690. 23:24>> Um so I actually went to graduate school
691. 23:26after I had been in industry for 7
692. 23:29years, which was a little bit unusual.
693. 23:31And so when I went to graduate school, I
694. 23:34was so excited to go back and learn.
695. 23:36And I'd learned a little bit about
696. 23:37myself and how I learned, but I also was
697. 23:40was really I wanted to not just pass the
698. 23:44tests and get the degree. I really
699. 23:45wanted to learn the material
700. 23:48that would give me that foundation. You
701. 23:50used that word before and I think it's
702. 23:51perfect. The foundation to be able to
703. 23:54open up so many new pathways for me. And
704. 23:56I actually believe that there's a lot of
705. 23:59things to study that can be that
706. 24:01foundation.
707. 24:02And I and I think a sense of what am I
708. 24:05intrinsically
709. 24:07motivated to learn about and would enjoy
710. 24:10because
711. 24:11for me at least,
712. 24:12it is an order of magnitude more work to
713. 24:14learn about something I don't care about
714. 24:16than something that I enjoy, which which
715. 24:18has this ease associated with it. And if
716. 24:21that lets you
717. 24:23learn how you learn, have some set of
718. 24:26experience, Um, and then I think the
719. 24:29second thing is try to make something. I
720. 24:31think that that always serves you well.
721. 24:33Doesn't really even matter what, but to
722. 24:35build something, to create something, to
723. 24:38demystify what the process looks like.
724. 24:41You're going to get some hands-on
725. 24:42experience, uh, and it's going to, you
726. 24:44know, be something that you can tell
727. 24:46other people about as you're trying to
728. 24:47get a new opportunity for yourself. And
729. 24:49you're going to, you know, you have more
730. 24:51information for yourself about I think
731. 24:53that this is something I would like to
732. 24:54spend more of my career What do I want
733. 24:56to be when I grow up question? I thought
734. 24:58I wanted to be a doctor. It turns out
735. 25:00that what I wanted to be didn't exist
736. 25:02when I was a kid. Um, and so just
737. 25:04continuing to try stuff and to be a
738. 25:06lifelong learner, um, for me has, I
739. 25:09think, been great advice that has served
740. 25:11me well and it's what I would encourage
741. 25:12today's 22-year-olds to consider.
742. 25:14>> I love that. We didn't have cell phones
743. 25:17like when I was young, there were no
744. 25:18cell phones.
745. 25:19>> We had pagers.
746. 25:20>> [laughter]
747. 25:21>> Right? So, there the world is constantly
748. 25:24evolving and changing, but that doesn't
749. 25:27mean um, we have to be lost. We we can
750. 25:30try different things and evolve with it.
751. 25:32I love that message. What do you think
752. 25:34are skills that matter more with AI over
753. 25:37time?
754. 25:38>> I think that I was reflecting on my own
755. 25:41career and then, um, a conversation that
756. 25:43actually happened earlier this week at
757. 25:45the research at conference, which is my
758. 25:47career has been at the frontier of sort
759. 25:49of what was technologically possible and
760. 25:51and what that meant for science. And,
761. 25:54um, that meant that I was not becoming
762. 25:56an expert in one kind of technology. I
763. 25:58was having to evolve. And so, I do think
764. 26:01that that ability to be a learner, I'm
765. 26:03slightly repeating my last answer, um,
766. 26:05is is a really important skill. And, uh,
767. 26:09and that means kind of letting yourself
768. 26:11be a little bit uncomfortable because
769. 26:13you're going to be in the in the
770. 26:14learning phase, right? The, um,
771. 26:17consciously incompetent is is a term
772. 26:19that we use for parts of the growth
773. 26:21curve. Um, but the willingness to let
774. 26:23yourself be there and be motivated to
775. 26:25get out of it propels you forward. In
776. 26:27terms of the the specific skill, um,
777. 26:30there was a term that I'm not going to
778. 26:31take credit for. This came from uh,
779. 26:33Sinead Bovell who is a futurist and she
780. 26:36said, "You know, you know, we have
781. 26:36optometrists, like we have, you know,
782. 26:39neurologists, [snorts]
783. 26:40maybe we're going to have an
784. 26:41AI-ologist." And
785. 26:44you know, I don't know what that means
786. 26:45in an exact sense, but I do believe
787. 26:48we're always going to need people at the
788. 26:51in at the connection point between
789. 26:53technologies and how to turn those
790. 26:54technologies into products that help and
791. 26:58are beneficial and are valuable. And I
792. 27:00think that's a really exciting and and
793. 27:02valuable skill.
794. 27:03>> Maybe we'll have some cool job titles
795. 27:05coming up in the next few years.
796. 27:07>> Absolutely.
797. 27:08>> Yeah. If you were to fill in the blank
798. 27:09for this sentence, by 2031 AI will have
799. 27:15something. What what would it be?
800. 27:16>> I both hope and believe that AI, before
801. 27:212031, uh, will have dramatically reduced
802. 27:24the amount of technological drudgery
803. 27:29that everyday people have to experience
804. 27:32and it enables them to um, spend much
805. 27:36more time on the things that they enjoy
806. 27:37and it opens up new opportunities for
807. 27:39them in their lives.
808. 27:40>> That's great. I'm going to watch this
809. 27:41video again in 2031.
810. 27:43>> [laughter]
811. 27:44>> And you can send me my scorecard.
812. 27:46>> Well, thank you for your time. It was
813. 27:48great having you.
814. 27:49>> This was such a pleasure. I really
815. 27:50appreciate it.
816. 27:51>> Thanks.
817. 27:52>> Thank you.