WEBVTT - TechStuff Classic: AI Assistants and You 0:00:04.440 --> 0:00:12.320 Welcome to tech Stuff, a production from iHeartRadio. Hey there, 0:00:12.360 --> 0:00:15.600 and welcome to tech Stuff. I'm your host, Jonathan Strickland. 0:00:15.640 --> 0:00:19.239 I'm an executive producer with iHeartRadio, and how the tech 0:00:19.320 --> 0:00:23.200 are you here? We are on a Friday again. That 0:00:23.280 --> 0:00:27.560 means it's time for a classic episode. This episode originally 0:00:27.640 --> 0:00:31.680 published on June fifteenth, twenty sixteen. It is called AI 0:00:31.960 --> 0:00:36.720 Assistants and You. This is clearly an episode that could 0:00:36.880 --> 0:00:40.720 use an update now twenty sixteen. Man boy, We've got 0:00:40.760 --> 0:00:43.400 a lot more to say about AI assistants these days, 0:00:43.400 --> 0:00:46.960 but I'll leave that for the autro. Let's take a listen. 0:00:48.040 --> 0:00:53.320 So we want to talk today. We being me and you. 0:00:53.320 --> 0:00:54.080 You can talk back. 0:00:54.120 --> 0:00:57.000 I just won't be able to hear you about these 0:00:57.040 --> 0:01:00.800 personal digital assistants, but not the pen of the past. 0:01:00.840 --> 0:01:04.000 We want to talk about the series and the quirtanas 0:01:04.000 --> 0:01:07.360 and the Google assistants and things of that nature. And 0:01:07.400 --> 0:01:10.399 I want to specifically look into how are these going 0:01:10.440 --> 0:01:13.160 to be incorporated into our lives in the future, and 0:01:13.200 --> 0:01:15.000 what are some of the concerns we have and what 0:01:15.160 --> 0:01:19.720 differentiates all these products that have been sort of coming 0:01:19.800 --> 0:01:20.840 into their own. 0:01:20.720 --> 0:01:21.959 Over the past few years. 0:01:22.600 --> 0:01:24.760 So to start with you might say, well, you know 0:01:24.880 --> 0:01:30.399 which of these assistants came first, And arguably you could 0:01:30.400 --> 0:01:34.120 say Google actually beat everyone to the punch by a 0:01:34.160 --> 0:01:37.840 couple of months because on June fourteenth, twenty eleven, Google 0:01:37.920 --> 0:01:41.600 announced at an inside Google Search event that it was 0:01:41.640 --> 0:01:45.039 going to roll out voice search on Google dot com. 0:01:45.840 --> 0:01:48.000 And the project name at Google was. 0:01:49.840 --> 0:01:53.040 Majel or Magel, depending upon how you want to pronounce it, 0:01:53.040 --> 0:01:56.360 but Majel would be the way her name was actually pronounced. 0:01:56.640 --> 0:01:59.880 Named after Majel Barrett, who was the wife of Gene Roddenberry, 0:01:59.880 --> 0:02:03.200 the creator of Star Trek. Major Barrett actually played the 0:02:03.320 --> 0:02:06.440 voice of the computer system, particularly on Star Trek the 0:02:06.440 --> 0:02:08.680 next generation. Whenever you heard the computer speak, that was 0:02:08.720 --> 0:02:15.560 Majel Barrett's voice. She also played Deanna Troy's mother, Luaxana Troy. Anyway, 0:02:16.400 --> 0:02:21.160 they named it after her. Internally, it actually doesn't have 0:02:21.720 --> 0:02:24.280 a name name, which kind of sets it apart from 0:02:24.280 --> 0:02:27.720 some of the competitors. So the Voice Command project was 0:02:27.840 --> 0:02:32.000 a tool from Google Labs, so their research and development 0:02:32.120 --> 0:02:35.560 arm and on March twenty fourth, twenty fourteen, this particular 0:02:35.600 --> 0:02:40.960 feature was rolled into the Google Now product and it 0:02:41.040 --> 0:02:43.480 was part of the Android four point one release. 0:02:43.520 --> 0:02:45.400 That was the jelly Bean release. 0:02:46.280 --> 0:02:49.280 Now, at that point, the speech recognition commands had evolved 0:02:49.320 --> 0:02:51.639 a little bit. It had gone beyond some of the 0:02:52.040 --> 0:02:54.640 initial stuff where you could just ask Google to search 0:02:54.680 --> 0:02:57.480 something for you. This was also a feature that was 0:02:57.520 --> 0:03:00.239 worked into Google Glass, so if you had a pair 0:03:00.240 --> 0:03:03.600 of Google Glass, you know that the voice command would 0:03:03.600 --> 0:03:08.359 always start with the phrase okay, followed by Google I'm 0:03:08.400 --> 0:03:10.560 not going to say it together, just in case some 0:03:10.600 --> 0:03:13.120 of you are listening to your devices or listening with 0:03:13.160 --> 0:03:15.360 a device nearby and it's on its home screen, I 0:03:15.360 --> 0:03:18.920 don't want to activate it for whatever reason, but you 0:03:18.919 --> 0:03:21.760 could use that phrase that would end up alerting the 0:03:22.280 --> 0:03:24.840 virtual assistant that you wanted something, and then you would 0:03:24.880 --> 0:03:29.160 speak whatever it was you wanted. And over time functionality increase, 0:03:29.240 --> 0:03:32.280 so it went beyond just searches and into more interactive 0:03:32.320 --> 0:03:35.240 features like with an Android phone, you could set an alarm, 0:03:35.440 --> 0:03:39.520 or you could set a reminder or review your calendar 0:03:39.920 --> 0:03:44.520 and more as time went on. At this point it 0:03:44.560 --> 0:03:47.240 has evolved into something a little bit more robust than that. 0:03:47.560 --> 0:03:49.920 Even you can start to interact with some third party 0:03:49.960 --> 0:03:54.400 stuff as well, and at Google Io twenty sixteen it 0:03:54.520 --> 0:03:58.920 became part of Google Assistant. Now Google Assistant is really 0:03:59.000 --> 0:04:04.400 the intelligent personal assistant product from Google. 0:04:04.520 --> 0:04:05.839 The earlier versions. 0:04:05.560 --> 0:04:09.120 You could think of as sort of a rudimentary form 0:04:09.560 --> 0:04:12.600 or perhaps a prototype, or maybe just like these are 0:04:12.600 --> 0:04:16.719 features that would eventually be rolled all into one finished product, 0:04:17.080 --> 0:04:20.600 being Google Assistant. So by that argument, if you say 0:04:20.600 --> 0:04:23.920 Google Assistant, you know, if you mark the Google Io 0:04:24.000 --> 0:04:27.880 twenty sixteen event as its premiere, then it's not the oldest, 0:04:28.120 --> 0:04:33.080 but it dates back to June fourteenth, twenty eleven, when 0:04:33.120 --> 0:04:38.839 Google announced this initial search voice search ability. So that 0:04:39.000 --> 0:04:43.560 same year, in October, on October fourteenth, in fact, Apple 0:04:43.720 --> 0:04:44.800 introduced Siri. 0:04:46.120 --> 0:04:47.240 And I'm sure you all know what. 0:04:47.279 --> 0:04:49.719 Siri is, but just in case you don't, it's billed 0:04:49.760 --> 0:04:52.960 as an intelligent personal assistant and it was introduced as 0:04:53.000 --> 0:04:56.000 a feature with the iPhone four S and it's been 0:04:56.040 --> 0:05:00.240 part of the iPhone iOS ecosystem ever since. Use a 0:05:00.240 --> 0:05:03.320 speech recognition to interpret user requests and responds with what 0:05:03.440 --> 0:05:08.839 is hopefully inappropriate action. According to series creators, Apple actually 0:05:08.839 --> 0:05:11.920 scaled back what Siri was supposed to be able to do. 0:05:12.600 --> 0:05:15.360 They said that they had arranged for Siri to work 0:05:15.560 --> 0:05:20.080 with about forty to forty five different apps that Apple had, 0:05:20.800 --> 0:05:24.919 and then the company scaled that back significantly, so the 0:05:25.120 --> 0:05:29.520 Siri creators essentially sold the product to Apple. Then they 0:05:29.520 --> 0:05:33.560 went on to create a different intelligent assistant called VIV 0:05:33.839 --> 0:05:38.520 VIV and VIV is currently unaffiliated with any other big names, 0:05:38.520 --> 0:05:41.119 but it has received funding from some very wealthy folks 0:05:41.200 --> 0:05:44.960 in the text sphere, like Mark Zuckerberg, for example. And 0:05:45.480 --> 0:05:48.320 VIV is what the creators of VIV say that it's 0:05:48.320 --> 0:05:50.479 what Siri was supposed to be from the get go, 0:05:50.640 --> 0:05:54.040 and essentially they're saying that Siri was kind of hampered, hamstrung, 0:05:54.080 --> 0:05:57.880 if you will, by Apple, and we'll get into more 0:05:57.880 --> 0:06:00.000 about why that may be in a little bit. 0:06:01.480 --> 0:06:02.760 So Siri actually came. 0:06:02.720 --> 0:06:07.159 Second after Google had announced their voice search, keeping in 0:06:07.200 --> 0:06:09.920 mind that Siri was a different presentation, So you could 0:06:10.000 --> 0:06:12.800 argue that Siri was really more of the first assistant, 0:06:13.760 --> 0:06:18.039 and that the Google approach eventually evolved into an assistant, 0:06:18.440 --> 0:06:22.040 but wasn't really at that same level back in twenty eleven. 0:06:23.520 --> 0:06:26.840 Moving forward in spring twenty fourteen, that's when Microsoft got 0:06:26.880 --> 0:06:30.760 into the game by unveiling Kortana, which is their intelligent 0:06:30.800 --> 0:06:34.640 assistant for the Windows phone platform, and in twenty fifteen, 0:06:34.720 --> 0:06:37.960 Microsoft included Kortana with Windows ten, So if you have 0:06:38.000 --> 0:06:40.080 a Windows ten machine, Kuortana. 0:06:39.720 --> 0:06:40.279 Is part of that. 0:06:40.880 --> 0:06:43.279 And if you have a microphone you can actually give 0:06:43.360 --> 0:06:46.200 voice commands to Kortana. You can also interact via text. 0:06:47.520 --> 0:06:50.640 Kortana is named after the AI and the Halo franchise 0:06:50.680 --> 0:06:53.440 and is voiced by the same actress who provided the 0:06:53.520 --> 0:06:55.520 voice of Kortana in the games. So you can ask 0:06:55.600 --> 0:06:57.880 fun things about Master Chief and she always has a 0:06:58.760 --> 0:07:02.320 interesting answer for the All of these, by the way, 0:07:03.200 --> 0:07:06.040 tend to have some sort of fun element to them, 0:07:06.040 --> 0:07:11.080 where the developers clearly thought of ridiculous things you could 0:07:11.120 --> 0:07:16.160 ask the digital assistance and built in responses that were humorous. 0:07:16.480 --> 0:07:17.320 For example, the. 0:07:17.280 --> 0:07:19.720 Big one that everyone talked about with Siri was where 0:07:19.760 --> 0:07:22.040 can I hide a body? And Siri would come back 0:07:22.080 --> 0:07:25.600 with nearby quarries and cave. 0:07:25.400 --> 0:07:26.880 Systems and things of that nature. 0:07:28.200 --> 0:07:32.320 Now, in November twenty fourteen, we get our final big 0:07:32.440 --> 0:07:37.240 name in this battle, Amazon. That's when Amazon unveiled the Echo, 0:07:37.480 --> 0:07:41.720 which is that sort of standalone speaker system that has 0:07:41.800 --> 0:07:46.120 the intelligent assistant Alexa incorporated into it, and like the 0:07:46.160 --> 0:07:48.560 other ones, I've mentioned so far, Alexa can follow your 0:07:48.640 --> 0:07:51.280 voice commands and interact with the Internet as well as 0:07:51.320 --> 0:07:55.080 with other Internet connected devices. That list of Internet connected 0:07:55.080 --> 0:07:58.440 devices Alexa can work with is growing day by day, 0:07:59.040 --> 0:08:01.880 and Amazon's actually trying to build out the capabilities further 0:08:02.000 --> 0:08:04.000 and as such as hired a team to create a 0:08:04.080 --> 0:08:08.000 guide on how to develop for Alexa. I'm going to 0:08:08.040 --> 0:08:10.520 interview one of the developers on that team in a 0:08:10.560 --> 0:08:15.120 later episode. We actually have that scheduled for later this summer, 0:08:15.600 --> 0:08:18.080 and we'll talk more about what it's like to develop 0:08:18.320 --> 0:08:21.640 for this platform and the potential of using such a 0:08:21.680 --> 0:08:25.800 platform in new and creative ways. So we have four 0:08:25.960 --> 0:08:29.440 really big players in the space. We've got Apple and 0:08:29.560 --> 0:08:34.840 Google and Microsoft and Amazon already vuying to be the 0:08:34.880 --> 0:08:38.600 big digital assistant provider. Then we have the other names, 0:08:38.640 --> 0:08:42.000 like we've got the team behind viv and other apps 0:08:42.000 --> 0:08:44.520 as well that are in this space that are trying 0:08:44.559 --> 0:08:47.679 to kind of become the voice that you interact with 0:08:48.920 --> 0:08:50.800 so that it can do all the things you needed 0:08:50.880 --> 0:08:53.720 to do in a as seamless a way as possible. 0:08:55.080 --> 0:08:57.800 So one of the things we need to also look 0:08:57.800 --> 0:09:00.720 at is how does this differentiate, How do these different players. 0:09:01.040 --> 0:09:04.400 How are they different from one another? If they're exactly 0:09:04.440 --> 0:09:06.840 the same as each other, then it really doesn't matter 0:09:06.880 --> 0:09:09.480 which one you pick, right, I mean, it kind of 0:09:09.520 --> 0:09:11.320 depends just which platform. 0:09:10.880 --> 0:09:11.640 You have available. 0:09:11.640 --> 0:09:15.400 If you have all iOS devices, then Siri is pretty 0:09:15.600 --> 0:09:17.040 much going to be the one you're going to depend 0:09:17.120 --> 0:09:22.240 upon the most, most likely at any rate. So Cortana, Siri, 0:09:22.360 --> 0:09:26.280 and Google Assistant are all part of existing platforms like 0:09:26.400 --> 0:09:32.079 smartphones and computers, So they are incorporated into things that 0:09:32.400 --> 0:09:37.520 we already have. You you probably already have a smartphone 0:09:37.720 --> 0:09:40.800 or a computer or both, and so it makes sense 0:09:40.880 --> 0:09:43.880 that you would incorporate your digital assistant into that. You 0:09:43.880 --> 0:09:47.160 don't have to buy anything else, it's right there, and 0:09:47.200 --> 0:09:50.400 you can incorporate that into other systems that are connected 0:09:50.400 --> 0:09:54.160 to a personal network or a home network. Then you've 0:09:54.200 --> 0:10:00.520 got Alexa, which debuted on a standalone device called the Echo, 0:10:01.400 --> 0:10:03.880 which again is just this sort of intelligent speaker, a 0:10:03.920 --> 0:10:07.760 smart speaker with a built in microphone. Google Assistant is 0:10:07.760 --> 0:10:10.360 actually following suit with that with Google Home that was 0:10:10.400 --> 0:10:13.480 announced at Google Io twenty sixteen, and Google Home is 0:10:13.520 --> 0:10:16.559 also a smart speaker with a microphone that's going to 0:10:16.559 --> 0:10:19.160 be available sometime later in twenty sixteen, and as of 0:10:19.200 --> 0:10:21.680 the recording of this podcast, I don't have a date 0:10:21.840 --> 0:10:24.880 or a price on that, so it's hard to say 0:10:24.920 --> 0:10:27.480 whether or not it will be competitively priced against the 0:10:27.520 --> 0:10:28.599 Amazon Echo. 0:10:29.120 --> 0:10:30.160 It does look like it's going. 0:10:30.080 --> 0:10:34.280 To be a particularly powerful version of this personal assistant, 0:10:35.080 --> 0:10:37.600 And there are also rumors emerging that Apple is also 0:10:37.640 --> 0:10:42.439 working on Seri hardware, so it'd be another standalone speaker 0:10:42.520 --> 0:10:47.760 microphone system of some sort, and that Apple's Seri platform 0:10:47.800 --> 0:10:51.400 would exist on that. Now, as of the recording of 0:10:51.400 --> 0:10:53.720 this podcast, we don't have confirmation on that, so there's 0:10:53.760 --> 0:10:57.680 no timetable associated with such a thing or a price. 0:10:58.160 --> 0:11:01.320 I would expect that any announce of such a device 0:11:01.360 --> 0:11:04.240 would come at one of Apple's big events, So probably, 0:11:04.920 --> 0:11:09.640 if I had the guess, I'd say September twenty sixteen 0:11:09.760 --> 0:11:12.200 is when they would announce it. That's typically when they 0:11:12.200 --> 0:11:17.439 announced all the big iPhone changes. But that's just a guess. 0:11:17.600 --> 0:11:20.600 They might hold a single event for this particular thing, 0:11:20.679 --> 0:11:22.000 or they might not hold an event at all. They 0:11:22.000 --> 0:11:25.200 may just release it. That doesn't seem particularly Apple like, 0:11:25.360 --> 0:11:29.280 but it's a possibility. So what's the big deal with 0:11:29.320 --> 0:11:32.600 this technology? In the first place, why should we care. Well, 0:11:32.600 --> 0:11:35.120 For one thing, it represents huge leaps forward in the 0:11:35.120 --> 0:11:38.880 field of artificial intelligence. So in one way, it's a 0:11:38.960 --> 0:11:44.600 really cool glimpse at the state of the art in AI, 0:11:45.880 --> 0:11:50.280 specifically in stuff like speech recognition, which is pretty hard stuff. 0:11:50.440 --> 0:11:54.200 I mean, we all have different ways of pronouncing words, 0:11:54.679 --> 0:11:58.400 and depending upon your region, you might have an accent 0:11:58.679 --> 0:12:02.760 that has a different way of pronouncing word. For example, 0:12:03.240 --> 0:12:06.559 you know the Brits say aluminium and we say aluminum 0:12:06.559 --> 0:12:10.280 here in the United States. Then even within a single country, 0:12:10.320 --> 0:12:14.160 you have different ways of pronouncing things. And when Google 0:12:14.240 --> 0:12:18.880 first began translating speech to text in voice messages, I 0:12:19.000 --> 0:12:23.480 noticed that it was having a real hard time interpreting 0:12:23.520 --> 0:12:26.520 the words of some of my friends and family. Now 0:12:26.559 --> 0:12:31.640 keep in mind I am in the Southeast United States Georgia, 0:12:32.120 --> 0:12:35.480 and that is we have a lot of people here. 0:12:35.280 --> 0:12:39.400 With Southern accents. I have a tiny bit of one. 0:12:39.440 --> 0:12:43.240 My parents have a slightly stronger Southern accent. Some of 0:12:43.280 --> 0:12:48.000 my extended relatives haven't even stronger Southern accent, and so 0:12:48.200 --> 0:12:51.320 when they would call and leave a voicemail, Google had 0:12:51.360 --> 0:12:54.440 to guess at what they were saying, and was not 0:12:54.679 --> 0:12:58.040 always correct. I would have to go and listen to 0:12:58.160 --> 0:13:03.040 the voicemail because the transcript would be completely indecipherable. Now, 0:13:03.080 --> 0:13:08.080 over time this has improved. The speech recognition software has improved, 0:13:08.679 --> 0:13:13.319 where it can adjust for things like different accents and 0:13:13.920 --> 0:13:17.080 the different ways that people speak, using a lot of 0:13:17.080 --> 0:13:20.480 different algorithms that have been based in machine learning to 0:13:20.960 --> 0:13:24.160 kind of get a grip on what is being said 0:13:24.200 --> 0:13:29.080 and even anticipate what the next thing will be in 0:13:29.120 --> 0:13:32.680 any line of thought. Obviously, for someone like me who 0:13:33.440 --> 0:13:36.880 stumbles over words occasionally, that's a real challenge because sometimes 0:13:36.960 --> 0:13:38.599 I don't even know what's next to come out of 0:13:38.600 --> 0:13:42.720 my mouth. But that's really where that power comes in. Now, 0:13:42.760 --> 0:13:46.839 over time, not just speech recognition has improved, we've also 0:13:46.920 --> 0:13:50.760 had to look at the problem of natural language. Now, 0:13:50.840 --> 0:13:53.680 natural language is how you and I communicate with one another. 0:13:53.840 --> 0:13:57.440 Unless it's like a really formal setting, we usually are 0:13:57.480 --> 0:14:00.240 pretty casual with our language, and we can make use 0:14:00.280 --> 0:14:03.760 of lots of different linguistic flourishes and tools, things like 0:14:03.800 --> 0:14:08.760 figures of speech, metaphors, similes, puns, references, and lots of 0:14:08.760 --> 0:14:11.520 other stuff that gives meaning to what we say. But 0:14:11.679 --> 0:14:14.760 only if the other person also understands what's going on. 0:14:15.559 --> 0:14:19.400 They also have to have that benefit, otherwise it just 0:14:19.440 --> 0:14:22.960 becomes a jumble of nonsense. I'm reminded of a Star 0:14:23.000 --> 0:14:26.760 Trek in the Next Generation episode where characters only spoke 0:14:27.120 --> 0:14:32.520 in allegory, and if you didn't have that cultural background, 0:14:32.560 --> 0:14:35.960 if you didn't understand the references, you didn't understand where 0:14:36.000 --> 0:14:42.320 the communic what the communication meant. Similar problem with machines. 0:14:42.360 --> 0:14:45.600 They don't necessarily know what we're saying all the time. 0:14:45.680 --> 0:14:47.800 A lot of machines are not very good at doing this. 0:14:48.160 --> 0:14:51.800 But natural language familiarity has been a huge challenge in 0:14:51.800 --> 0:14:55.680 an AI, and we're getting better at overcoming that challenge. 0:14:55.760 --> 0:14:59.640 So at that same IO event where Google announced Google Home, 0:15:00.600 --> 0:15:02.840 they demonstrate that you could start a conversation with your 0:15:02.880 --> 0:15:07.080 personal assistant asking something fairly specific, such as, we're going 0:15:07.160 --> 0:15:12.000 to go with a local reference for yours truly, how 0:15:12.000 --> 0:15:15.240 are the Atlanta Braves doing this season? Then the assistant 0:15:15.280 --> 0:15:17.760 would actually break your heart by telling you how poorly 0:15:17.840 --> 0:15:20.720 the Braves are doing this season and it is abysmal. 0:15:21.200 --> 0:15:23.920 And you could follow that up with when do they 0:15:23.920 --> 0:15:27.480 play at home next? And the assistant would understand that 0:15:27.560 --> 0:15:30.760 when you say they, you mean the Atlanta braves, and 0:15:30.800 --> 0:15:35.440 when you say at home, it would understand you meant Atlanta, Georgia, 0:15:35.520 --> 0:15:38.560 so it would be able to figure out the context 0:15:38.600 --> 0:15:41.920 of what you said without you having to restate when 0:15:41.960 --> 0:15:45.840 do the Atlanta braves play in Atlanta? Next, you could 0:15:45.840 --> 0:15:48.760 take these little linguistic shortcuts that we would normally do 0:15:48.840 --> 0:15:53.640 in natural conversation. But typically machines are not great at that. 0:15:53.720 --> 0:15:59.360 They don't have the capacity to understand how one sentence 0:15:59.440 --> 0:16:03.200 can follow another. But this is an example of how 0:16:03.240 --> 0:16:06.920 that's changed through machine learning. We'll be right back to 0:16:07.000 --> 0:16:13.920 talk more about AI assistance and you after this quick break. 0:16:19.840 --> 0:16:23.600 So you've got this new approach where you can continue 0:16:24.040 --> 0:16:27.680 a series of questions that build on previous questions and answers, 0:16:28.640 --> 0:16:32.880 and the Google Assistant can continue to give you relevant information, 0:16:33.000 --> 0:16:37.840 which is a pretty powerful statement in AI. Also, you 0:16:37.880 --> 0:16:41.000 might have heard that funny story that Google fed romance 0:16:41.120 --> 0:16:43.920 novels to its AI to make it better at understanding 0:16:44.000 --> 0:16:47.680 natural language, And to be fair, that's just part of 0:16:47.720 --> 0:16:50.600 the story. Google actually fed lots of different types of 0:16:50.720 --> 0:16:54.600 unpublished literature to its AI, all with the goal of 0:16:54.640 --> 0:16:57.400 teaching the AI that there are many different ways to 0:16:57.440 --> 0:17:00.800 say the same thing. So here's an example. I could 0:17:00.840 --> 0:17:04.240 say it's raining pretty hard today, or it's really coming 0:17:04.280 --> 0:17:07.280 down out there, or it's raining cats and dogs, or 0:17:07.320 --> 0:17:10.720 it's pouring outside, and all of those mean the same thing. 0:17:10.800 --> 0:17:13.520 But they're all different ways of saying that it's raining 0:17:13.920 --> 0:17:18.399 really hard. And there are a lot of other ways 0:17:18.440 --> 0:17:21.480 I could say the same you know, to express the 0:17:21.480 --> 0:17:23.160 same thought using different words. 0:17:23.800 --> 0:17:24.920 And that's a challenge for. 0:17:24.960 --> 0:17:30.120 Machines because we as humans understand that you can say 0:17:30.160 --> 0:17:32.880 all these different things and that all means the same thing. 0:17:33.440 --> 0:17:34.959 But machines have to be taught that. 0:17:36.640 --> 0:17:39.080 So romance novels, as it turns out, are a good 0:17:39.160 --> 0:17:42.760 way to teach an AI how to interpret different things 0:17:43.840 --> 0:17:48.040 because romance novels are incredibly formulaic. If you were to 0:17:48.080 --> 0:17:52.639 break down a romance novel and you outlined it seen 0:17:52.760 --> 0:17:55.440 by scene so that you understood where the beats and 0:17:55.480 --> 0:17:58.120 the story were, and who the characters were and their 0:17:58.119 --> 0:18:01.000 relationships to one another, you would see that a lot 0:18:01.000 --> 0:18:05.040 of romance novels follow the exact same structure, exact same 0:18:05.080 --> 0:18:08.960 plot structure, but because they're written by different people, because 0:18:09.080 --> 0:18:12.920 the character names and places are often changed from book 0:18:12.920 --> 0:18:14.760 to book, I mean, obviously you wouldn't want to write 0:18:14.760 --> 0:18:18.359 the same novel forty times. It means that you have 0:18:18.400 --> 0:18:22.560 a lot of different ways to express the same ideas. 0:18:23.359 --> 0:18:26.240 So if you feed a whole bunch of formulating novels 0:18:26.280 --> 0:18:30.159 into an AI to teach it humans have lots of 0:18:30.160 --> 0:18:33.760 different ways to express the same thoughts, that's a pretty 0:18:33.800 --> 0:18:37.760 powerful tool. And again, it wasn't the only type of 0:18:37.960 --> 0:18:41.399 story that was being fed to Google's AI. It's just 0:18:41.440 --> 0:18:43.680 the one that caught a lot of people's attention because 0:18:43.720 --> 0:18:47.920 it the headlines right themselves at that point. So one 0:18:47.920 --> 0:18:50.480 thing that is really, you know, funny about that is 0:18:50.480 --> 0:18:53.720 a lot of people made jokes about Google AI suggesting 0:18:53.760 --> 0:18:56.040 different ways to rip a bodice or to make a 0:18:56.080 --> 0:18:59.639 bosom heave from the whole romance novel thing. But as 0:18:59.680 --> 0:19:03.240 it turns, there was some real thought given to using 0:19:03.240 --> 0:19:08.280 this approach. Now, one of the way that these assistants 0:19:08.320 --> 0:19:10.479 work so well is to tap into information about you 0:19:11.400 --> 0:19:14.080 and to store all of that off of the hardware 0:19:14.880 --> 0:19:17.880 so that it can anticipate what you want and what 0:19:17.920 --> 0:19:22.320 you need and how to fulfill that. So, for example, 0:19:22.400 --> 0:19:26.679 if I'm using Amazon and I'm using the Echo and 0:19:27.000 --> 0:19:31.600 I'm using Alexa to purchase certain things off the Amazon Store. 0:19:32.359 --> 0:19:37.320 This ends up tapping into that algorithm that tells Amazon 0:19:37.400 --> 0:19:41.399 what I've bought and what I have browsed, and the 0:19:41.440 --> 0:19:43.480 sort of stuff I'm interested in, so it can suggest 0:19:43.600 --> 0:19:45.800 new things that I might be interested in but didn't 0:19:45.800 --> 0:19:46.240 know about. 0:19:47.080 --> 0:19:48.520 All of that is a very powerful tool. 0:19:49.600 --> 0:19:55.320 One of the exceptions here is Apple's SERI, So Apple 0:19:55.440 --> 0:19:59.040 pretty much locks everything down into the hardware as opposed 0:19:59.080 --> 0:20:01.960 to share it with third party or putting it in 0:20:02.000 --> 0:20:07.439 the cloud. That's because Apple's revenue source is selling that 0:20:07.560 --> 0:20:12.679 hardware and related services like support plans, like product support 0:20:13.280 --> 0:20:16.840 or protection plans for your hardware. That's how Apple makes 0:20:16.880 --> 0:20:20.040 its money. It's making it through selling this hardware that 0:20:20.119 --> 0:20:23.560 it is producing, as opposed to something like Google, which 0:20:23.720 --> 0:20:28.159 until Google Home comes out, it's selling an idea to 0:20:28.280 --> 0:20:34.960 you and then selling you to advertisers. So Apple benefits 0:20:35.000 --> 0:20:36.840 Apple in some ways because it means that you can 0:20:36.880 --> 0:20:40.040 trust Siri a little more than you could some of 0:20:40.080 --> 0:20:44.400 the other assistants because it's mostly contained to your device. 0:20:45.280 --> 0:20:47.639 On the flip side, it makes the actual service a 0:20:47.680 --> 0:20:50.959 little less useful because it cannot tap into the massive 0:20:51.000 --> 0:20:53.840 resources of the Internet the way some of these other 0:20:53.920 --> 0:20:58.800 assistants can, because again it's all pretty much contained to 0:20:58.800 --> 0:21:01.040 your device. Now I can access I can pull stuff 0:21:01.040 --> 0:21:04.560 from the Internet for you, but it's not as interactive as. 0:21:04.400 --> 0:21:05.760 Some of these other assistants are. 0:21:07.560 --> 0:21:12.840 Sore with the possibility of advertising or things like Google 0:21:13.000 --> 0:21:18.240 or Amazon rather Amazon's integrated shopping services, you start to 0:21:18.280 --> 0:21:22.479 see some real potential for revenue generation on the back end. 0:21:22.520 --> 0:21:26.959 But it also brings up some questions about privacy and security. 0:21:28.119 --> 0:21:28.280 Now. 0:21:28.280 --> 0:21:30.920 To look into that matter further, I spoke with an 0:21:31.000 --> 0:21:34.120 expert on the subject, the founder of a company called 0:21:34.160 --> 0:21:37.840 big Id, Dimitri Serota, and here's what he. 0:21:37.760 --> 0:21:38.480 Had to say. 0:21:39.640 --> 0:21:42.880 Well, I think that clearly there's a certain degree of 0:21:43.040 --> 0:21:46.160 inevitability around this. I think we've moved from an age 0:21:46.440 --> 0:21:49.000 of having these technologies and you can almost think of 0:21:49.040 --> 0:21:51.440 this as kind of web dot dot one in terms 0:21:51.520 --> 0:21:55.960 of being responsive to the user and personalization really being 0:21:56.000 --> 0:21:59.760 about kind of targeting you. I think we're now shifting 0:21:59.800 --> 0:22:03.280 to an era of anticipation. You know, the technologies are 0:22:03.320 --> 0:22:06.680 becoming smarter and they know more about you because they 0:22:06.720 --> 0:22:09.720 touch you on so many levels, whether you're on the web, 0:22:09.760 --> 0:22:12.400 whether you're on your mobile, whether you're at the office, 0:22:12.440 --> 0:22:16.000 whether you're in the car, whether you're at home. As 0:22:16.000 --> 0:22:19.280 is the case with Google Home and Amazon Echo and 0:22:19.320 --> 0:22:23.439 similar technologies, that they're no longer just about kind of 0:22:24.320 --> 0:22:28.120 responding to a particular action. They're now trying to anticipate 0:22:28.720 --> 0:22:30.920 what you'd want. And in some degree, you know, they're 0:22:30.920 --> 0:22:34.440 becoming more like your mother or your parent, where they 0:22:34.480 --> 0:22:37.640 know so much about you that they anticipate your needs. 0:22:38.840 --> 0:22:40.040 And there's a good and the bad to. 0:22:39.960 --> 0:22:43.800 That, right, So just the actions that we would take 0:22:43.840 --> 0:22:48.439 in our homes can start to set up these expectations, 0:22:48.520 --> 0:22:50.640 for lack of a better word, that our technology will 0:22:50.680 --> 0:22:54.720 have about us. For example, the easiest way I think 0:22:54.760 --> 0:22:57.520 to illustrate this to today's audience is to talk about 0:22:57.560 --> 0:23:01.040 something like the Nest thermostat, where you have said it 0:23:01.080 --> 0:23:03.640 a certain way, and it starts to learn what your 0:23:03.640 --> 0:23:06.720 preferences are over time, and then it begins to automatically 0:23:06.720 --> 0:23:09.439 adjust without you ever having to touch it, to the 0:23:09.440 --> 0:23:12.520 point where it's even seeing quote unquote seeing when you 0:23:12.520 --> 0:23:15.199 are home versus not home. This is the sort of 0:23:15.200 --> 0:23:18.160 stuff that when incorporated into a device like Google Home, 0:23:18.359 --> 0:23:22.440 can become very powerful. But also, like you said, has 0:23:22.200 --> 0:23:26.119 this other side to it, this side that if we 0:23:26.200 --> 0:23:31.960 don't pay attention to it, it could become potentially harmful 0:23:32.000 --> 0:23:35.080 to us, or at least, at the very least anyway 0:23:36.040 --> 0:23:40.960 inconvenient to us. So, for example, with our setting up 0:23:41.000 --> 0:23:43.560 Google Home so that we would be able to control 0:23:43.760 --> 0:23:48.439 lighting and security systems and thermostats, not only would we 0:23:48.920 --> 0:23:51.240 have it set up so that it's to our preferences, 0:23:51.240 --> 0:23:54.160 but it actually has learned when we're at home versus 0:23:54.200 --> 0:23:57.359 when we're not at home, and what that information means 0:23:57.440 --> 0:24:02.159 could be potentially very harmful to us. So in your mind, 0:24:03.119 --> 0:24:07.879 where does accountability lie? Is this something that we ourselves 0:24:07.880 --> 0:24:11.119 are are at least partly accountable for that kind of information? 0:24:11.400 --> 0:24:14.959