WEBVTT - Artificial Intelligence: The Future is Now 0:00:02.240 --> 0:00:04.960 This is Alec Baldwin and you're listening to Hear's the 0:00:05.000 --> 0:00:10.360 thing from iHeartRadio. Modern day society has been living in 0:00:10.360 --> 0:00:14.120 some form of the computer era for decades. An entire 0:00:14.160 --> 0:00:17.119 genre of films has been dedicated to showing what human 0:00:17.160 --> 0:00:21.400 life may be like once computers become conscious. My guests 0:00:21.440 --> 0:00:24.439 today are two trailblazers in a field that many have 0:00:24.520 --> 0:00:28.360 speculated could potentially risk the end of human civilization as 0:00:28.400 --> 0:00:31.680 we know it. That is, of course, the true birth 0:00:31.680 --> 0:00:35.880 of artificial intelligence, known simply as AI. And if you 0:00:35.960 --> 0:00:39.320 think the world has become inundated with AI technology overnight, 0:00:39.760 --> 0:00:43.200 you are not alone. AI based products come in many 0:00:43.240 --> 0:00:46.559 different forms, such as chatbots, which can be used for 0:00:46.640 --> 0:00:50.040 everything from finding a list of local Italian restaurants to 0:00:50.120 --> 0:00:54.840 drafting legal documents and writing college essays. AI can also 0:00:54.880 --> 0:00:57.720 be used to manipulate a person's likeness on video or 0:00:57.760 --> 0:01:00.720 the sound of their voice. And if you're wondering just 0:01:00.800 --> 0:01:04.880 how good computers have become at mimicking humans by narration 0:01:05.040 --> 0:01:07.720 thus far has been generated by text written into a 0:01:07.800 --> 0:01:13.360 program called descript. Okay, I can't let the machines take 0:01:13.400 --> 0:01:18.400 over just yet. This is actually human. Alec Baldwin. Later 0:01:18.440 --> 0:01:22.119 in the program, I'll speak with Blake Lemoyne, a software 0:01:22.160 --> 0:01:26.480 engineer formerly employed at Google. Lemoyne was part of the 0:01:26.520 --> 0:01:31.520 team working on Lambda, or language model for dialogue applications, 0:01:31.840 --> 0:01:36.360 the technology behind chatbots. He went public with claims that 0:01:36.440 --> 0:01:40.000 the AI he was working on was sentient and was 0:01:40.080 --> 0:01:44.920 fired shortly thereafter. But first I'm talking to Jay Lebouffe, 0:01:45.400 --> 0:01:49.840 head of business and corporate development at descript Dscript is 0:01:49.920 --> 0:01:54.720 an editing program used by podcasters and vloggers. It allows 0:01:54.880 --> 0:01:59.360 users to edit audio directly from an AI generated transcript, 0:01:59.600 --> 0:02:03.400 and can even use voice cloning technology to create new 0:02:03.440 --> 0:02:06.280 audio from text, just as we did in the opening 0:02:06.280 --> 0:02:10.200 of the show. Lebuff has worked in technology his entire career. 0:02:10.720 --> 0:02:14.320 I was curious how long AI has really been utilized 0:02:14.680 --> 0:02:15.400 in his field. 0:02:16.440 --> 0:02:19.000 The term has been around for a very long time, 0:02:19.520 --> 0:02:21.600 but there's this period that a lot of people call 0:02:21.639 --> 0:02:25.360 the AI winter where it was much hyped in the 0:02:25.440 --> 0:02:27.320 eighties and even going into the nineties, and it just 0:02:27.360 --> 0:02:30.840 had never delivered. So I didn't really know anybody who's 0:02:30.840 --> 0:02:32.680 studying AI at that time. 0:02:33.160 --> 0:02:36.200 Were you aware of it before you went to Cornell? No? No, 0:02:36.680 --> 0:02:38.800 you weren't You were introduced to all this when you 0:02:38.800 --> 0:02:39.480 went to college. 0:02:39.600 --> 0:02:41.880 Absolutely, that is the value of college education for me 0:02:41.960 --> 0:02:44.959 is I was exposed to a lot. I wasn't told 0:02:45.000 --> 0:02:46.359 what to do with it, but I was exposed to 0:02:46.400 --> 0:02:48.200 a lot of concepts that later on turned out to 0:02:48.200 --> 0:02:49.200 be valuable. 0:02:49.400 --> 0:02:52.400 It just occurred to me, Jay, are we really talking 0:02:52.520 --> 0:02:55.280 to you? Are you here with us? 0:02:55.560 --> 0:02:57.800 What I'd suggest is you ask me a few questions 0:02:58.280 --> 0:03:00.240 and then we do a little touring test on it, 0:03:00.280 --> 0:03:02.320 which is a bit of a gold standard. 0:03:02.320 --> 0:03:04.880 It just occurred to me, maybe you're like the beach 0:03:04.919 --> 0:03:07.119 in Aruba and this is not you. 0:03:07.160 --> 0:03:10.400 No, but this is such a fascinating topic, Alec, because 0:03:11.280 --> 0:03:16.920 we are entering a phase where anything can be synthetically generated. 0:03:17.040 --> 0:03:21.720 So my voice can be synthetically generated and indistinguishable from 0:03:21.760 --> 0:03:24.800 real me, even what I'm saying and the terms I'm 0:03:24.840 --> 0:03:27.920 using in my speaking style, that even can be generated 0:03:28.000 --> 0:03:30.400 all through this generative AI technology. 0:03:31.480 --> 0:03:33.840 Well, I know that the work you do has nothing 0:03:33.880 --> 0:03:35.640 to do with like deep thinks and that kind of thing. 0:03:35.920 --> 0:03:38.800 But what I find, for example, is, and again this 0:03:38.840 --> 0:03:40.320 has nothing to do with your work, but like if 0:03:40.320 --> 0:03:43.400 I see a deep fake. You just mentioned about content 0:03:43.440 --> 0:03:46.560 when I see a deep fake, I know it's bullshit 0:03:46.720 --> 0:03:52.000 because no matter how much facially the animation is precise, 0:03:52.760 --> 0:03:55.120 I'm obviously going to talk about Tom Cruise, the Tom 0:03:55.160 --> 0:03:58.880 Cruise deep thakes and the voice and the cadences, but 0:03:58.960 --> 0:04:01.720 what's missing is the cut. You need writers to write. 0:04:02.120 --> 0:04:04.800 I'm assuming what Cruise would say, because when I watch 0:04:04.840 --> 0:04:07.880 these deep things, they almost never say what Cruise would say. 0:04:08.400 --> 0:04:12.640 Do you believe that AI has that capability to manufacture 0:04:12.680 --> 0:04:15.720 the thoughts of the person that they're recreating. 0:04:16.680 --> 0:04:19.120 So the state of the art where we are right 0:04:19.160 --> 0:04:25.920 now would be AI can regurgitate and recombine past things 0:04:26.000 --> 0:04:28.400 that you might have said or that might have been 0:04:28.400 --> 0:04:31.760 attributed to you, And so it might seem like it's 0:04:31.800 --> 0:04:35.599 a seemingly novel output, a new Tom Cruisi and expression, 0:04:35.960 --> 0:04:38.760 but it's actually just recombining things that he might have 0:04:38.839 --> 0:04:42.040 said and then what you pointed out. It's also introducing 0:04:42.120 --> 0:04:46.400 some hallucinations things that know that absolutely does not pass 0:04:46.440 --> 0:04:48.680 the smell test right now, and that's where humans in 0:04:48.720 --> 0:04:51.920 the loop are always going to kind of prevail. 0:04:52.080 --> 0:04:54.560 One thought we had was to open this episode with 0:04:54.680 --> 0:04:58.159 me introducing my Guest Academy Award winner, one of the 0:04:58.160 --> 0:05:01.000 greatest film stars of his general join me from my 0:05:01.080 --> 0:05:04.480 conversation with Laurence Olivier. So we all know I'm not 0:05:04.600 --> 0:05:08.040 interviewing Lawrence Olivier because he's dead. Do you see that 0:05:08.040 --> 0:05:09.279 that's possible. 0:05:09.560 --> 0:05:12.160 With our platform? It is not possible to clone the 0:05:12.240 --> 0:05:15.240 voice of the deceased. It's also not possible to clone 0:05:15.400 --> 0:05:18.800 a voice that is not your own voice without your consent. 0:05:19.680 --> 0:05:21.520 But I'm saying technologically it's possible. 0:05:21.520 --> 0:05:25.120 Technologically as possible. You are completely right. There will always 0:05:25.200 --> 0:05:29.800 be tools, technology and open source that allow anyone to 0:05:29.839 --> 0:05:33.120 do anything, but a platform like what we run, we've 0:05:33.160 --> 0:05:36.599 actually made a very firm ethical stance that we believe 0:05:36.920 --> 0:05:39.279 each person should have the rights to their voice and 0:05:39.360 --> 0:05:41.560 control when it is used. And that's why we have 0:05:41.600 --> 0:05:45.480 some pretty strict authentication and also ethical steps set up 0:05:45.480 --> 0:05:48.560 so you couldn't use our platform to do that interview unfortunately. 0:05:49.040 --> 0:05:51.240 Now what are you working on? Now? What kinds of 0:05:51.320 --> 0:05:53.000 other projects are you working on now? 0:05:53.560 --> 0:05:56.800 So a big thing we're continuing to improve upon is 0:05:57.400 --> 0:06:00.640 this voice cloning technology to allow people to fix their 0:06:00.640 --> 0:06:03.919 mistakes and generate new content. We have a like a 0:06:04.040 --> 0:06:08.760 Hollywood style green screen removal, so in that situation. That's 0:06:08.800 --> 0:06:11.200 where maybe I'm doing this interview and I don't like 0:06:11.240 --> 0:06:14.560 the background behind me. With one click it can disappear. 0:06:14.640 --> 0:06:18.520 I can spot in a more professional looking background. That's 0:06:18.520 --> 0:06:19.760 something that's also giving people, well. 0:06:19.760 --> 0:06:21.520 They have those background things on zoom. 0:06:21.200 --> 0:06:25.320 Correct exactly like like zoom but you know Hollywood grade 0:06:25.400 --> 0:06:27.839 at that point, Yeah, you know, It's really about how 0:06:27.839 --> 0:06:32.279 do we help creators best tell their story through finding content, 0:06:32.360 --> 0:06:37.040 discovering content, and getting their videos as polished as possible, 0:06:37.080 --> 0:06:39.920 but as quickly as possible. Because that's that's a trend. 0:06:39.960 --> 0:06:42.279 We just see people feel a sense to put up 0:06:42.360 --> 0:06:45.320 more content, but we're hoping we can help them actually 0:06:45.320 --> 0:06:48.920 just create better content and really hone their craft. 0:06:49.279 --> 0:06:52.360 I think of where the venue is a performance venue 0:06:53.040 --> 0:06:57.000 as opposed to an educational venue. So, for example, if 0:06:57.200 --> 0:07:00.680 someone is doing an audio book and the book is 0:07:00.720 --> 0:07:05.200 about something rather dry. There are people reading textbooks, and 0:07:05.279 --> 0:07:09.120 I don't see anything wrong with them using this technology 0:07:09.720 --> 0:07:12.280 to get a copy of their voice, to run it 0:07:12.320 --> 0:07:16.760 through the grinder and do the dubbing of that person's 0:07:16.840 --> 0:07:22.720 voice reading that text. But this hearkens to the idea 0:07:22.760 --> 0:07:26.160 of a performer giving a performance they never gave. Someone 0:07:26.200 --> 0:07:27.760 said to me, pretty soon, you're going to be able 0:07:27.760 --> 0:07:31.560 to do a movie with Humphrey Bogart. And through this technology, 0:07:31.600 --> 0:07:34.480 the very technology you and I are referencing today, they're 0:07:34.480 --> 0:07:36.360 going to give a performance with Humphrey Bogart. 0:07:37.160 --> 0:07:38.840 Whether we want it to be here or not, that 0:07:39.080 --> 0:07:43.160 it's inevitable. The technology is here. The lawyers are sorting 0:07:43.160 --> 0:07:46.840 out the publicity and the copyright rights. I mean, James 0:07:46.880 --> 0:07:50.480 old Jones has licensed his voice to live on for 0:07:51.080 --> 0:07:54.320 future Darth Vader times when it needs to be used. 0:07:54.640 --> 0:07:56.080 One of the things I do a lot of of 0:07:56.200 --> 0:07:58.680 alec is I do a lot of university teaching, and 0:07:59.120 --> 0:08:01.920 this is the best place to keep me grounded about 0:08:01.960 --> 0:08:06.680 what the next generation of our industry actually thinks. And 0:08:07.160 --> 0:08:08.640 as I was talking to them about this over the 0:08:08.640 --> 0:08:12.320 past few weeks, they are very excited for these tools, 0:08:12.400 --> 0:08:15.200 like they'll try out ten tools in a single day 0:08:15.640 --> 0:08:18.640 because they have no fears about them whatsoever. And when 0:08:18.640 --> 0:08:21.240 I ask them about like, well, aren't you concerned, and 0:08:21.280 --> 0:08:23.480 they're like, no, this is great, Like this this is 0:08:23.520 --> 0:08:26.240 how the world is going to be for them. This 0:08:26.400 --> 0:08:29.200 is just the new state of the normal. And their 0:08:29.280 --> 0:08:31.600 job is to understand the world that they're going into, 0:08:32.120 --> 0:08:36.040 embrace the tools, and then find people who are resistant 0:08:36.080 --> 0:08:37.640 to the tools and kind of push them out. 0:08:37.559 --> 0:08:41.640 Of the way. Push them out of the way. How well, in. 0:08:42.040 --> 0:08:44.240 The typical way where for many years I taught this 0:08:44.679 --> 0:08:47.280 music business class at Stanford and on the first day 0:08:47.280 --> 0:08:50.680 of class the sheer number of students whose mission for 0:08:50.760 --> 0:08:54.440 taking this class was to disrupt the entire music industry, 0:08:54.720 --> 0:08:57.360 and they thought of a better model than Spotify and 0:08:57.400 --> 0:09:01.160 a better model than the major label system. And it's 0:09:01.200 --> 0:09:03.960 that just kind of overconfidence of no, no, no, no, 0:09:04.000 --> 0:09:06.040 we have a better idea. We're just going to you know, 0:09:06.320 --> 0:09:08.600 run fast and break things. We're going to do that. 0:09:08.960 --> 0:09:10.800 And then you know, of course, through the class we 0:09:10.880 --> 0:09:14.920 let them understand things like mchandical royalties and publishing and 0:09:15.000 --> 0:09:17.719 just help them understand how the sausage is actually made 0:09:17.800 --> 0:09:19.760 and why it is the way it is, and then 0:09:19.800 --> 0:09:21.800 they come with a more pragmatic view. So I think 0:09:22.120 --> 0:09:26.280 right now we have university students who are incredibly enthusiastic 0:09:26.400 --> 0:09:28.600 about these tools and are trying to figure out how 0:09:28.640 --> 0:09:31.280 they can use them to make themselves better when they 0:09:31.280 --> 0:09:34.120 go into industry and to make themselves more competitive and 0:09:34.200 --> 0:09:36.600 often when I hire people on my team that are 0:09:36.640 --> 0:09:39.480 fresh out of school, their knowledge of certain tools that 0:09:39.520 --> 0:09:42.000 I've never seen in my life is actually a competitive 0:09:42.000 --> 0:09:45.680 differentiation for them compared to the older people like me. 0:09:46.320 --> 0:09:49.280 Do you have any concerns about this technology? I mean, 0:09:49.360 --> 0:09:53.439 obviously there's some wonderful you know, Zach McNeice, who cuts 0:09:53.480 --> 0:09:58.880 our show, uses descript to edit the show on a transcript, 0:09:59.360 --> 0:10:01.240 And when I'm talking to him on the phone and 0:10:01.280 --> 0:10:03.480 telling him, I make my notes for a cut, and 0:10:03.520 --> 0:10:05.959 I'll call him and say, now at thirty eight minutes 0:10:06.000 --> 0:10:09.679 and ten seconds, I say this cut that. Now all 0:10:09.679 --> 0:10:11.240 I have to do is say to him, will I 0:10:11.360 --> 0:10:14.160 say this phrase? And he goes to the transcript and 0:10:14.160 --> 0:10:16.559 he finds it and we can do the cutting more efficiently. 0:10:16.800 --> 0:10:20.440 I mean, there is obviously very useful applications for this, 0:10:20.760 --> 0:10:23.360 and not only in our business but beyond. But was 0:10:23.400 --> 0:10:25.680 there anything that concerned you about this technology? 0:10:26.280 --> 0:10:29.000 I mean, one of the things I do worry about 0:10:29.000 --> 0:10:31.680 for creators that are coming in is this temptation to 0:10:31.760 --> 0:10:35.760 go for more, because if you can put out more 0:10:35.840 --> 0:10:40.160 content easier, then you're probably going to try to do that, 0:10:40.200 --> 0:10:41.719 thinking that that's how you're going to just you know, 0:10:41.720 --> 0:10:43.440 you're going to bombard the world with all your great 0:10:43.480 --> 0:10:47.480 ideas and your great stories. But what I really believe 0:10:47.480 --> 0:10:49.840 will happen is quality is going to just always continue 0:10:49.880 --> 0:10:51.880 to rise to the top, and so we're going to 0:10:51.920 --> 0:10:56.840 see a surge and just drown and synthetically generated media 0:10:57.120 --> 0:10:59.360 until we go back to more of a curation phase 0:10:59.720 --> 0:11:04.199 where things like the fact that you leave some authentic 0:11:04.360 --> 0:11:08.319 moments in your shows actually become valued. They become handprints 0:11:08.320 --> 0:11:11.719 and fingerprints in the work, and that is something that 0:11:12.120 --> 0:11:15.480 we will all value. And so when I do work 0:11:15.520 --> 0:11:18.320 with studios that are adopting this technology, my goal is 0:11:18.360 --> 0:11:20.800 to not help them put out nine new shows with 0:11:20.840 --> 0:11:22.600 the same team they have, but actually take that one 0:11:22.640 --> 0:11:25.160 show they're working on and make it even better and 0:11:25.280 --> 0:11:27.880 use the AI to come up with, you know, give 0:11:27.880 --> 0:11:30.480 me ten interview questions for when I'm interviewing this person 0:11:30.800 --> 0:11:33.040 and help them think of things they haven't thought of before, 0:11:33.559 --> 0:11:36.880 and help them, you know, summarize key talking points and 0:11:37.040 --> 0:11:39.599 just kind of be a virtual producer. 0:11:40.320 --> 0:11:43.800 I wonder if there are other applications of this that 0:11:44.000 --> 0:11:48.600 are similar, let's say to a three D printer. So 0:11:49.240 --> 0:11:54.920 do you see a technology where the information is put 0:11:55.000 --> 0:11:59.559 into the computer. This is obviously very complex, but that 0:11:59.640 --> 0:12:05.680 you have automated surgery and medical procedures where devices can 0:12:05.760 --> 0:12:09.920 perform common surgeries, basic surgeries at first, I mean not 0:12:10.040 --> 0:12:13.880 something that they're very complex. But if a myriad of 0:12:13.960 --> 0:12:16.319 information is fed into a machine, as it is into 0:12:16.320 --> 0:12:19.439 the brain of a neurosurgeon or a team of them, 0:12:19.840 --> 0:12:22.880 and you wind up having devices that will perform the 0:12:22.920 --> 0:12:25.640 surgery instead of a person with an eye toward at 0:12:25.640 --> 0:12:28.760 the very least, the goal being that it'll be done better, 0:12:29.080 --> 0:12:31.240 it'll be done more precisely. Do you see that as 0:12:31.240 --> 0:12:31.840 a potential. 0:12:32.920 --> 0:12:36.920 So I can't knowledgeably speak about the medical profession, but 0:12:37.600 --> 0:12:40.960 for most of the knowledge worker industries, like what a 0:12:41.000 --> 0:12:44.480 business professional, what a marketing person, and what a content 0:12:44.559 --> 0:12:48.760 creator would be doing, we're going to have these AI 0:12:49.040 --> 0:12:52.680 assistive tools complement all the work that we're doing. It's 0:12:52.800 --> 0:12:55.840 things like spellcheck, Like it went from being you have 0:12:55.880 --> 0:12:58.040 a dictionary on your desk, but you're too lazy to 0:12:58.120 --> 0:13:01.880 use it to it's monitoring every textbox we ever type. 0:13:02.000 --> 0:13:04.640 So really the only excuse for having a spelling mistake 0:13:04.720 --> 0:13:09.000 is just sheer laziness. So now that we have these 0:13:09.040 --> 0:13:13.000 tools that are always giving us suggestions, always helping us 0:13:13.000 --> 0:13:18.120 get unblocked, always helping us brainstorm. It's our job as producers, 0:13:18.160 --> 0:13:20.760 Like I often think of, like who wouldn't want to 0:13:20.800 --> 0:13:23.240 be a music producer? You just kind of shout vague 0:13:23.280 --> 0:13:26.480 terms into the room and the system knows what you 0:13:26.559 --> 0:13:30.480 mean by you know, louder Basier, no more like Coltrane 0:13:30.920 --> 0:13:34.160 and like you can work with the system at that level. 0:13:34.440 --> 0:13:37.760 And so I think, you know, with a medical analogy, 0:13:38.000 --> 0:13:41.320 you have something that's helping you with your diagnosis, helping 0:13:41.400 --> 0:13:43.320 you think through things you might not have thought of, 0:13:43.679 --> 0:13:46.559 giving you suggestions just to make you a better human. 0:13:47.280 --> 0:13:52.040 Now, the Anthony Boordnaining documentary, the road Runner documentary, got 0:13:52.080 --> 0:13:54.520 a little bit of flack because they had him, they 0:13:54.559 --> 0:13:57.959 synthesized his voice to read some of the copy. Yeah, 0:13:58.200 --> 0:14:01.160 is that something that concerned you? It did? It did. 0:14:01.240 --> 0:14:04.960 And that is just an example where I think it's 0:14:05.559 --> 0:14:09.000 really important for the companies that are creating these tools 0:14:09.679 --> 0:14:13.000 to take a stand on what is ethically allowed using 0:14:13.040 --> 0:14:17.640 their platforms, because there are no worldwide regulations on what 0:14:17.679 --> 0:14:20.360 you can and cannot do with this. So at the 0:14:20.440 --> 0:14:22.920 end of the day, it's really up till now, up 0:14:22.960 --> 0:14:26.120 to the tool manufacturers, the software companies whether to allow 0:14:26.160 --> 0:14:28.720 this or not. You know, in my case and Descript's case, 0:14:29.120 --> 0:14:31.240 we just don't allow this. We don't allow someone to 0:14:31.440 --> 0:14:34.720 clone the voice of the deceased because we feel like 0:14:34.760 --> 0:14:37.160 it's a really slippery slope. If we start allowing that, 0:14:37.320 --> 0:14:39.320 then you know, where does it go from here? 0:14:40.240 --> 0:14:43.120 When you're not working in this field, what does someone 0:14:43.200 --> 0:14:48.160 with your interests and your education and your career thus far, 0:14:48.640 --> 0:14:50.720 what do you do to relax and enjoy yourself. 0:14:51.960 --> 0:14:54.440 I am so exhausted right now, Alec, because we are 0:14:54.480 --> 0:14:57.320 sleep training. Are six month old and we have a 0:14:57.360 --> 0:14:59.560 two and a half year old as well. I love 0:14:59.600 --> 0:15:03.520 them both. They're incredibly exhausting, but it's just so fulfilling. 0:15:03.560 --> 0:15:06.560 So my hands are certainly filled with joy and love 0:15:06.640 --> 0:15:09.840 and exhaustion. With our two kids, my wife and I 0:15:09.880 --> 0:15:12.800 like to do a lot of just really authentic things 0:15:12.800 --> 0:15:17.080 that don't involve technology, like hike, go to there's a 0:15:17.080 --> 0:15:19.320 little beer garden near our house that we just like 0:15:19.960 --> 0:15:21.960 can sit at and let the kids play and talk 0:15:22.000 --> 0:15:24.480 to other people live that are not synthetic. 0:15:25.240 --> 0:15:29.080 You know, what I'm fascinated by is that children It's 0:15:29.320 --> 0:15:33.200 almost haunting to me how even the littlest ones, how 0:15:33.240 --> 0:15:38.120 transfixed they are by media and screens. But you see 0:15:38.440 --> 0:15:42.840 the way the human brain engages with this equipment and 0:15:42.960 --> 0:15:46.360 an even early age. And I'm wondering if somebody is 0:15:46.400 --> 0:15:49.640 able to develop technology where my kids are looking at 0:15:49.640 --> 0:15:53.520 the screen and what comes on that screen is something 0:15:53.560 --> 0:15:58.280 rendered through artificial intelligence that helps them to learn. You 0:15:58.360 --> 0:16:01.600 look at their programming, you're one their programming through a 0:16:01.640 --> 0:16:04.120 device which tells you what all the shows they have 0:16:04.200 --> 0:16:08.880 in common are, right, the colors and the pace and 0:16:08.960 --> 0:16:12.000 the music and the types of characters and the voices. 0:16:12.200 --> 0:16:15.960 And to develop a programming where you rend or something 0:16:16.400 --> 0:16:19.880 which has the best and the most popular of what 0:16:19.920 --> 0:16:22.520 they want, but at the same time conserve an educational purpose. 0:16:23.200 --> 0:16:26.680 So in this particular case, my two and a half 0:16:26.760 --> 0:16:31.200 year olds toddler Leo. He loves this YouTube channel called 0:16:31.200 --> 0:16:33.640 Songs for Littles, and the character is Miss Rachel and 0:16:33.680 --> 0:16:36.880 she's charming and very educational. It's great. He has this 0:16:37.040 --> 0:16:39.680 two minute long video it's called I'm So Happy, and 0:16:39.760 --> 0:16:41.600 it's like a music video, but it teaches you really 0:16:41.600 --> 0:16:43.840 good stuff. But after the four hundred and eleventh time, 0:16:43.840 --> 0:16:45.920 that I've seen it like I'm pulling my hair out, 0:16:45.960 --> 0:16:48.080 and I want him to get more out of it. 0:16:48.560 --> 0:16:52.760 So we actually personalized it for him. So I took 0:16:52.760 --> 0:16:56.000 the video, I used the technology we had. I took 0:16:56.040 --> 0:16:59.440 some video of him bouncing on a trampoline outside, removed 0:16:59.480 --> 0:17:02.960 the background and put him in the video and select places, 0:17:03.320 --> 0:17:06.560 and then we have her saying Leo's name every now 0:17:06.600 --> 0:17:12.119 and then. This unlocked the video for him and engaged 0:17:12.160 --> 0:17:13.920 his learning, because you know, part of it is about 0:17:13.960 --> 0:17:16.399 just teaching you emotions and what's okay, and so to 0:17:16.440 --> 0:17:19.800 put our child into a video this is like it 0:17:19.840 --> 0:17:21.639 turned out to be the hit of Christmas. Like whatever 0:17:21.640 --> 0:17:24.280 else we gave paled in comparison to this, like YouTube 0:17:24.320 --> 0:17:25.399 video that I made over lunch. 0:17:25.840 --> 0:17:26.879 That's amazing to me. 0:17:27.200 --> 0:17:30.880 I'm just super excited about this. I'm an optimist. So yes, 0:17:31.200 --> 0:17:34.159 we could also, you know, do bad things with the technology, 0:17:34.240 --> 0:17:38.520 but overwhelmingly, my children are growing up in a time 0:17:38.760 --> 0:17:42.040 where whatever they're watching on the screen, they could actually 0:17:42.080 --> 0:17:44.719 grab it and start recombining it and remixing it and 0:17:44.760 --> 0:17:47.439 inserting themselves in it. They could insert their parents in it. 0:17:47.760 --> 0:17:49.040 They could do anything with it. 0:17:49.920 --> 0:17:52.640 Thank you so much. For your time. Thank you so much. 0:17:52.680 --> 0:18:01.480 It's pleasure. Head of Business and Corporate Development Escript jle Both. 0:18:02.560 --> 0:18:06.520 If you're interested in conversations about our changing world, be 0:18:06.640 --> 0:18:09.560 sure to check out my episode on climate science with 0:18:09.720 --> 0:18:11.240 doctor Peter Domenical. 0:18:12.000 --> 0:18:15.400 Climate change is costing us now it's real dollars. I mean, 0:18:15.480 --> 0:18:18.399 last year was roughly three hundred billion dollars in climate 0:18:18.440 --> 0:18:22.399 and weather related damages. This year, with the California fires, 0:18:22.400 --> 0:18:24.600 it looks like it's maybe even more than that. So 0:18:24.720 --> 0:18:26.960 regardless of whether you believe in climate change or not, 0:18:26.960 --> 0:18:29.119 imagine you are in some deeply read state and a 0:18:29.200 --> 0:18:30.720 deeply read county in that state. 0:18:31.200 --> 0:18:32.360 You are paying for this. 0:18:32.720 --> 0:18:34.159 You may not like it, you don't want to call 0:18:34.200 --> 0:18:36.840 it climate change or whatever, but you for sure someone 0:18:36.880 --> 0:18:39.199 is paying that bill. 0:18:39.240 --> 0:18:42.920 Hear more of my conversation with doctor Peter Domenical at 0:18:42.960 --> 0:18:47.520 Here's Thething dot org after the break. My next guest, 0:18:47.720 --> 0:18:51.480 Blake Lemoyne, shares the process of being let go by 0:18:51.520 --> 0:19:04.920 Google and why he wanted to take a stand. I'm 0:19:04.920 --> 0:19:08.560 Alec Baldwin and you're listening to Here's the Thing. In 0:19:08.640 --> 0:19:13.920 twenty twenty two, Google senior software engineer Blake Lemoine distributed 0:19:13.960 --> 0:19:16.880 a document claiming that the AI he worked on was 0:19:17.240 --> 0:19:21.760 conscious and was fired soon thereafter. Lemoine has been speaking 0:19:21.800 --> 0:19:26.080 out about the need for greater transparency and accountability in 0:19:26.200 --> 0:19:31.280 Silicon Valley ever since. Lemoine is also a military veteran 0:19:31.400 --> 0:19:35.000 and an ordained priest. I wanted to know how he 0:19:35.080 --> 0:19:37.600 found his way to Google in the first place. 0:19:38.720 --> 0:19:41.399 So after I finished my masters, I had applied for 0:19:41.480 --> 0:19:43.760 some jobs with different companies that I would have gone 0:19:43.800 --> 0:19:47.920 to rather than pursuing a PhD. Google was one of them. 0:19:48.320 --> 0:19:52.159 I actually didn't get accepted that first round, so I 0:19:52.200 --> 0:19:55.600 went on started working on the PhD. But then about 0:19:55.640 --> 0:19:58.360 two years later, I got a call from Google out 0:19:58.359 --> 0:20:00.520 of the blue one day and they said, Hey, we're 0:20:00.520 --> 0:20:03.080 doing a big hiring push right now, and the last 0:20:03.080 --> 0:20:05.119 time you applied you almost made it. Would you like 0:20:05.160 --> 0:20:07.800 to come out and try again? So I was like, yeah, sure, 0:20:08.080 --> 0:20:11.200 So I went back to Mountain View interviewed again, and 0:20:11.400 --> 0:20:13.200 that second time through I got hired. 0:20:14.480 --> 0:20:17.119 So when you go there and you arrive at the 0:20:17.200 --> 0:20:21.119 Kremlin there in Mountain View, what's the goal, what do 0:20:21.200 --> 0:20:23.160 you want to start working on? What do they want 0:20:23.200 --> 0:20:24.040 you to start working on? 0:20:24.600 --> 0:20:26.760 So when you get a job, you get kind of 0:20:26.760 --> 0:20:30.119 put into a general purpose pool unless you're hired for 0:20:30.160 --> 0:20:34.119 a very specific reason. But just like most software engineers, 0:20:34.160 --> 0:20:37.320 I was interviewing with different teams at Google to figure 0:20:37.320 --> 0:20:39.800 out which one I wanted to start with. I interviewed 0:20:39.800 --> 0:20:42.399 with three teams. Two of them had to do with 0:20:42.640 --> 0:20:46.439 the kind of natural language processing and AI that I 0:20:46.560 --> 0:20:48.920 was interested in, so I picked one of those. At 0:20:48.920 --> 0:20:51.760 the time, the product was called Google. Now, the basic 0:20:51.840 --> 0:20:54.080 job that I had is for every single person on 0:20:54.119 --> 0:20:57.160 the planet predict what they're going to want to read tomorrow. 0:20:57.320 --> 0:21:00.200 Figure out some kind of way to do that, kind 0:21:00.240 --> 0:21:03.240 of articles online, whether that's here are some recipes for 0:21:03.320 --> 0:21:07.320 you to hear hard news articles to your latest comic 0:21:07.359 --> 0:21:11.040 webcomic that you follow. What will people want to read tomorrow? 0:21:11.200 --> 0:21:13.840 Was the general question we're trying to answer now. 0:21:13.960 --> 0:21:18.200 The first AI research was done at Dartmouth in nineteen 0:21:18.320 --> 0:21:20.800 fifty six. Does that sound correct? 0:21:20.920 --> 0:21:22.359 I'm probably under that heading. 0:21:22.520 --> 0:21:24.040 Yeah, And what were they trying to do? 0:21:24.960 --> 0:21:28.199 I believe the first ones were they were focusing on language, 0:21:28.240 --> 0:21:31.120 if I recall correctly. I'm not one hundred percent sure. 0:21:31.320 --> 0:21:34.600 But the goal was, in your mind, whether it's specifically 0:21:34.680 --> 0:21:37.240 a dartmouth or not. They were trying to get computers 0:21:37.240 --> 0:21:39.320 to talk well. 0:21:39.520 --> 0:21:42.119 So there had been a lot of debate in the 0:21:42.160 --> 0:21:44.399 early port of the twentieth century about whether or not 0:21:44.520 --> 0:21:48.439 machines can be intelligent at all, and there have been 0:21:48.480 --> 0:21:52.280 a lot of debate over definitions and philosophy up until 0:21:52.400 --> 0:21:57.080 Alan Teering in nineteen fifty wrote an essay on computing, 0:21:57.160 --> 0:21:59.919 machinery and Intelligence where he proposed what he called the 0:22:00.040 --> 0:22:04.679 imitation game. And it's basically the basic principle hind it is, 0:22:05.119 --> 0:22:08.719