1 00:00:00,340 --> 00:00:05,365 You have probably heard of Watson from it's famous appearance on Jeopardy 2 00:00:05,365 --> 00:00:10,082 in 2011, where it competed against previous champions and won. 3 00:00:10,082 --> 00:00:14,909 Or more recently, you might've seen commercials about Watson being 4 00:00:14,909 --> 00:00:18,932 used by H&R Block to help file taxes in the United States. 5 00:00:18,932 --> 00:00:20,940 >> Yeah, Armin, I've heard of these examples. 6 00:00:20,940 --> 00:00:24,770 But I didn't quite understand what it meant to be powered by Watson. 7 00:00:24,770 --> 00:00:27,600 It wasn't until we started to prepare for this course that I really kind of 8 00:00:27,600 --> 00:00:31,890 understood what's available and what other kinds of things people are working on. 9 00:00:31,890 --> 00:00:35,505 At first, I thought Watson was some kind of artificial intelligence system. 10 00:00:35,505 --> 00:00:38,340 That's partially true, isn't it? 11 00:00:38,340 --> 00:00:43,640 >> Yes, but it is really a platform of APIs that IBM has 12 00:00:43,640 --> 00:00:49,430 worked hard to train, that you can now use as a service for your own software. 13 00:00:49,430 --> 00:00:54,660 Essentially, though, Watson APIs allow you to process speech, 14 00:00:54,660 --> 00:00:58,860 text and images in an easy and natural way. 15 00:00:58,860 --> 00:01:03,840 IBM has trained these APIs on immense data sets, but you can also 16 00:01:03,840 --> 00:01:08,460 train them with your own data to make customized decisions and responses. 17 00:01:09,990 --> 00:01:12,780 It's important to note that Ginni Rometty, 18 00:01:12,780 --> 00:01:18,380 IBM's CEO, has coined the term augmented intelligence. 19 00:01:18,380 --> 00:01:24,050 This means that Watson is helping people and machine work together 20 00:01:24,050 --> 00:01:29,230 to create knowledge from data that enhances human expertise. 21 00:01:29,230 --> 00:01:29,730 >> That's awesome. 22 00:01:30,860 --> 00:01:34,799 >> Let me tell you about a recent application that some students at 23 00:01:34,799 --> 00:01:36,520 an MIT Hackathon created. 24 00:01:38,180 --> 00:01:42,730 The winning team at the MIT Hackathon used the Speech 25 00:01:42,730 --> 00:01:47,360 to Text and the Tone Analyzer services working in tandem. 26 00:01:48,450 --> 00:01:51,990 Their premise was the therapist's office, 27 00:01:51,990 --> 00:01:57,350 where couples would get counseling on their grievances and daily bickerings. 28 00:01:57,350 --> 00:02:01,120 With the therapist's permission, the conversation was recorded. 29 00:02:01,120 --> 00:02:04,040 The sound files were converted to text. 30 00:02:04,040 --> 00:02:07,486 The text was then fed into the tone analyzer service, 31 00:02:07,486 --> 00:02:11,425 where it depicted sentences by highlighting them in color. 32 00:02:11,425 --> 00:02:16,640 [SOUND] For example, the deep red color meant an angry tone. 33 00:02:16,640 --> 00:02:20,443 Slighter shades of red meant somewhat angry tone. 34 00:02:20,443 --> 00:02:25,192 Green stood for happy, yellow for sad, and so forth. 35 00:02:25,192 --> 00:02:30,533 The couple were able to identify sentences that triggered, 36 00:02:30,533 --> 00:02:36,616 that had become habitual and a de facto way of speaking to each other. 37 00:02:36,616 --> 00:02:41,420 And they practiced saying what they meant, except with a different tone. 38 00:02:41,420 --> 00:02:46,030 But not so combative or trenchant or sarcastic. 39 00:02:46,030 --> 00:02:49,600 This allowed the therapist to continue their session 40 00:02:49,600 --> 00:02:52,730 even after the office visit was over. 41 00:02:52,730 --> 00:02:57,324 They would practice at home answering some focused questions that 42 00:02:57,324 --> 00:03:02,749 the therapist depicted as trigger points for argumentative conversations. 43 00:03:02,749 --> 00:03:04,624 >> We're getting a little bit ahead of ourselves. 44 00:03:04,624 --> 00:03:08,618 But I think we can show you what you can do with some hands-on examples and 45 00:03:08,618 --> 00:03:10,280 some more stories like this. 46 00:03:10,280 --> 00:03:12,330 Let's take a look at a hands-on example in the next video.