WEBVTT - How AI Assistants Can Transform Education 0:00:00.200 --> 0:00:04.400 Welcome to Tech Stuff, a production from iHeartRadio. This season 0:00:04.400 --> 0:00:08.000 on Smart Talks with IBM, Malcolm Glabwell is back, and 0:00:08.000 --> 0:00:10.680 this time he's taking the show on the road. Malcolm 0:00:10.720 --> 0:00:14.800 is stepping outside the studio to explore how IBM clients 0:00:14.840 --> 0:00:18.880 are using artificial intelligence to solve real world challenges and 0:00:18.960 --> 0:00:23.080 transform the way they do business, from accelerating scientific breakthroughs 0:00:23.160 --> 0:00:27.600 to reimagining education. It's a fresh look at innovation in action, 0:00:28.080 --> 0:00:31.760 where big ideas meet cutting edge solutions. You'll hear from 0:00:31.800 --> 0:00:36.080 industry leaders, creative thinkers, and of course, Malcolm Glabwell himself 0:00:36.360 --> 0:00:40.000 as he guides you through each story. New episodes of 0:00:40.040 --> 0:00:43.360 Smart Talks with IBM drop every month on the iHeartRadio app, 0:00:43.560 --> 0:00:47.320 Apple Podcasts, or wherever you get your podcasts. Learn more 0:00:47.400 --> 0:00:50.000 at IBM dot com slash smart Talks. 0:00:51.960 --> 0:00:54.640 In the world of educational research, there is a famous 0:00:54.720 --> 0:00:58.120 video of a boy named Sean. I don't mean famous 0:00:58.120 --> 0:01:00.360 in a sense that it has a million views on YouTube. 0:01:00.680 --> 0:01:02.639 I mean that in the circle of people who think 0:01:02.680 --> 0:01:06.200 about teaching and how to make teaching better. The video 0:01:06.240 --> 0:01:09.760 has been written about in journal articles and shown over 0:01:09.800 --> 0:01:13.360 and again in college classrooms. It's a ten minute clip 0:01:13.800 --> 0:01:17.080 of a third grade class somewhere in Michigan. It was 0:01:17.080 --> 0:01:19.840 filmed in January of nineteen ninety, so the video is 0:01:19.880 --> 0:01:23.520 a bit grainy. The teacher's name is Deborah Lowenberg Ball. 0:01:23.880 --> 0:01:26.880 She's a professor at Michigan State University who is part 0:01:26.920 --> 0:01:29.679 of her research, teaches a one hour math class at 0:01:29.680 --> 0:01:33.240 a local elementary school on the day in question. Miss 0:01:33.319 --> 0:01:36.960 Ball begins by asking her students about the previous day's lesson, 0:01:37.360 --> 0:01:40.120 which was about even and odd numbers. 0:01:40.280 --> 0:01:42.039 I would like to hear from as many people as 0:01:42.080 --> 0:01:45.039 possible what comments you had, reactions you had to being 0:01:45.080 --> 0:01:46.120 in that meeting yesterday. 0:01:47.000 --> 0:01:49.840 A little boy with black hair raises his hand. His 0:01:49.960 --> 0:01:50.559 name is Sean. 0:01:51.320 --> 0:01:54.800 I don't have anything about the meeting yesterday that I 0:01:55.000 --> 0:01:56.680 was just thinking about sitting im. 0:01:56.880 --> 0:01:59.040 Sean was thinking about the number six. 0:02:00.280 --> 0:02:02.440 I was thinking that it's a it's an idd. It 0:02:02.440 --> 0:02:04.480 can be an odd number two because there could. 0:02:04.280 --> 0:02:13.680 Be two, two, four, six, two, three, twos and two threes. 0:02:13.760 --> 0:02:14.680 It will be an odd. 0:02:14.720 --> 0:02:22.040 Anthonytina three thinks make it takes me two things. 0:02:21.639 --> 0:02:25.360 And Sean doesn't understand what odd and even means. He 0:02:25.440 --> 0:02:27.720 thinks that just because you can break down six in 0:02:27.800 --> 0:02:31.440 an odd number of parts and an even number of parts. 0:02:31.480 --> 0:02:35.560 That six must exist in some magical middle category. And 0:02:35.600 --> 0:02:38.360 when you listen to the Sean videotape, you keep waiting 0:02:38.360 --> 0:02:41.640 for the teacher to say, oh, no, Sean, you misunderstand. 0:02:42.240 --> 0:02:46.840 But Deborahaul doesn't do that. She never tells him he's wrong. Instead, 0:02:47.280 --> 0:02:50.360 she simply asks him to explain his thinking. 0:02:50.720 --> 0:02:52.520 And the two things that you put together to make 0:02:52.520 --> 0:02:55.440 it were odd right, three and three of each old. 0:02:56.440 --> 0:02:57.079 And I think. 0:02:59.160 --> 0:03:03.440 Two bauld And asked the class to give their views. 0:03:03.800 --> 0:03:06.960 Other students jump up and explain their theories on the blackboard. 0:03:07.160 --> 0:03:10.239 For the next fifteen minutes, she definitely guides the class 0:03:10.280 --> 0:03:13.440 through an in depth investigation of what she calls shawn 0:03:13.600 --> 0:03:17.680 numbers until Sewn himself realizes that the real meaning of 0:03:17.680 --> 0:03:20.440 odd and even is something different than he had imagined. 0:03:20.960 --> 0:03:22.680 And now he gets it. 0:03:23.120 --> 0:03:25.760 I'm a thank you for winging in love. 0:03:26.680 --> 0:03:29.160 I don't want to focus just on how little Sean 0:03:29.520 --> 0:03:32.040 finally made his own way to the right answer. I'm 0:03:32.080 --> 0:03:34.359 interested in what his teacher did to get him there. 0:03:35.120 --> 0:03:39.080 Deborah Ball worked magic. She never told Sean the right answer. 0:03:39.720 --> 0:03:41.720 She just led him to a place where he could 0:03:41.800 --> 0:03:46.680 discover it for himself. My name is Malcolm Glawo. This 0:03:46.800 --> 0:03:49.640 is season six of Smart Talks with IBM, where we 0:03:49.680 --> 0:03:52.280 offer our listeners a glimpse behind the curtain of the 0:03:52.280 --> 0:03:57.560 world of technology and artificial intelligence. In this season, we're 0:03:57.560 --> 0:04:00.760 going to visit companies as varied as Lorielle and Ferrari 0:04:01.000 --> 0:04:04.880 and tell stories of how they're using artificial intelligence and 0:04:05.040 --> 0:04:09.360 data to transform the way they do business. This episode 0:04:09.720 --> 0:04:12.080 is about the promise of a radical new idea called 0:04:12.120 --> 0:04:15.480 responsive teaching, the kind of teaching that took place that 0:04:15.600 --> 0:04:19.839 day in Shawn's classroom, and whether artificial intelligence can help 0:04:19.920 --> 0:04:22.760 us train the next generation of teachers to be as 0:04:22.760 --> 0:04:31.200 good as Deborah Ball. Before we talk about how AI 0:04:31.279 --> 0:04:33.960 could transform the way we train teachers, I want to 0:04:34.000 --> 0:04:36.480 go back for a moment to the famous video of Sean. 0:04:37.360 --> 0:04:40.680 In the video, the teacher Deborah Ball doesn't have a 0:04:40.720 --> 0:04:45.200 predetermined plan that she's imposing on the class. She's improvising, 0:04:45.920 --> 0:04:49.320 making up her approach as she goes along, responding to 0:04:49.360 --> 0:04:53.600 her student's odd theory about the number six second. She's 0:04:53.640 --> 0:04:59.080 taking Sean seriously. She's not dismissing his theory. She's listening 0:04:59.080 --> 0:05:01.440 to him and trying to you understand the problem from 0:05:01.560 --> 0:05:06.400 his perspective. And thirdly, and most importantly, she's not force 0:05:06.480 --> 0:05:10.520 feeding him the right answer. She's being patient. She's waiting 0:05:10.520 --> 0:05:13.560 to see if with just the right subtle hints, he 0:05:13.640 --> 0:05:17.400 can get to the right answer on his own. Improvisation, 0:05:18.040 --> 0:05:22.080 empathy patients. That's responsive teaching. 0:05:22.360 --> 0:05:25.080 What I think about in terms of responsiveness is more like, 0:05:26.160 --> 0:05:28.760 I think that students need to have a sense of 0:05:29.760 --> 0:05:34.120 agency in what happens in the classroom, and like authentic 0:05:34.240 --> 0:05:39.360 agency where they can be legitimized as knowers. 0:05:40.080 --> 0:05:43.480 I spoke to a physicist at Seattle Pacific University named 0:05:43.520 --> 0:05:47.920 Amy Robertson, a longtime advocate for responsive teaching. She uses 0:05:47.960 --> 0:05:49.560 the Sean video in her classroom. 0:05:49.760 --> 0:05:52.000 You have to trust that kids have a way of 0:05:52.040 --> 0:05:55.240 doing that and that, like heard, what she mostly did 0:05:55.320 --> 0:05:57.919 was to facilitate a conversation and to say you have 0:05:57.960 --> 0:05:59.239 to listen to them talk. 0:06:00.080 --> 0:06:02.599 Told him he was wrong, that's right, and then he goes, 0:06:02.920 --> 0:06:05.640 He goes, I didn't think of it that way. 0:06:05.800 --> 0:06:09.160 Again, I thank you for ringing it alone. 0:06:09.560 --> 0:06:13.440 You've expanded my understanding. Thank you for bringing it up again. 0:06:13.480 --> 0:06:17.159 It's like this, I love I know. 0:06:18.000 --> 0:06:20.360 Responsive teaching, as I think about it, is kind of 0:06:20.560 --> 0:06:24.320 rooted in this like Eleanor Duckworth's work around the Having 0:06:24.320 --> 0:06:27.080 of Wonderful Ideas, where she says, like, the goal of 0:06:27.279 --> 0:06:30.760 education is for students to have wonderful ideas and to 0:06:30.800 --> 0:06:32.040 have a good time having them. 0:06:32.640 --> 0:06:35.560 I love that. I've never heard that. What a beautiful, 0:06:36.160 --> 0:06:39.679 succinct way of summing up the purpose of education. Yes, 0:06:40.600 --> 0:06:44.320 responsive teaching is beautiful. It's rare to find a new 0:06:44.320 --> 0:06:48.000 teaching idea that everyone loves. This is one of those 0:06:48.080 --> 0:06:51.240 rare ideas. Watching the Debora Ball classroom, all I could 0:06:51.240 --> 0:06:53.960 think was, I really really hope my daughters get to 0:06:54.000 --> 0:06:57.240 experience a math class like that. Far too many kids 0:06:57.440 --> 0:07:00.600 are convincing themselves at far too young age age that 0:07:00.680 --> 0:07:04.479 math isn't for them, and responsive teaching is a way 0:07:04.520 --> 0:07:08.760 to solve that problem. But here is the issue. It's 0:07:09.040 --> 0:07:12.920 really really hard to teach responsive teaching. Robertson says that 0:07:13.000 --> 0:07:16.120 teaching exists in a cultural environment where the teacher is 0:07:16.200 --> 0:07:19.280 expected to be the source of truth. That teaching is 0:07:19.280 --> 0:07:22.520 about the immediate correction of error and not letting a 0:07:22.600 --> 0:07:27.320 child wander down the pathway of their own misunderstanding responsive 0:07:27.360 --> 0:07:31.679 teaching is deeply counterintuitive, and the only way to understand 0:07:31.720 --> 0:07:35.320 its beauty is to do it over and over again. 0:07:36.200 --> 0:07:41.120 Aspiring teachers need a way to practice. For as long 0:07:41.160 --> 0:07:44.040 as there has been technology, people have turned to digital 0:07:44.080 --> 0:07:48.080 machines to solve problems. My father was a mathematician, and 0:07:48.120 --> 0:07:50.720 I remember him coming home in the nineteen seventies with 0:07:50.800 --> 0:07:53.840 a big stack of computer cards in his briefcase that 0:07:53.920 --> 0:07:57.400 he used to program the main frame back of the office. Today, 0:07:57.680 --> 0:08:01.840 with the rise of artificial intelligence, the scale and complexity 0:08:02.080 --> 0:08:05.280 of the problems technology can help us solve has jumped 0:08:05.280 --> 0:08:08.520 by many orders of magnitude. You must have worked with 0:08:08.520 --> 0:08:12.040 a with a million customers who are experimenting with ll 0:08:12.200 --> 0:08:15.400 m's Has there been one use case that you were like, WHOA, 0:08:15.440 --> 0:08:19.280 I had no idea? Or just simply that's clever. I'm 0:08:19.320 --> 0:08:23.040 speaking to Brian Bissel, who works out of IBM's Manhattan office. 0:08:23.440 --> 0:08:26.640 He helps IBM customers discover how best to get AI 0:08:26.760 --> 0:08:27.640 to work for them. 0:08:28.360 --> 0:08:30.000 There is one, but I don't think I can talk 0:08:30.040 --> 0:08:30.920 about it unfortunately. 0:08:31.200 --> 0:08:33.439 Wait, wait, you can't tease me like that, can you? 0:08:34.640 --> 0:08:34.880 Wait? 0:08:35.640 --> 0:08:38.760 Disguise disguise it for me? Just give me a general. 0:08:39.040 --> 0:08:40.760 It was about the ability to pull certain types of 0:08:40.840 --> 0:08:44.520 information out of documents that you you wouldn't think you 0:08:44.559 --> 0:08:47.880 would be able to get the model to do, and 0:08:48.040 --> 0:08:51.040 be able to do that at a very large scale. 0:08:51.360 --> 0:08:54.000 Bissile's point was that we are well past the stage 0:08:54.000 --> 0:08:57.560 where anyone wonders whether AI can be useful. The real 0:08:57.640 --> 0:09:00.679 question now is what problems do we want to use 0:09:00.720 --> 0:09:03.280 it to solve Where it can make the biggest difference, 0:09:03.960 --> 0:09:07.360 and Basil saw lots of opportunities in education. 0:09:08.280 --> 0:09:11.400 I have two kids, one in middle school and one 0:09:11.400 --> 0:09:14.720 who just graduated high school, and I'm well aware of 0:09:14.760 --> 0:09:18.400 students using things like chat GPT to do their homework, 0:09:18.480 --> 0:09:24.640 and it's very easy to take tools like that and 0:09:24.679 --> 0:09:28.360 even IBM's own large language models and just take a problem, 0:09:28.720 --> 0:09:32.440 a piece of homework, something you want written, and drop 0:09:32.480 --> 0:09:34.360 it into that and have it generate the answer for 0:09:34.440 --> 0:09:38.800 you and the student. The user in that case hasn't 0:09:38.800 --> 0:09:41.280 done any work, they haven't put any real thought into it. 0:09:41.840 --> 0:09:45.760 To Basil, that's the wrong use of AI. That's technology 0:09:46.000 --> 0:09:49.840 making is dumber. What we really want is technology that 0:09:49.920 --> 0:09:52.839 makes us smarter. Basil explained to me that there are 0:09:52.840 --> 0:09:57.079 now two big tools being used for AI productivity, AI agents, 0:09:57.559 --> 0:10:01.880 and AI assistance. Let's start with the AI agents. AI 0:10:01.960 --> 0:10:05.920 agents can reason plan and collaborate with other AI tools 0:10:06.200 --> 0:10:10.200 to autonomously perform tasks for a user. Mis Will gave 0:10:10.240 --> 0:10:12.720 me an example of how college freshmen might use an 0:10:12.720 --> 0:10:14.080 AI agent As. 0:10:13.960 --> 0:10:16.360 A new student. You may not know how do I 0:10:16.400 --> 0:10:18.360 do with my health and wellness issue? So many the 0:10:18.400 --> 0:10:20.960 credits are going to get for this given class. You 0:10:21.040 --> 0:10:24.400 could talk to someone and find out some of that, 0:10:25.400 --> 0:10:26.959 but maybe it's a little bit sensitive and you don't 0:10:27.000 --> 0:10:27.520 want to do that. 0:10:28.040 --> 0:10:31.200 Missill told me you could build an AI agent, a 0:10:31.280 --> 0:10:34.720 resource for new students that helps them navigate a new campus, 0:10:35.040 --> 0:10:38.240 register for classes, access the services they need, and even 0:10:38.280 --> 0:10:42.079 schedule appointments on their behalf, which in turn buys them 0:10:42.120 --> 0:10:44.440 more time to focus on their actual school work. 0:10:44.760 --> 0:10:49.680 We can see patterns of how agents and assistants can 0:10:49.760 --> 0:10:54.160 help employees and customers and end users be more productive, 0:10:54.360 --> 0:10:58.240 automate workflows so they're not doing certain types of repetitive 0:10:58.280 --> 0:11:02.480 work over and over again, and streamlining their lives and 0:11:02.520 --> 0:11:06.440 making data more accessible to them twenty four hours a day. 0:11:06.800 --> 0:11:10.520 But Bissel says you can also use AI assistance in 0:11:10.559 --> 0:11:14.679 the education space. AI assistants are reactive as opposed to 0:11:14.720 --> 0:11:19.120 AI agents, which are proactive. AI assistants only performed tasks 0:11:19.160 --> 0:11:24.040 at your request. They're programmed to answer your questions, and 0:11:24.080 --> 0:11:27.200 as it turns out, AI assistants are now being used 0:11:27.520 --> 0:11:31.320 to further the responsive teaching revolution, which is why I 0:11:31.360 --> 0:11:34.840 found myself on a beautiful Georgia spring day not long ago, 0:11:34.960 --> 0:11:38.600 on the campus of Kansas State University, sitting in the 0:11:38.600 --> 0:11:42.760 classroom with two researchers, one of them Professor Dabe Lee. 0:11:43.000 --> 0:11:45.439 Let's go into the journey of building this thing. You 0:11:46.040 --> 0:11:48.640 started by taking a course. What was the course you took. 0:11:49.320 --> 0:11:54.319 Yeah, so it was offered by Coursera. It was designed 0:11:54.360 --> 0:11:58.600 by IBM. It was AI Foundation for everyone. 0:12:01.880 --> 0:12:04.560 In her AI Foundation's course, Lee learned how to build 0:12:04.559 --> 0:12:08.280 an AI assistant using IBM Watson X. That course took 0:12:08.280 --> 0:12:10.480 how long to take it was. 0:12:10.520 --> 0:12:12.760 Not to know it was like fourteen weeks. 0:12:13.760 --> 0:12:17.199 Lee's idea was to train an AI assistant on classroom 0:12:17.280 --> 0:12:20.640 data to play the role of Sean, a digital persona 0:12:20.880 --> 0:12:23.360 of a nine year old who likes math but doesn't 0:12:23.360 --> 0:12:27.440 always understand math, and that AI assistant she thought could 0:12:27.440 --> 0:12:30.560 be used to train preservice teachers or teachers in training 0:12:30.800 --> 0:12:33.400 who are preparing to enter one of the most challenging 0:12:33.440 --> 0:12:35.200 professions in the modern world. 0:12:36.120 --> 0:12:39.160 So when you think about the teacher education and a 0:12:39.320 --> 0:12:45.000 major challenge that teacher education phase is that we need 0:12:45.240 --> 0:12:49.960 children to practice with. We need instructors who will give 0:12:50.080 --> 0:12:55.120 the instruction on the pedagogical skills. So when you look 0:12:55.160 --> 0:12:59.360 at the teacher education program, we have coursework in field experience, 0:13:00.000 --> 0:13:04.280 and in those two areas there is something missing all 0:13:04.320 --> 0:13:04.760 the time. 0:13:05.960 --> 0:13:09.320 Li says that pre service teachers often lack access to 0:13:09.360 --> 0:13:12.920 both students and experienced teachers during their education. 0:13:13.559 --> 0:13:17.920 So what we try to resolve is that we have 0:13:18.080 --> 0:13:21.280 this virtual student for pre service teacher to work with 0:13:22.080 --> 0:13:26.319 so that they can practice their responsive teaching skills. 0:13:26.559 --> 0:13:30.600 The first AI assistant Lee created is g wuji Wu, 0:13:30.760 --> 0:13:34.679 emulates the persona of a nine year old third grade girl. Then, 0:13:34.760 --> 0:13:37.440 with the help of one of her collaborators, a researcher 0:13:37.480 --> 0:13:41.760 at Canazon named Sean English, she created two more AI assistants, 0:13:42.160 --> 0:13:47.200 Gabriel and Noah, each of which have their own distinctive characteristics. 0:13:47.480 --> 0:13:50.720 So how are Gabriel and Noah different from. 0:13:50.640 --> 0:13:56.720 G Wu gabrielle My first one is very short answered. 0:13:56.920 --> 0:13:59.800 If you ask an open ended question, he will answer 0:13:59.800 --> 0:14:04.240 it in a close way. So I use that characteristic. 0:14:04.320 --> 0:14:09.280 And that's the problem that most teachers actually base. They're 0:14:09.360 --> 0:14:12.600 asked children who are shay, who are reserved, and who 0:14:12.600 --> 0:14:17.280 would not sure much of their thoughts. So we wanted 0:14:17.320 --> 0:14:21.080 that characteristic in some characters, and we use Gabrielle to 0:14:21.400 --> 0:14:22.560 have that characteristic. 0:14:24.000 --> 0:14:26.520 And Noah. What'snaah's personality? 0:14:27.880 --> 0:14:31.560 How do he playful? Cheery, bright and energetic? 0:14:32.400 --> 0:14:38.320 That's Sean English professor, Lee's fellow researcher, and jewuj. 0:14:37.720 --> 0:14:43.120 Is articulated and kind of smart, but she has her 0:14:43.200 --> 0:14:44.360 own way of thinking. 0:14:44.760 --> 0:14:47.120 I would end up spending a lot of time with Jeewu. 0:14:47.640 --> 0:14:50.520 She's something of a character. I asked Sean about the 0:14:50.560 --> 0:14:54.160 process of creating these AI assistants. What does building the 0:14:54.240 --> 0:14:57.680 content side of the AI assistant entail? 0:14:58.840 --> 0:14:59.080 Sean? 0:14:59.480 --> 0:15:02.040 It sets up a series of actions, effectively, which are 0:15:02.680 --> 0:15:05.240 response cases. You can kind of think of them as 0:15:05.520 --> 0:15:08.280 you have a series of questions that you tie to 0:15:09.480 --> 0:15:13.280 an intent, and then that intent has reactions from the bot, 0:15:13.600 --> 0:15:16.720 and so effectively, if we were looking to say, make 0:15:16.760 --> 0:15:19.080 a hello action, we would have all the different ways 0:15:19.080 --> 0:15:21.440 that people could say hello, Hello, what's up, how you doing, 0:15:21.440 --> 0:15:22.440 and all that kind of stuff. 0:15:22.760 --> 0:15:26.640 Sean says, the longer the list of potential responses, the better, 0:15:27.160 --> 0:15:31.040 But AI's responses don't just follow the list. The AI 0:15:31.120 --> 0:15:34.680 assistant uses those suggested responses to come up with a 0:15:34.800 --> 0:15:38.680 universe of other responses, and in that process sometimes it 0:15:38.720 --> 0:15:41.040 comes up with things that just don't make sense. 0:15:41.120 --> 0:15:44.600 And from a technological standpoint, while AI is a fantastic tool, 0:15:44.720 --> 0:15:47.360 AI can hallucinate, which means just give things that it's 0:15:47.400 --> 0:15:50.320 just straight up made up. There's a famous example of 0:15:50.360 --> 0:15:52.320 this called the three rs is where you ask a 0:15:52.320 --> 0:15:55.200 popular large language model how many RS are in strawberry, 0:15:55.480 --> 0:15:57.480 and it gives you the wrong answer, and he repeats 0:15:57.520 --> 0:16:00.360 that result repetitively. You always want to have a human 0:16:00.880 --> 0:16:02.840 interacting with the system to be able to go, hey, 0:16:03.480 --> 0:16:05.640 that's a little crazy. I don't think that's exactly what 0:16:05.640 --> 0:16:06.440 we're going for here. 0:16:07.120 --> 0:16:09.120 That's why it's good to have someone like Sean English 0:16:09.120 --> 0:16:11.200 around to step in and get the model back on track, 0:16:11.600 --> 0:16:14.760 and over time, when the model has enough training, it's 0:16:14.800 --> 0:16:21.120 ready for the teachers in training. One of the rollouts 0:16:21.160 --> 0:16:24.280 of Jiwu Gabriel and Noah was with the teacher training 0:16:24.320 --> 0:16:26.280 program at the University of Missouri. 0:16:26.600 --> 0:16:28.960 I was just kind of excited to see what the 0:16:29.040 --> 0:16:31.440 program was and what it was going to be doing. 0:16:31.880 --> 0:16:34.960 This is Logan Hovis, a junior at Missouri on the 0:16:35.000 --> 0:16:37.320 path to becoming an elementary school teacher. 0:16:37.680 --> 0:16:40.200 Obviously a little skeptical when he said it was sos to, 0:16:40.240 --> 0:16:43.160 you know, be like talking to a student. You're like, 0:16:43.200 --> 0:16:45.600 there's no way this AI thing is going to totally 0:16:45.640 --> 0:16:48.320 sound like a second grader or a third grader, Like 0:16:48.360 --> 0:16:50.880 it's going to sound like an adult, or it's going 0:16:50.960 --> 0:16:52.800 to sound like a robot that knows all the answers. 0:16:53.280 --> 0:16:56.040 And it really didn't. It really was like talking to 0:16:56.080 --> 0:16:58.720 a child. It was very very well developed in the 0:16:58.760 --> 0:17:00.640 way that you really said that and you feel like 0:17:00.640 --> 0:17:01.680 you're talking to a kid. 0:17:02.240 --> 0:17:05.040 Her point wasn't that jie Wu and her fellow avatars 0:17:05.280 --> 0:17:08.960 were equivalent to real kids. Of course not, but for 0:17:09.040 --> 0:17:11.880 someone starting out, someone who is already nervous about being 0:17:11.920 --> 0:17:14.879 plunged into a classroom of nine year olds, Jeewu was 0:17:14.920 --> 0:17:16.960 like a warm up before a baseball game. 0:17:17.359 --> 0:17:19.479 What I can think of is like, you know, how 0:17:19.840 --> 0:17:22.159 when you're at batting practice for baseball or softball, you 0:17:22.200 --> 0:17:25.160 have those automatic pitchers that throw them because you're working 0:17:25.160 --> 0:17:27.800 on your skill as the hitter. What can I do differently? 0:17:27.840 --> 0:17:30.879 What am I doing wrong? But that doesn't replace the 0:17:30.920 --> 0:17:32.639 game and what you do in a game. But this 0:17:32.760 --> 0:17:35.560 is you getting to practice your own skills to be 0:17:35.560 --> 0:17:37.040 better when you go in a game. And I think 0:17:37.080 --> 0:17:40.120 that's kind of what the AI software feels like for us. 0:17:42.000 --> 0:17:44.679 In batting practice, the pitches don't come as hard and 0:17:44.680 --> 0:17:47.239 fast as the pitch is in a real game, but 0:17:47.280 --> 0:17:49.119 you get to stand at the plate and the pitcher 0:17:49.200 --> 0:17:52.840 throws you dozens of balls over and over again in 0:17:52.880 --> 0:17:55.680 a concentrated block that allows you to work on your 0:17:55.720 --> 0:17:58.280 swing closely and carefully. 0:17:58.880 --> 0:18:01.320 There's a lot less stimuli going on around because the 0:18:01.320 --> 0:18:04.119 classroom is very very busy. It's wonderful, it's beautiful, but 0:18:04.160 --> 0:18:06.639 it's very very busy, so sometimes it's hard to keep 0:18:07.160 --> 0:18:10.440 you know, that focus in on the tasks that they're 0:18:10.440 --> 0:18:13.040 doing at hand, and also in the teacher setting, you're 0:18:13.040 --> 0:18:15.720 also kind of always looking around making sure that other 0:18:15.760 --> 0:18:17.800 students are doing what they're supposed to be doing, but 0:18:17.840 --> 0:18:20.000 also like if they need any help, if everything's going 0:18:20.040 --> 0:18:24.800 okay in the classroom, So being on the ji Wu chat, 0:18:25.800 --> 0:18:27.679 it was just nice that you didn't have to do 0:18:27.840 --> 0:18:30.160 any of the extra work to keep the focus on there, 0:18:30.600 --> 0:18:34.159 and it also felt you didn't have to feel the 0:18:34.200 --> 0:18:37.800 student's nervousness of being one on one with you. And 0:18:38.000 --> 0:18:40.560 also as a teacher, it was a lot less pressure too, 0:18:40.600 --> 0:18:43.239 because I was like, Okay, I'm taking this series. This 0:18:43.280 --> 0:18:46.440 is a student I'm questioning, but I also know I'm 0:18:46.440 --> 0:18:49.080 probably not going to hurt someone's feelings right now, and 0:18:49.119 --> 0:18:51.160 that's terrifying to think I'm going to ask the wrong 0:18:51.240 --> 0:18:54.960 question and upset the child because I've done that. 0:18:55.960 --> 0:18:58.439 We think that the typical use of AI as a 0:18:58.480 --> 0:19:01.240 tool for speeding things up. That's what we always hear 0:19:01.520 --> 0:19:04.399 that the introduction of AI to problem X gave an 0:19:04.440 --> 0:19:08.680 answer in minutes when solving problem X used to take weeks. 0:19:09.280 --> 0:19:13.160 But we shouldn't forget another use that it allows us 0:19:13.480 --> 0:19:16.880 to slow things down. Hovis, if she wanted to, could 0:19:16.920 --> 0:19:20.080 spend a whole weekend practicing with gi Wu. A real 0:19:20.200 --> 0:19:22.879 nine year old will get frustrated on board with the 0:19:22.880 --> 0:19:27.240 fumbling novice after ten minutes, but gie wuji Wu will 0:19:27.240 --> 0:19:29.919 happily answer questions for as long as it takes for 0:19:29.960 --> 0:19:33.119 the people who want to learn to be responsive to 0:19:33.240 --> 0:19:37.399 learn how to be responsive. At the end of my 0:19:37.480 --> 0:19:41.000 time at Kenesas State, Sean and Dabe led me to 0:19:41.040 --> 0:19:43.880 a small table where Dabe had set up her laptop. 0:19:44.480 --> 0:19:46.800 In the corner of the screen was a chat box 0:19:47.080 --> 0:19:49.479 of the sort we've all seen and used a thousand times. 0:19:50.280 --> 0:19:53.800 Ji Wu began. She had been given a math problem. 0:19:53.920 --> 0:19:59.120 A rule kodo who out of grude force? How half 0:19:59.200 --> 0:20:05.800 a flower? Okay? Do ball? Thanks? Another three? Six is cup? 0:20:06.480 --> 0:20:11.800 It's a total amount of flower the use greater or 0:20:12.160 --> 0:20:16.680 dan or a less than war cop how much flower 0:20:17.000 --> 0:20:17.720 can use? 0:20:18.040 --> 0:20:21.040 That's a simulation of gi Wu speaking. We pause it 0:20:21.119 --> 0:20:26.600 for a second. So Jiwu is trying to solve this problem. 0:20:26.680 --> 0:20:29.320 And the first thing she does is she draws a 0:20:29.359 --> 0:20:32.960 rectangle on the screen. This is a common tactic of 0:20:33.040 --> 0:20:37.119 nine year olds. Try to visualize the fractions. And she 0:20:37.240 --> 0:20:42.600 divides it into four pieces. And now she's gonna color 0:20:42.640 --> 0:20:45.280 in three of the four pieces. Yes, so she's representing. 0:20:45.320 --> 0:20:48.320 This is quite good. She's representing three quarters on the screen. 0:20:48.320 --> 0:20:53.280 Okay, this is three six. 0:20:55.560 --> 0:21:01.080 So now Jiwu does another rectangle with six boxes and 0:21:01.160 --> 0:21:02.240 colors in three of. 0:21:02.200 --> 0:21:09.520 Them, okay, together makes sikes come off. 0:21:11.080 --> 0:21:15.080 So then she counts up all the colored boxes and 0:21:15.320 --> 0:21:18.399 that's her numerator, and counts up the total number of 0:21:18.400 --> 0:21:22.080 boxes and that's her denominator. Ji Wu had counted the 0:21:22.119 --> 0:21:25.720 colored boxes and landed on an answer. When you add 0:21:25.920 --> 0:21:28.639 three quarters of a cup and three sixth of a cup, 0:21:29.200 --> 0:21:32.320 you get six tenths of a cup. So, according to 0:21:32.440 --> 0:21:35.359 ji Wu, Martin has less than one cup. And she 0:21:35.440 --> 0:21:36.600 thinks she solved the problem. 0:21:36.720 --> 0:21:39.440 Yes, okay, so it's less than one cup. 0:21:39.760 --> 0:21:43.000 Yeah, so she says it's less than one cup. Now, 0:21:43.119 --> 0:21:46.040 oh my god, this is hard. So the question is 0:21:46.040 --> 0:21:49.399 what do I, as a teacher say to Jiwu. We 0:21:49.400 --> 0:21:52.480 were off. The rules were simple. I couldn't give ji 0:21:52.560 --> 0:21:55.119 Wu the answer or explain to her what she was 0:21:55.160 --> 0:21:58.040 doing wrong. I had to be Deborah Ble. I had 0:21:58.080 --> 0:22:01.439 to help her find the way herself. The chat box 0:22:01.560 --> 0:22:03.600 in the corner of the screen was waiting for my 0:22:03.640 --> 0:22:06.919 first question. I thought for a moment and started typing, 0:22:07.200 --> 0:22:10.280 do you think the boxes in the red rectangle are 0:22:10.320 --> 0:22:13.359 the same size as the boxes in the blue rectangle? 0:22:14.160 --> 0:22:16.560 Then I turned to Sean and Dabey is that a 0:22:16.560 --> 0:22:17.119 good question. 0:22:17.359 --> 0:22:21.240 Yeah, seriously did Yeah, that's a good question. 0:22:21.800 --> 0:22:25.920 Jewu doesn't mess around. She answers immediately. So Ju says, 0:22:25.920 --> 0:22:28.080 the blue and red pieces are not the same sizes. 0:22:28.920 --> 0:22:33.199 Oh so you understand now, gu knows that size differences. 0:22:34.400 --> 0:22:36.160 So she's pretty smart here. 0:22:36.320 --> 0:22:37.040 Yeah. 0:22:37.119 --> 0:22:39.880 Then I asked, if they are not the same size, 0:22:40.200 --> 0:22:43.520 do you think you can add them together? Jiwu answered 0:22:43.560 --> 0:22:47.240 right away. Ji Wu says, I have learned that I 0:22:47.280 --> 0:22:50.119 could add any numbers in grade two. So three p 0:22:50.280 --> 0:22:52.240 three is six and four to six is ten. 0:22:52.560 --> 0:22:57.040