WEBVTT - Amazon Part 2: The Obsessive Secrecy Around Alexa 0:00:00.160 --> 0:00:03.760 Jeff Adams as a computer scientist. In two thousand and eleven, 0:00:03.880 --> 0:00:07.120 he was at a really tiny startup, just thirteen employees, 0:00:07.440 --> 0:00:10.360 called the App. They were working on technology that could 0:00:10.400 --> 0:00:15.080 transcribe voicemail into text messages. And then one day Jeff's 0:00:15.080 --> 0:00:19.239 fairy godmother showed up in the form of executives from Amazon. 0:00:19.720 --> 0:00:22.000 We went to a show and some people from Amazon 0:00:22.160 --> 0:00:24.640 came and talked to us, and they might be interested 0:00:24.720 --> 0:00:28.360 in possibly even acquiring us. And I thought, that's absurd. 0:00:28.480 --> 0:00:32.120 What could Amazon possibly be interested in speech for? Maybe 0:00:32.440 --> 0:00:35.600 they want you to be able to, you know, call 0:00:35.680 --> 0:00:38.600 up on the phone and order a book. Jeff figured 0:00:38.600 --> 0:00:41.040 the talks wouldn't amount too much, but over the course 0:00:41.080 --> 0:00:45.000 of the next few months they became serious. Naturally, he 0:00:45.080 --> 0:00:48.280 wanted to know why Amazon was interested. They said, you know, 0:00:48.360 --> 0:00:51.000 don't ask, it's our business. And I thought, well, it's 0:00:51.000 --> 0:00:53.000 gonna be my business too, and you might want to 0:00:53.040 --> 0:00:58.280 know whether whether we can do it um, But they wouldn't. 0:00:58.280 --> 0:01:00.160 They wouldn't say anything. They said, I'm sorry, were not 0:01:00.200 --> 0:01:02.639 at liberty to discuss any of this. There was one 0:01:02.720 --> 0:01:05.120 hint that Amazon would give, so one of the things 0:01:05.200 --> 0:01:07.600 they did let us know was that the guy who 0:01:07.600 --> 0:01:10.440 was kind of running this was Greg Hart, who we 0:01:10.560 --> 0:01:14.080 knew was the right hand person of Jeff Bezos. So 0:01:14.160 --> 0:01:16.280 they go to all the meetings with them. They are 0:01:16.319 --> 0:01:20.360 like the confidante, the consilior or whatever. Uh. And so 0:01:20.400 --> 0:01:23.800 we knew, okay, something this does have visibility at the 0:01:23.840 --> 0:01:27.520 highest levels, and Bezos must be behind it in in 0:01:27.680 --> 0:01:31.000 some form. As the deal was being finalized, Jeff's team 0:01:31.040 --> 0:01:34.760 traveled to Florence, Italy, to a computer speech conference along 0:01:34.800 --> 0:01:38.479 with several Amazon managers. They even stayed in a villa together, 0:01:38.959 --> 0:01:41.399 but the Amazon people didn't want to be seen with 0:01:41.440 --> 0:01:45.320 the voice transcription people during the conference. From day to 0:01:45.400 --> 0:01:49.120 day in the conference, we had to let we we 0:01:49.200 --> 0:01:51.800 I rented a van and we drove down into town 0:01:51.880 --> 0:01:54.240 and to the conference together. I had to let them 0:01:54.240 --> 0:01:57.160 out around the block, around the corner. It's like being 0:01:57.200 --> 0:01:59.920 a teenager and trying to hide the person you're dating 0:02:00.160 --> 0:02:02.280 from your parents. First of all, they didn't want anyone 0:02:02.280 --> 0:02:04.160 to know that they were from Amazon. They didn't want 0:02:04.160 --> 0:02:06.320 anyone to see them with us, so we had to 0:02:06.440 --> 0:02:08.480 avoid them at the conference. We would like, you know, 0:02:08.560 --> 0:02:12.280 look at each other and kind of smile from across 0:02:12.360 --> 0:02:15.000 the room or across the courtyard or whatever. But we 0:02:15.000 --> 0:02:18.160 couldn't sit together, we couldn't talk together, or whatever. That 0:02:18.280 --> 0:02:22.440 fall the deal was done. Amazon buys the speech company, 0:02:22.520 --> 0:02:25.480 and Jeff is still living in mystery. We had been 0:02:25.520 --> 0:02:28.200 employees of Amazon for like a week, but they had 0:02:28.240 --> 0:02:31.040 still not told us any anything. They said, no, we 0:02:31.080 --> 0:02:34.000 have to tell you in a closed room. So we 0:02:34.040 --> 0:02:36.360 all came out to Seattle. We all got in the 0:02:36.400 --> 0:02:41.360 room together. They closed the door, locked the door, put 0:02:41.680 --> 0:02:46.200 paper over the window in the door, closed the exterior windows. Uh. 0:02:46.440 --> 0:02:51.320 It was very secret, very mysterious, and hush, hush. Jeff 0:02:51.360 --> 0:02:55.520 and his team leaned forward. This was the moment they said, 0:02:55.919 --> 0:02:59.519 imagine something the size of a coke can that's sitting 0:02:59.560 --> 0:03:02.000 on your able and we're going to sell this for 0:03:04.720 --> 0:03:06.760 uh and people will be able to talk to it. 0:03:06.840 --> 0:03:10.200 And we thought, you know, we can't do this. The 0:03:10.280 --> 0:03:13.360 technology isn't there yet, it doesn't exist yet. And of 0:03:13.400 --> 0:03:17.359 course they did do it by inventing the technology, because 0:03:17.360 --> 0:03:22.440 the product they're talking about becomes Alexa, Amazon's AI virtual assistant, 0:03:22.720 --> 0:03:26.000 and that twenty co can is the Echo smart speaker. 0:03:26.360 --> 0:03:29.640 Though it ends up selling for more than twenty Alexa 0:03:29.800 --> 0:03:33.480 is one of Amazon's first forays into artificial intelligence, and 0:03:33.520 --> 0:03:37.080 it would bring Amazon into people's living rooms, further weaving 0:03:37.120 --> 0:03:43.760 it into their lives. You're listening to Foundering. I'm your host, 0:03:43.960 --> 0:03:47.320 Brad Stone. In this episode, we're going to tell the 0:03:47.400 --> 0:03:51.080 story of the creation of Alexa. How Jeff Bezos almost 0:03:51.160 --> 0:03:55.440 singlehandedly conceived the idea for a voice activated computer and 0:03:55.640 --> 0:04:00.400 drove the device's creation. Alexa helped to solidify Amazon's image 0:04:00.400 --> 0:04:04.560 as an innovator and in the process challenge conventional notions 0:04:04.560 --> 0:04:07.840 of privacy. Will tell you more after a quick break. 0:04:19.080 --> 0:04:22.000 In the last episode, we talked about how Amazon survived 0:04:22.000 --> 0:04:25.000 the dot com bust, how Jeff Bezos said to colleagues, 0:04:25.240 --> 0:04:27.320 the only way out of this is to invent her 0:04:27.320 --> 0:04:31.080 way out. During those next few years, Bezos launched Amazon 0:04:31.160 --> 0:04:35.159 Prime and the Kindle. But there's another innovation from this era, 0:04:35.560 --> 0:04:38.880 one that generates huge profits for Amazon and sets the 0:04:38.960 --> 0:04:42.680 stage for the creation of its pioneering voice assistant. It's 0:04:42.720 --> 0:04:47.560 Amazon's cloud service called AWS, or Amazon Web Services. It's 0:04:47.600 --> 0:04:51.440 the most difficult of Bezos's inventions for lay people to understand, 0:04:52.080 --> 0:04:54.799 because as much as it may already seem like Amazon 0:04:54.920 --> 0:04:58.279 is everywhere a massive store, a network of warehouses, and 0:04:58.320 --> 0:05:01.039 the fleet of delivery people. There's one more way in 0:05:01.040 --> 0:05:05.960 which Amazon is ubiquitous indispensable. Amazon is literally holding up 0:05:06.040 --> 0:05:09.520 much of the Internet today. AWS is a highly profitable, 0:05:09.640 --> 0:05:14.320 fifty billion dollar annual business for Amazon. When you watch Netflix, 0:05:14.480 --> 0:05:18.440 you're often streaming video from Amazon. When you send a Snapchat, 0:05:18.640 --> 0:05:23.120 you're using Amazon servers. By two thousand and ten, Bezos 0:05:23.240 --> 0:05:25.760 was asking everyone in the company, what are you doing 0:05:25.800 --> 0:05:30.039 for AWS? You wanted Amazon to deepen and exploit the 0:05:30.080 --> 0:05:33.000 early lead it already had in running massive, cutting edge 0:05:33.040 --> 0:05:36.320 data centers. Around this time, he was at lunch with 0:05:36.360 --> 0:05:39.800 Greg Hart, the guy described as his consiliary, and the 0:05:39.880 --> 0:05:43.200 conversation turned how Google was starting to let people search 0:05:43.640 --> 0:05:47.640 just by talking into their smartphones. Here's Greg and and 0:05:47.680 --> 0:05:49.680 I said, look at how convenient is And I just 0:05:49.800 --> 0:05:53.640 did a Google voice search on pizza and your me um. 0:05:53.760 --> 0:05:57.000 And you know, it's just so much faster than actually typing, 0:05:57.360 --> 0:05:59.080 you know, with your fingers or your thumbs pizza and 0:05:59.120 --> 0:06:01.920 your me on your phone waiting for results to come back. Now, 0:06:02.120 --> 0:06:04.960 Bezos has long been a believer in voice. Here he 0:06:05.080 --> 0:06:07.160 is all the way back in the year two thousand 0:06:07.560 --> 0:06:12.040 talking to Charlie Rose, I believe that for mobile commerce, 0:06:12.440 --> 0:06:14.640 the thing that's going to be the biggest part of 0:06:14.640 --> 0:06:17.919 that is voice, and I think there's a lot that 0:06:18.000 --> 0:06:20.640 can happen in the In the short term, it'll be 0:06:20.720 --> 0:06:23.520 kind of a stilted special purpose language for talking to 0:06:23.560 --> 0:06:25.839 Amazon dot Com. But in the long term it could 0:06:25.839 --> 0:06:29.400 even be natural language processing. I think that is a 0:06:29.480 --> 0:06:33.240 real mind bender. Now. A decade later, it seemed like 0:06:33.279 --> 0:06:36.360 this early prediction might come true. A few weeks after 0:06:36.400 --> 0:06:39.680 their fateful lunch, Bazo sent an email to Greg that read, 0:06:40.240 --> 0:06:43.240 we should build a twenty device with its brains in 0:06:43.240 --> 0:06:47.200 the cloud that's completely controlled by your voice. The central 0:06:47.240 --> 0:06:50.279 insight here was that the product could be inexpensive because 0:06:50.279 --> 0:06:53.400 it's brains resided in Amazon's data centers and could be 0:06:53.480 --> 0:06:57.000 constantly improved, So if you buy it, the product would 0:06:57.000 --> 0:07:00.159 be upgrading itself and you wouldn't even know it. A 0:07:00.200 --> 0:07:03.440 few weeks later, Greg started recruiting people inside the company 0:07:03.480 --> 0:07:07.200 for the project. He proceeded with total secrecy. The only 0:07:07.240 --> 0:07:09.760 thing he told colleagues was that it had the potential 0:07:09.800 --> 0:07:12.200 to be bigger than the Kindle, and at the time 0:07:12.240 --> 0:07:15.440 that seemed like laughable that that we could create something 0:07:15.480 --> 0:07:17.560 that could be bigger than Kindle, because Kindle, you know, 0:07:17.560 --> 0:07:22.480 at that time was now for years into its life, 0:07:23.760 --> 0:07:27.480 dominating the e reader and e ink market and was 0:07:27.560 --> 0:07:30.400 changing the way that people thought about meeting. Next, Greg 0:07:30.400 --> 0:07:33.960 acquired YAP, the speech recognition company where Jeff Adams worked. 0:07:34.480 --> 0:07:36.480 That team didn't think he knew what he was doing either. 0:07:37.280 --> 0:07:39.520 Amazon wanted people to be able to talk to a 0:07:39.600 --> 0:07:43.160 device across a room and haven't understand them. That's called 0:07:43.200 --> 0:07:47.320 far field speech recognition, and the technology for it didn't exist. 0:07:47.720 --> 0:07:52.960 Here's Jeff Adams again. The big problem is speech recognition 0:07:53.000 --> 0:07:56.840 at the time really relied on a close talking microphone. 0:07:56.880 --> 0:07:59.400 You needed to capture the speech close to the person's mouth. 0:07:59.800 --> 0:08:01.880 And they were talking about, Oh, I'm going to be 0:08:01.880 --> 0:08:04.080 in the garage and this thing is gonna coach me 0:08:04.160 --> 0:08:07.320 through changing my car's oil, and it's gonna be over 0:08:07.360 --> 0:08:09.000 on the other side of the garage, and I'm gonna 0:08:09.040 --> 0:08:11.480 be shouting things to it, and it's gonna be shouting 0:08:11.480 --> 0:08:15.520 instructions back or whatever. And we thought, you can't do that. 0:08:15.560 --> 0:08:19.200 There's there's too many reflective surfaces in the room. They're 0:08:19.200 --> 0:08:22.480 gonna mess up the audio. Basically, if you're shouting across 0:08:22.520 --> 0:08:25.600 the room, that introduces lots of echo and the computer 0:08:25.680 --> 0:08:28.920 gets confused. So I realized that this far field issue 0:08:28.960 --> 0:08:30.520 was going to be a problem. I didn't want to 0:08:30.560 --> 0:08:33.240 say anything in in front of the group. I didn't wanna, 0:08:33.320 --> 0:08:38.480 you know, um, appear unsupportive. But afterwards I tracked down 0:08:38.480 --> 0:08:41.400 Greg Hart in the hallway and I said, Greg, we're 0:08:41.400 --> 0:08:43.560 excited about this, but I think you should know that 0:08:43.920 --> 0:08:47.400 what you want to do, the technology isn't there yet. 0:08:47.440 --> 0:08:50.720 It doesn't exist yet. We don't we we don't have 0:08:50.960 --> 0:08:54.720 that technology to solve this far field speech problem. And 0:08:55.000 --> 0:08:58.960 he was unflapped. He said, uh, I appreciate that, thank 0:08:59.000 --> 0:09:01.679 you for telling me, But solve it. We are an Amazon, 0:09:02.000 --> 0:09:04.839 We've got resources, hire as many people as you need, 0:09:04.880 --> 0:09:07.679 to take as long as you want, but you know, 0:09:07.720 --> 0:09:11.360 solve the problem. The team gave itself the code name Doppler, 0:09:11.600 --> 0:09:14.319 as in the Doppler effect, which describes the way a 0:09:14.400 --> 0:09:18.240 sound wave moves with respect to a listener. They hotly 0:09:18.280 --> 0:09:21.079 debated everything from what to name the device, to what 0:09:21.120 --> 0:09:23.679 it should do, to how to market that to the public. 0:09:24.520 --> 0:09:27.920 At first, they met with Bezos once a month. He 0:09:27.920 --> 0:09:31.960 would get deeply involved in the technology. UM, not just 0:09:32.160 --> 0:09:35.280 the business or the product, but deeply involved in the technology. 0:09:35.360 --> 0:09:37.400 He I would say, he stayed very close to the 0:09:37.440 --> 0:09:41.200 project throughout. As a project progressed, those meetings would increase 0:09:41.440 --> 0:09:44.880 in frequency UM and UH. And by the end we 0:09:44.880 --> 0:09:47.400 were meeting with him, you know, leading up to launch, 0:09:47.440 --> 0:09:49.440 I would say, there probably wasn't a day that went 0:09:49.480 --> 0:09:51.960 by that we didn't have at least one meeting with 0:09:52.040 --> 0:09:55.000 Jeff UM. And you know, sometimes there were multiple meetings 0:09:55.000 --> 0:09:59.880 with Jeff in a single day. UM. That is both 0:10:00.280 --> 0:10:03.320 a blessing and a curse. The blessing came in the 0:10:03.360 --> 0:10:06.360 form of money. The team could spend whatever they needed 0:10:06.400 --> 0:10:09.760 to break through the technical obstacles. This is a big 0:10:09.800 --> 0:10:13.120 project that Jeff is personally invested in. It was his 0:10:13.240 --> 0:10:17.120 idea originally. He can be very demanding. UM, he can 0:10:17.160 --> 0:10:20.680 push teams UM. But at the same time, like being 0:10:20.720 --> 0:10:24.240 able to go through that experience and have his brain 0:10:24.400 --> 0:10:31.360 on your side is an immensely powerful opportunity and experience 0:10:31.480 --> 0:10:34.520 if you let it, like if you get really frustrated 0:10:34.520 --> 0:10:36.200 by it. And at times some of the people on 0:10:36.280 --> 0:10:39.600 my product team would say, it feels like Jeff is 0:10:39.600 --> 0:10:43.600 the product manager for Alexa. Bezo set the vision for Alexa. 0:10:44.160 --> 0:10:46.760 He wanted it to be the Star Trek computer. He 0:10:46.840 --> 0:10:51.600 pictured a versatile, conversational machine that could respond to any question. 0:10:52.160 --> 0:10:54.000 You should be able to sit in your living room 0:10:54.240 --> 0:10:58.280 and ask Alexa anything. It sounds simple, but it's not 0:10:58.800 --> 0:11:00.840 because they needed a lot of data to train the 0:11:00.880 --> 0:11:04.760 AI algorithms. For example, for Alexa to tell you the weather, 0:11:05.120 --> 0:11:07.559 they would have to understand the phrasing of your question 0:11:07.800 --> 0:11:10.880 and your dialect from across a noisy room, and sort 0:11:10.960 --> 0:11:14.240 through databases for the right answer. This would take Alexa 0:11:14.320 --> 0:11:17.679 a step further than Syrior Google Voice, which only worked 0:11:17.679 --> 0:11:20.880 when you spoke directly into your phone. But the Alexa 0:11:20.960 --> 0:11:23.520 team just didn't have the data to get the AI smart. 0:11:24.000 --> 0:11:27.959 The early prototypes worked so badly that even Amazon employees 0:11:28.000 --> 0:11:31.280 didn't really want to test them. Bassos was getting impatient. 0:11:31.720 --> 0:11:34.480 He actually walked out of a number of internal meetings 0:11:34.480 --> 0:11:37.800 and frustration. Then Greg and his colleague struck upon an 0:11:37.840 --> 0:11:43.240 idea they called Project Amped. Basically, rents houses or apartments 0:11:43.280 --> 0:11:46.600 in cities all over the US, and we would put 0:11:46.640 --> 0:11:51.120 devices in those apartments. They were all camouflaged, and they 0:11:51.160 --> 0:11:54.080 were not just Amazon devices. There were other companies devices 0:11:54.160 --> 0:11:57.120 there as well, UM, some of which were visible, some 0:11:57.200 --> 0:12:01.080 of which were not visible. And characteristic of all Amazon projects, 0:12:01.120 --> 0:12:05.720 secrecy was the priority. Employees were careful to conceal Amazon's 0:12:05.720 --> 0:12:08.079 identity when they set up the room, and that was 0:12:08.120 --> 0:12:10.400 all about obfuscation, you know, we didn't want people to 0:12:10.480 --> 0:12:13.600 understand what company we were working for. UM, and so 0:12:13.760 --> 0:12:15.800 you could see an xbox, you could see this, you 0:12:15.840 --> 0:12:18.199 could see that UM. And then we would have all 0:12:18.200 --> 0:12:21.440 these Alexa devices hidden throughout the room or echo devices, 0:12:21.960 --> 0:12:25.680 prototype devices, and then they brought in testers, thousands of 0:12:25.720 --> 0:12:28.640 people paid an hourly wage, coming in at all hours 0:12:28.640 --> 0:12:30.640 of the day and days of the week to train 0:12:30.679 --> 0:12:34.920 the machines. Then we would bring in we would recruit participants, 0:12:35.320 --> 0:12:38.600 have people with different accents, you know, male, female, different 0:12:38.600 --> 0:12:41.520 ages who would come in and we would ask them 0:12:41.600 --> 0:12:45.640 to do a mixture of reading scripted things and then 0:12:45.720 --> 0:12:48.319 also talking in a much more off the cuff all 0:12:48.320 --> 0:12:51.600 the cart fashion to ask for things that people would 0:12:51.640 --> 0:12:55.920 be we hoped asking Alexa when we launched. Amazon conducted 0:12:55.920 --> 0:12:58.800 the data gathering effort with such stealth that neighbors started 0:12:58.840 --> 0:13:02.280 to get suspicious. House that we rented in Boston that 0:13:03.440 --> 0:13:05.400 the neighbors thought that because we had a lot of 0:13:05.400 --> 0:13:08.000 cars showing up and individuals getting out on their own 0:13:08.040 --> 0:13:11.240 and coming in and spending I think maybe an hour, 0:13:11.920 --> 0:13:13.680 and so there's a lot of sort of you know, 0:13:13.720 --> 0:13:16.679 transient in and out traffic, and neighbors thought that maybe 0:13:16.679 --> 0:13:19.760 there was you know, a drug running ring or something 0:13:19.800 --> 0:13:22.240 else going on, and so the police actually showed up 0:13:22.480 --> 0:13:26.680 despite the attention from police. Bezos loved the inventiveness of 0:13:26.679 --> 0:13:30.960 the data gathering program. When we first took AMPED to him, 0:13:32.240 --> 0:13:36.959 his response was effectively like, now you're talking, like let's 0:13:37.000 --> 0:13:40.120 do this, um, Like tell me if you want more money, 0:13:40.440 --> 0:13:43.720 Greg says. The project AMPED ran in thirteen cities and 0:13:43.760 --> 0:13:47.080 included over ten people. The devices were placed all over 0:13:47.120 --> 0:13:49.400 the room, and so we were trying to capture a 0:13:49.440 --> 0:13:53.960 massive amount of acoustic data about, you know, how noise 0:13:54.040 --> 0:13:56.760 performs in a room in all kinds of different rooms, 0:13:56.760 --> 0:14:00.160 in bedrooms and living rooms, in kitchens, in bathrooms. The 0:14:00.200 --> 0:14:04.920 result Amazon basically solved the far field voice problem. It 0:14:05.000 --> 0:14:07.959 took six months for this company to solve a problem 0:14:08.000 --> 0:14:11.960 that had stumped speech scientists for decades. But since the 0:14:12.000 --> 0:14:15.120 project was top secret, they weren't able to tell anyone, 0:14:15.440 --> 0:14:19.640 not even the speech science community, about their big technological breakthrough. 0:14:20.560 --> 0:14:24.119 Here's Ahmed Boozid. He had been working in voice technology 0:14:24.200 --> 0:14:27.320 for almost twenty years when Amazon tried to recruit him 0:14:27.400 --> 0:14:30.720 in early two fifteen. I turned him down. I said, 0:14:30.840 --> 0:14:33.720 you're not showing me anything. You have never done anything 0:14:33.760 --> 0:14:36.880 in voice, and you're telling me that you are going 0:14:36.920 --> 0:14:39.640 to any uly to to go from DC and go 0:14:39.720 --> 0:14:42.720 to Seattle and they'll let you. Just trust us. I know, no, 0:14:42.920 --> 0:14:44.800 I don't think so. I mean, I've seen the Kindle 0:14:44.880 --> 0:14:48.160 and it's fine, it's okay, but Kindall is like, you know, 0:14:49.040 --> 0:14:50.800 it's a toy compared to what you guys are trying 0:14:50.800 --> 0:14:54.000 to do. So anyway, I said no twice. Eventually they 0:14:54.040 --> 0:14:57.480 flew Ahmed out to Amazon headquarters to see Alexa in person. 0:14:57.880 --> 0:15:00.400 So it was, you know, on the table, and so 0:15:00.440 --> 0:15:02.880 the first thing I did is I asked it for 0:15:02.880 --> 0:15:04.960 for music, and he plays some Miles Davis and it 0:15:05.080 --> 0:15:09.400 did I'm like, okay, cool, And then I suppos I said, 0:15:09.440 --> 0:15:12.320 you know what time is it? Man stopping and told 0:15:12.320 --> 0:15:14.280 me the time, and then it continued playing Miles datas, 0:15:14.320 --> 0:15:17.640 which was great. Interacting with Alexa was almost a moving 0:15:17.720 --> 0:15:21.360 moment for Ahmed. He felt like decades of scientific research 0:15:21.400 --> 0:15:24.880 had been realized in this little device, and I, you know, 0:15:24.960 --> 0:15:27.600 I was like, Okay, this is amazing, This is amazing. 0:15:28.480 --> 0:15:31.080 This is clearly an important moment in technology. And this 0:15:31.120 --> 0:15:32.720 is what I've been doing all my life, right, solving 0:15:32.720 --> 0:15:35.360 the problem problems. See, you know, we people who are 0:15:35.400 --> 0:15:38.680 in the speech were are always saying this expression, you know, 0:15:38.720 --> 0:15:41.120 speeches around the corner. We've been saying it's since the 0:15:41.120 --> 0:15:44.120 mid nineties, species around the corner, because we do believe 0:15:44.160 --> 0:15:47.000 that if you do speech well, a lot of stuff 0:15:47.920 --> 0:15:52.120 becomes easy to do for people. Ahmed was particularly impressed 0:15:52.120 --> 0:15:56.240 by two innovations that Amazon made. One that they solved 0:15:56.240 --> 0:16:00.120 the far field speech issue and secondly a let so 0:16:00.200 --> 0:16:03.760 it was fast to speed right. The fact that it's 0:16:03.800 --> 0:16:05.720 just I was, I was just amazed, Like, how the 0:16:05.720 --> 0:16:09.680 hell does come back within within like two seconds? I mean, 0:16:09.880 --> 0:16:12.200 if I if I type, you know, if I want 0:16:12.200 --> 0:16:14.640 to launch a page on my browser, or it takes 0:16:14.680 --> 0:16:17.840 like sometimes like three four or five seconds. Right, How 0:16:17.920 --> 0:16:19.520 is how the hell is this thing doing all of 0:16:19.520 --> 0:16:23.520 these things, getting my voice going to the cloud, processing it, 0:16:24.040 --> 0:16:27.640 coming back talking back within two seconds. It's like a 0:16:27.680 --> 0:16:30.360 magic even for someone like me who was been like 0:16:30.440 --> 0:16:33.640 you should be jaded by by by then, Right, okay, 0:16:33.680 --> 0:16:36.200 I know exactly how it's happened. I was astonished. Right, 0:16:36.440 --> 0:16:39.080 how the hell did they do that? We'll be right back. 0:16:50.080 --> 0:16:53.400 I want to take a moment to address Amazon's obsessive secrecy. 0:16:54.040 --> 0:16:57.360 Many companies are secretive, but tech companies today have the 0:16:57.400 --> 0:16:59.800 eyes of the world on them, so they are taking 0:17:00.000 --> 0:17:03.600 corporate secrecy to a whole other level. Like that story. 0:17:03.680 --> 0:17:07.679 Jeff Adams, the speech scientist, told that the Amazon executives 0:17:07.840 --> 0:17:10.600 were so secretive when the acquired a company that they 0:17:10.600 --> 0:17:14.160 wouldn't even be seen together at a conference. That's typical 0:17:14.160 --> 0:17:18.040 of Amazon. This intense drive for secrecy comes straight from 0:17:18.119 --> 0:17:21.480 Jeff Bezos. He wants to tightly control the messaging around 0:17:21.480 --> 0:17:25.639 Amazon's new products. The idea is that complete secrecy pays 0:17:25.720 --> 0:17:29.600 off with a surprising, almost magical reveal once the product 0:17:29.680 --> 0:17:34.119 is launched. That man keeping Alexa under wraps until launch, 0:17:34.560 --> 0:17:38.480 preserving the details of how it actually worked, and choosing 0:17:38.680 --> 0:17:42.520 the perfect voice here's Greg Hart, how do you define 0:17:42.560 --> 0:17:46.000 the characteristics that you want the voice to have? And 0:17:46.040 --> 0:17:48.120 so there was sort of a brief that we wrote 0:17:48.160 --> 0:17:51.639 up about the qualities you wanted the person to be knowledgeable. 0:17:51.680 --> 0:17:54.560 We very quickly early on, decided the first voice would 0:17:54.560 --> 0:17:57.199 be female. We knew there would be additional voices, but 0:17:57.280 --> 0:17:59.800 we felt that the first voice should be female in 0:18:00.160 --> 0:18:04.400 because yeah, so that's the logical question why In part 0:18:04.480 --> 0:18:07.439 because the we knew that the device would be in 0:18:07.560 --> 0:18:12.040 the kitchen, and we felt that a female voice would 0:18:12.080 --> 0:18:15.680 be more open and inviting and warm than a male 0:18:15.800 --> 0:18:18.719 voice would be in that environment, and more appropriate in 0:18:18.760 --> 0:18:22.720 that environment. Not because of any sexist things, but just 0:18:22.760 --> 0:18:25.480 because of the fact that we knew, um that we 0:18:25.600 --> 0:18:28.920 knew the way that, um, that's the right way to 0:18:28.960 --> 0:18:34.959 say this. We had seen evidence that people respond differently 0:18:35.080 --> 0:18:39.040 to male computer voices than to female computer voices, and 0:18:39.040 --> 0:18:42.760 they respond more positively to female computer voices, and we 0:18:42.880 --> 0:18:45.520 wanted because the device was in the home, we wanted 0:18:45.560 --> 0:18:48.919 it to be a device that everybody would respond positively to. 0:18:49.400 --> 0:18:52.280 Putting a female voice in the kitchen, so to speak, 0:18:52.440 --> 0:18:55.879 would turn out to be a somewhat controversial choice, and 0:18:55.960 --> 0:18:59.520 Amazon isn't the only company to do this. Serie. Google 0:18:59.600 --> 0:19:04.439 Voice and Microsoft Cortana are all women by default, So 0:19:04.600 --> 0:19:07.640 Amazon moved forward in their search for a voice. They 0:19:07.680 --> 0:19:10.640 contracted with the same studio that had developed the voice 0:19:10.640 --> 0:19:13.520 of Sirie for Apple. Okay, I set up your meeting 0:19:13.520 --> 0:19:16.480 with David tomorrow. Shall I schedule it? Who is voiced 0:19:16.480 --> 0:19:19.960 by Susan Bennett, a career voice actress. Hi, my name 0:19:20.000 --> 0:19:22.800 is Susan Bennett, and I have a voice actor and 0:19:23.440 --> 0:19:27.240 the original voice of Sirie. So to find their Alexa, 0:19:27.520 --> 0:19:30.760 the studio had half a dozen female voice actresses read 0:19:30.800 --> 0:19:35.320 for hours. They read entire books and random articles, and 0:19:35.400 --> 0:19:39.600 finally Greg Hart and Jeff Bezos picked one woman. Her 0:19:39.680 --> 0:19:44.520 identity hasn't been revealed in all this time. It's amazing 0:19:44.560 --> 0:19:47.600 to me as an Amazon reporter that Amazon has still 0:19:47.640 --> 0:19:49.800 been able to keep the voice of Alexa a secret. 0:19:50.200 --> 0:19:54.520 In I started canvassing voice over actors, asking if they 0:19:54.600 --> 0:19:58.639 knew the identity of Alexa. No one knew. Finally, I 0:19:58.680 --> 0:20:00.919 got a tip from someone who had worked with the 0:20:00.960 --> 0:20:04.800 studio that Amazon contracted with, and they said that Alexa 0:20:04.880 --> 0:20:07.600 was voiced by an actress and singer from Boulder, Colorado 0:20:07.880 --> 0:20:12.080 named Nina Raleigh reached out to Raleigh numerous times, she 0:20:12.119 --> 0:20:14.960 wouldn't confirm that she's the voice of Alexa, but she 0:20:15.040 --> 0:20:18.879 didn't deny it either. Eventually I confirmed with enough people 0:20:18.880 --> 0:20:23.000 that I feel confident she's the one. She's Alexa. Here's 0:20:23.000 --> 0:20:24.960 a clip from her website and add that she did 0:20:25.000 --> 0:20:29.000 for Time Warner Cable. Thanks for choosing Time Warner. Now 0:20:29.000 --> 0:20:31.760 that you've ordered your installation, let's set up a time 0:20:31.800 --> 0:20:35.760 to get a crew over to your place. And here's Alexa. 0:20:36.240 --> 0:20:39.800 Time Warner Cable, also simply known as Time Warner, was 0:20:39.800 --> 0:20:42.960 an American cable television company. It was ranked the second 0:20:43.040 --> 0:20:45.720 largest cable company in the United States. But imagine what 0:20:45.800 --> 0:20:48.720 it might be like to be Nina Raleigh. Your voice 0:20:48.760 --> 0:20:52.080 is piped into millions of homes every day, each new 0:20:52.119 --> 0:20:55.439 iteration of Alexa, each update for Amazon, you have to 0:20:55.480 --> 0:20:59.040 record something new, Like when Amazon Fresh released a product 0:20:59.040 --> 0:21:02.879 called the single Burger. She had to be available single 0:21:02.920 --> 0:21:05.719 Cow Burger, a beef burger made with meat from just 0:21:05.800 --> 0:21:09.560 a single cow. She's chained to Alexa. It sounds like 0:21:09.600 --> 0:21:14.080 a life of obscurity and loneliness. Compare that to Siri 0:21:14.240 --> 0:21:18.159 and Susan Bennett. She's spoken on CNN and countless talk shows. 0:21:18.520 --> 0:21:20.240 She's even been able to use her fame as the 0:21:20.320 --> 0:21:25.040 voice of Sirie as leverage for many more opportunities. Greg 0:21:25.080 --> 0:21:28.240 saw the burden that Amazon secrecy put on the human 0:21:28.359 --> 0:21:33.520 behind the voice of Alexa. I never met her um 0:21:34.560 --> 0:21:37.160 and I don't even know that I might have. I'm 0:21:37.160 --> 0:21:39.480 not have spoken with her once, but I never met her. 0:21:40.040 --> 0:21:43.600 And it would be interesting to be in her shoes now, 0:21:43.680 --> 0:21:46.840 because it's on the one hand, it's it's incredible that 0:21:46.920 --> 0:21:50.040 this thing that you contributed to is now so ubiquitous. 0:21:50.119 --> 0:21:52.480 On the other hand, I would think it would be, 0:21:52.640 --> 0:21:56.960 you know, maybe a little bit unsettling. Amazon introduced the 0:21:57.000 --> 0:22:01.160 Echo in November two, fourteen were re strived from Amazon. 0:22:01.359 --> 0:22:04.080 I didn't know what it was. With a YouTube video, 0:22:04.520 --> 0:22:12.720 Alexa play rock music, rock music, alexis stop. We want 0:22:12.720 --> 0:22:16.000 to try Alexa, what time is it? The time is three. 0:22:17.119 --> 0:22:19.159 You actually don't have to yell at it. Okay. It 0:22:19.240 --> 0:22:21.760 uses far Fueld technology so it can hear you from 0:22:21.840 --> 0:22:24.880 anywhere in the room, So I can just hear you anywhere. Yes, 0:22:25.600 --> 0:22:29.280 that promise was enough. A responsive computer that can tell 0:22:29.320 --> 0:22:32.840 the time or the weather, play music, and answer questions 0:22:32.840 --> 0:22:36.000 from across the room. Tens of thousands of people joined 0:22:36.000 --> 0:22:39.159 a waiting list to receive a device. Here's Bezos at 0:22:39.200 --> 0:22:42.800 the Recode Technology conference in two thousand sixteen saying that 0:22:42.840 --> 0:22:47.000 the future is Alexa. He's speaking to Walt Mossburg. But 0:22:47.080 --> 0:22:49.879 it has been a dream ever since, you know, people started, 0:22:49.920 --> 0:22:52.080 you know, in the early days of science fiction to 0:22:52.160 --> 0:22:55.280 have a computer that you can talk to. So are 0:22:55.320 --> 0:22:58.520 you deeply committed to this becoming a huge part of 0:22:58.560 --> 0:23:02.119 your business and what you Absolutely We've been working on it. 0:23:02.200 --> 0:23:04.320 You know, we worked on it. We have more than 0:23:04.359 --> 0:23:09.160 a thousand people dedicated us to Alexa in the Echo ecosystem, 0:23:09.240 --> 0:23:12.320