Oct 16, 2019

Major advances in Artificial Intelligence are the ability of computers to understand text and speech, comprehend information, and respond to users. And an area that's becoming more obvious to us is 'intelligent agents' or chatbots, where we can have a fairly normal chat conversation and get information delivered to in a natural way. They talk about the impact on students, when future careers will be changed by chatbots.

Dan talks about bots he's built with schools in the past, whilst Ray begs somebody to build a "Why did I come in here?" bot. More seriously, they talk about ways that they are providing support for students, and Ray talks about why he doesn't agree with the Silicon Valley view that we can replace teachers with bots and other forms of AI.

Dan mentions Q & A Maker, which is a Microsoft bot service you can find at QnAMaker.ai, and Ray talks about the work with bots being done at UNSW that is supporting students in their learning journey, and there's more on that UNSW case study here

TRANSCRIPT FOR The AI in Education Podcast
Series: 1
Episode: 4

This transcript and summary are auto-generated. If you spot any important errors, do feel free to email the podcast hosts for corrections.

This podcast excerpt features hosts Ray and Dan discussing conversational AI and chatbots, moving on from a previous conversation about AI prediction. They begin by humorously brainstorming ideal chatbot applications, such as a "Best Before" food inventory bot that suggests recipes and a bot to counter the momentary forgetfulness after walking into a room. Dan defines a chatbot as an interaction with a computer through chat, often seen as an automated Frequently Asked Questions (FAQ) service on websites, which saves organisations money. The discussion then explores the ease of building a bot using a knowledge base of questions and answers, like those created via Microsoft's Q&A Maker, noting that a bot’s ability to understand natural language provides a better service than static FAQs, especially for international users. Finally, the hosts consider the impact of automation on jobs and the education sector, concluding that while AI can augment learning, the human element of teachers cannot be replaced because their role extends beyond imparting knowledge to include passion, mentoring, and well-being.

Well, hello again. It's me, Ray, and
Dan from Microsoft.
We're here to carry on our conversation from last time. Last time we were talking about the AI and being able to predict the future. This time we said we'd talk a bit about conversational AI
and chatbots. Oh, chatbot's exciting. If you if you could have a chatbot invented, Dan, what what would you have it for?
I definitely have a chatbot. Let's have a think. So, around meals for me probably because I kind of uh I'm a one trick pony for meals, right? I have to say I'm kind of a spaghetti bolognese and chicken nuggets with my kids. So, I've got a lot of things in the in the kind of cupboard and the fridge. And when I kind of look back at that and I think if I had a chatbot which could just tell me what food would be out of date or what kind of things I've got available to me, give me some inspiration. some ideas around what I could make today. I'm sure there would be a chatbot that is much better.
I didn't know you're not the first to have that idea because a little while ago there was the AI for good schools competition. All
right.
And I went and judged the Tasmania state finals and it was year 8 to 10 kids came up with some brilliant ideas and and one of them was I think it was called best before and it was a bot that looked at what you put in the fridge, told you what was coming up.
That's exactly what I want.
Told you what was just about to go out of date and gave you a recipe that used the things that were going to go out a day.
That's a good idea. And they got good food in Tasmania as well.
I can't imagine anybody that would not want a bot like that. Um, so what would I want as a bot? Oh, Dan, could you invent one called Why Did I Come in Here? Cuz look, I'm a little bit older than you.
Okay.
And so, uh, you know that bit where you're at home and you go, I need a screwdriver. I need, um, I need to go and send that email. And so, you go to another room
and you go and you stand there and you go, why did I come in here? I want a bot that does that for me. Because I remember reading some research that says it's the walking through the doorway that makes you forget.
So if I had something where I go, "Oh, I just go and fetch my reading glasses." And then I walk into the study and I go, "Why did I come in here?" And the bot says, "You came in here for your reading glasses."
That's interesting though, isn't it? It's It's Yeah, the walking through the door thing where you kind of mind just kind of goes into another zone. I love it.
I thought it was just me. Thank goodness.
No, it's me, too.
So that's a kind of we've jumped straight into, oh, let's build a bot, but what just explain to me what a chatbot is, Dan.
So, a chatbot is your interaction with uh I suppose a computer um through chat. So, you see them all the time in things like um insurance websites and university application forms or you think you're speaking to somebody at the end of a web page, but you're not. You know, it's usually like a little small window that appears where you can ask simple questions and and often that might feel as if you're talking to somebody. And I suppose over the last couple years you were. They would employ people at the end of a phone to kind of ring people up and you can kind of speak to them about particular things to do with your car insurance. But nowadays you can kind of code a lot of that and get those questions inside your your chatbot which kind of saves organizations on that kind of money. It's basically your FAQ, your frequently asked questions in an automated way.
So I'm guess I'm guessing you talk to schools about it.
Yeah. Yeah. We done quite a bit with schools around around this and actually get people to actually do one.
So you know I've got kids to do it as well and and essentially from Microsoft point of view when I'm talking to the schools I normally get them to first of all create the knowledge base so you need to know the questions and the answers. Yeah.
So we've got a website called Q&A maker.ai. Easy for me to say and basically that either you point that at a um FAQ site on a page. So some of the kids used to point at like a Minecraft FAQ page. So
uh you could ask questions about what's a zombie, what's a zombie pigman, all the kind of fun stuff that you get in Minecraft. But you get your FAQ page of your questions and your answers and essentially you create a knowledge base. Once you've got a knowledge base, a questions and answers, you can edit that and then you then create a bot inside cloud-based service to so for example Azure and you connect the two together. So you get get the kind of interface and you get the knowledge base. So you basically can type it in and you can do a you know I probably made it sound very complicated here in this podcast but it literally takes about 15 minutes and I did it live at Edute Tech and
Yep. in this is what the ideas that teachers come up with for those things. So once you show them how to make one on something kind of a little bit boring like I don't know an IT help desk suddenly they think ah that actually might not just be as boring as you make it sound Dan that I might actually save me and save the pain of my users a lot of you know time to actually find out reset their password or how to where to take their machine if it breaks and things like that. So we've had we had some good examples of that at Catholic schools New South Wales for example in in Sydney where they've kind of done uh just that an IT bot. So it's quite easy to do. Showing people to do it and then getting people to think about ideas and applications and that is where it's been quite fun.
I I remember making one um for a university just to show them what was possible. And I went to their website and I found they had 200 Q FAQ documents on different web pages.
Oh wow.
So I pointed the bot at five and it went and read those FAQs and turned it into Q&A. And it was the bit that amazed me was the the language. capability, the the fact that it was able to understand what I was asking a question about even though that question wasn't in the FAQ.
Yes.
And help me find the answer. So, so the example I I really remember because, you know, it shows how thick I am. When I moved to Australia,
I used to drive through French's Forest and up at French's Forest there's a sign that says uh to the aquatic center
and for two years I thought it was somewhere that sold fish tanks and goldfish
because that's what aquatic meant to me it was aquarium.
Yeah.
But of course aquatic center is a swimming pool in Australia which is you know in the UK it's called the swimming pool and so
maybe in the UK where you came from.
Yeah. Okay. One of the Q&As's from this university or one of the FAQ pages was for the aquatic center and out of 35 questions it only mentioned swimming pool once
and I said to the bot what time is the swimming pool open
and it told me that the aquatic center was open 9 till 9. And it was that It was like well how did it know that swimming pool means aquatic center
and it's the smarts behind that
that a really impressed me what it could do but secondly it open unlocked that idea of
actually a bot was providing a better service than an FAQ because if I was an international student
I wouldn't know the answer to the question
because I'd be looking for something completely different and the bot provided that interface for me to not only know where to go and find the question
Yeah. but to give me an answer in my language
and I suppose the the way you can connect that to other services so around like language understanding like the lure system inside Azure or whatever other cloud providers people might be using and the the um uh visual kind of cognitive services the speech cognitive services is not you know it starts to augment it even further beyond that chat capability so when we talking about chat bots I suppose
chat is the kind of medium but you can actually include other things so I suppose there's the technology out there at the minute I could upload a copy of my tax for example to a bot and they could analyze that for me or I could, you know, it could be a point in time where I could start to upload documents and then using visual recognition they could read that.
Oh my goodness, you just unlocked for me. I was using a bot last weekend because I was trying to work out should I change my electricity provider and it said upload three bills and so I just took three pictures of three of my bills, uploaded them to the New South Wales energy commissioner site and it read that and then told me which was the most cost effective electric provider.
Yeah, fantastic.
I I didn't even think about that until you mention it. But Dan, you know, let's let's roll this into the bigger picture because we're talking about chat bots and stuff like that and where they rock up. So, I've used them on websites. Where else would you see them?
See, well, they kind of everywhere really, Ray, aren't they? And when you start thinking about I suppose when when we when we step back and talk about this generally, I suppose it begs the question of what is is AI and where do chat bots sit and are they in other forms and you know I you know you you they're really augmenting our own intelligence at the minute and I think it's quite interesting when we use technology in mobile phones for example so even though that's not directly a chatbot it's it's something that kind of augments what we're doing so I take my kids to cricket or sport every weekend and I'm looking at my phone and I'm following my mapping tool on my phone and that takes me to the ground and I'm kind of cognitive cog atively amputated which is one of my colleagues Travis used to say quite a lot you know it's because I'm not thinking about being in the present sometimes I'm just following the technology being guided by some form of automation so I think when we thinking about chatbots yes we've got that scenario but we've also got the fact that it's augmenting our intelligence generally and it always has so I think it's not something that's kind of coming and something new it's already here a lot of you know you could ask or be asked any question these days I kind of almost killed the pub quiz. You know, I used to love going to pub quizzes or in Australia, trivia nights and and you know, these days people are kind of checking their phones and the toilet, all of this kind of stuff. Not that I cheat in a pub quiz, right? Of course I'm Welsh, but at the end of the day, you know, I think some of the kind of uh the all the sum of all knowledge is kind of at our fingertips. So, so whether it's a chatbot, whether it's search, I think it's about the fact that we've always been augmented and we since mobile phones have come out and we is adding our intelligence to us.
So you might see chat bots turn up on your phone or in Messenger or something like that. It's not just something you'd see on a website, but
correct. But like in in Microsoft Teams or or or something like that, for example.
But but let's go to the bigger picture then cuz
if this technology is going to start becoming more pervasive and and I know you've got, you know, Alexexas in your house and so you talking to a bot then and getting answers. What is the impact then on people in terms of for example the jobs that people are going to be doing in the future or the opposite side of that which is oh my god robots coming to destroy our world. What are the jobs that are going to disappear because of this?
Well, there's a lot of that that manual process and a lot of jobs that kind of rely on that manual process. You know, we've we've had already moved through industrial revolutions and I suppose they call it now the fourth industrial revolution with digitization of a lot of the the technology and you know the robots coming in to do lots of the manual work and there's lots of those jobs that are moved out but lots of the you know the I I see all the time and I'm sure we all do when you sit back and just have a look. You know, I've just gone through the tax year and I'm looking at that and I'm thinking this is this can clearly be like uh kind of done digitally. Um so I think we we may not even see that as a front end in some places in terms of like a chatbot interface. We it'll just get done for us. Why I need to do my
tax every year kind of fascinates me. You know, it's like the the government know everything going in and out of my account. They know my pay is the deductions and things like that. But you know the the the things that are kind of process driven will go first. Um you know and I think accountancy and where those processes are where you seem to be double dipping all the time and then the the knowledge workers things like where you need legal advice you know that's an interesting one where sometimes you you know you can kind of find that information but sometimes you need interpretation of that law so I think that's when it might blur
so you might then see the kind of entry level jobs like the entry- level law jobs start to disappear because they get automated first by bots and conversational agents, people dealing with paperwork, especially people whose job it is is to see a piece of paper, type it into a computer,
you know, that kind of stuff might disappear.
Yeah.
But you still need the high level
lawyers, you see, you'll still need the creativity and all of that stuff that isn't rules driven.
So that's an interesting question for thinking about students of today is how do they get through a period where the low-level job disappear, but they need to get the experience for the high level jobs.
And and we've been talking about this for a while, haven't we? And I it just it's one of those things that kind of I don't know inspire me in one way, but kind of make me feel sad in another cuz we're always saying about the jobs of the future and preparing students for the jobs of the future. But we really are going through a transition. Like the last couple of years seems to have gone on that hockey stick. It's gone exponential. You know, like suddenly out of nowhere, Elon Musk was landing his rockets on a on a ship in the middle of the sea and like just like on on on one day we put it on the news we haven't heard anything about it and suddenly there's this thing landing and you went what happened there and similarly with the Tesla cars and the automated driving it seems to have just
accelerated the path of innovation has moved really quickly so yeah and I you know I really do think that you know I think my own kids what are they going to be doing in the future and what jobs will there be and I know we've talked about it for a while but I really do think we've got a skills gap now
so here's the question is it really the future you know, kind of a bot to to to have a conversation with because it's not because if I I remember a few months ago talking to Optus support and I wasn't sure if I was talking to a person or a bot. I give you the example of my electricity bill. There's more and more scenarios where I'm talking to a machine rather than a person. So, is it is it really the future? Actually, I think it's probably today. So, let me take you to the uh the question that I always find fascinating, which is so could you replace a teacher with a bot? Can AI replace teachers?
Oh, that's a big
I'm I'm gonna sit back and and see you handle that one.
No, no. Uh well, again, you know, teaching in the teaching profession is a lot wider than part of our knowledge. And and the interesting thing from an education point of view is we get measured on the the exam results, whether that's in Australia or whichever country of the world. So, you know, there's there's an element of imparting that knowledge and filling up a vessel rather than lighting a fire. So, uh to take your analogy earlier on, maybe there's element of um augmenting learning in the classroom using a bot. But but teachers are are worth more than just the knowledge they're parking and it's very much about well-being and and the the holistic education of our child. What do you think about this?
It's it's interesting because I think we come from different perspectives because I always say, you know, what's really important when I'm talking to people is I'm not an ex educator.
I'm not an ex teacher. I'm I'm an education technologist. That's five years of my career is education technology. You come from the background of, you know, you're a teacher, you're an ex- inspector, you've got a deep understanding. You know how to pronounce pedagogy correctly. But I look at it and I look at the people that talk about replacing teachers as being this Silicon Valley pipe dream, which is that people don't truly understand what happens in the classroom. And they go, well, the teacher is the variable. We can replace it with technology and therefore everyone will be taught in the same excellent way. it'll be great. And I just don't believe it because the role of the teacher is exactly right. It's building the passion for learning. It's the guiding and the steering, the coaching, the mentoring, the enthusiasm. It's all of those things as well as subject specific knowledge. And I think there's a deeper question about the purpose of exams because if we can get computers to pass exams,
that's absolutely right. Yes.
But but that role of I think that everybody needs to become a lifelong learner especially with the pace of technology change and the pace of change in societ That role is essential for a teacher to get a student connected to learning now and for the future.
But but say if you take from a Microsoft point of view, we we very much autonomous learners. So we
we don't get we get certain training that we have to do. But a lot of the time it's about your career path and your own development and your own interest and developing your own learning and that lifelong learning
mentality that I will go out and I will learn about bots or I will learn about AI. Um and you need to do yourself, especially because we're on the cutting edge of some of that stuff and it's new. So, we need to know about it to understand the implications. But when I think about it from an ex-teer point of view, there was a lot of things that that the kids could get from online learning. And you know, you're you're right about that entire kind of um uh the different facets that teachers bring to the classroom and to the whole student and to groups of students and to society. But I think there is a lot of the stuff that we do do can be automated. And I know it's a bit controversial, but you know as as a teacher a lot you know it was and it was predominantly because of what you said it was to do with the exam results it was to do with our results of driving kids literacy so how much could they read and how much what the reading level was and how I moved from reading level A to B that a lots of that can be automated but actually
the love of reading
can't well I would say can't there you know there is more to it than just those results but it is interesting from an autonomy point of view so quick example of that, you know, I I learned to play the drums, but I learned via YouTube. And then I got a guy in to help me who was like a famous kind of uh drummer and and he said, "No, I'm holding sticks wrong. I've got the and my seats incorrect and I did that wasn't on the YouTube video. All the YouTube videos about holding a Foo Fighters beat." So, um it was it was you know, I was doing it all wrong. So, there's there are elements of online learning that kind of support and I could do a better job to be fair. You know, a lot of the time good teach teachers will use technology to augmented anyway. So if I'm talking about um I remember you might want to close your ears here right this is a bit dodgy but I remember having to show robots uh you know the use of robots with the kids and you know I used to show a video every year and and I used to pick different videos and it was one video that that would make everybody squirm which was one which is like a robot doing a prostate surgery operation and because the robot was so big and the there was a guy laying down on this the kids used to freak out But they could see that this operation would kind of happen and it was like pretty frantic video, but it would it would capture their imaginations and panic on the faces of spec specifically boys. But um uh you know good teachers would always augment the classroom with extra videos and learning to really drive that education because I could I could explain about robots and I could explain about the way robots can change society and all these good things when I was teaching but showing them a video of an actual robot doing an operation and say that's better than a a a surgeon doing operation. It's going to cost less and you know all of the things laser surgery and stuff like that. It would it would really bring that home. So I think there's there's elements of that knowledge that that technology can do better.
So there there's an interesting thing there about logic says one thing but your heart says something different. You know the the logic might say that this robot surgeon is much better at performing operations.
M
but if it's you, would you want that recording on you?
And maybe maybe that's the same with things like um automated exam marking. If you had a if you had a bot that was marking exams, had a conversation with a student and was making an assessment of their capabilities, would you feel comfortable with the idea that um that AI can mark more consistently than a teacher? Yes. would but would you apply it in your own exam? Would you want it to apply in your assessment?
Wow, that's very profound, isn't it? Especially when when you look at some of the topics like say art or English where you you know there is a rubric behind it but people wanted to make it when people are marking those things you know when you do pair back a rubric from you know your marking criteria even when you're marking artwork it isn't a case of looking and going well that's pretty or not from the process. So
really everything when you are talking about marking even when it's a creative piece of writing that people will say, "Well, Ray's interpretation might be different to Dan's." You can't have that. So, even in the most creative things you think can't be marked, it does pair down to a rubric because you have to say
the Ray's got a 45% and Dan's got 32% and you've got to be able to mark that. So, actually there are rules in place and that's all of these things need. There are rules in place for all kinds of rubrics. So, could teachers be could AI take the job as teachers? Some elements Hopefully the ones that are very time consuming and have to be performed
marking
in the evening with a glass of red wine.
Yeah.
Then you can just have the glass of red wine, not the glass of red wine and the laptop.
Okay. So the stories I've heard kind of take us into yes, you can do some automation and and there are some ways it's going to change professions and things like that, but I think we both agree you can't replace a teacher with a robot.
But but you've done some work with University of New South Wales, right? Where where that you've gone into process there.
Yeah, that's interesting because the work that's being done there. So, uh Dr. David Calaman who's in the uh school of engineering at UNSW, he's got so so first of all, school of engineering is bigger than Stanford University on its own. So, this is a big university. But his first year uh mechanical engineering class is 500 students. So, imagine that lecture photo of a whole heap of people sitting in a room watching one lecturer. The challenge is how feel help people to feel connected and so they have done a lot of work around how do they use technology to support those students to feel connected to the learning process and to be able to personalize things down. So what they've done there is that they built from a bot perspective lots of other aspects to the project but from a bot perspective they've built something called questionbot and questionbot sits in teams and it allows a student to go in and ask a question and what questionbot does is try to help find an answer. So, it might go to a teaching adviser. It might already know the answer or it might go and look for references that it can bring back to the student to say, "Here's where you might find information." So, an example could be you ask about a concept, what it can do is go and search the previous lecture recordings to see if that concept is mentioned and then say, "Well, it was mentioned in lecture 3 at 48 minutes and here's the link to go and watch that." Or
fantastic,
it can look at the question and provide resources to help revision for a student. So, and let me give you a really specific example. So, students operate differently to the way that we operate. The first thing they often do is take a f photo of something. So, what they found was students were taking a picture of the assessment question and saying, "Can somebody help me with this?" Now, students have always done that. They've always done it in Facebook or
text message or whatever it is. Now, they do it in Teams. And what they've done is put a QR code on the assessment paper so that the bot can work out what the photograph is of. It's like that's question three. And so the bot can then say, "Well, this is where you go to to um reflect on
the skills you need to."
So they put pictures into the into the bot as well. So they put that in.
Yeah. And and so it leads to a very natural experience which is doing what students do. I'll take a picture of something and ask for help. It also integrates the teaching adv uh assistants and other students. So for example, if you ask a question of the bot, another student comes in with an answer, a teaching assistant might look at that and go, "Well, actually, there's a better answer." Provide the better answer and tell the bot that's a better answer. So then for the future, the bot remembers that if this question comes up, here's the here's the the best answer to that.
And that's supporting learning then rather than telling the answers and just saying, "Well, this is what you're doing." Right? Because that's is very there's a bit of a fine line, isn't there? You putting your AGL, you know, electricity things in Yes, it's going to cost you 35 bucks still like how do I do this and then the bot going this is the answer.
Yeah, it's it's very much about pointing you towards the resources to help you
to find the answer or to work through a problem rather than here's the answer. A lot of the bot scenarios we see in the commercial world are here's the answer to the question which is exactly what you want when you say to your Alexa what's the weather going to be like today? You don't want it to say hey you can go and look at these things.
You want it to go it's going to be six degrees. Take a coat. And so, but in student scenarios, we're encouraging inquiry and learning. And so, actually, what you're going to do is go
point them towards resources that going to help them to learn as opposed to point them towards the answer.
And but the interesting thing what we've seen with the UNSW case study is what it has led to is a more engaged community of learning because students now know that they can go and ask questions even though they're one of five. 00 in a in a class
and get a personalized answer that's going to help them take the next step in their learning journey.
And it doesn't get more complicated in an engineering class really. You know, I'm thinking of all of the different subjects out there, you know, the everybody kind of it's a always people get when when I've been working with teachers in lots of different capacities, you know, a maths teacher will always come back and go, "Well, it doesn't work for maths. You've got to have a whiteboard. You've got to have, you know, you've got to have a piece of paper. You can't digitize maths. You can't digitize elements of certain subjects. Well, it's actually, you know, it doesn't get more technically difficult than engineering.
And and when you see the teaching practice that happens in that lecture hall with ink, digital ink with surface devices where that's exactly what they've done is digitize a piece of paper.
But then you point that student back to the moment in the lecture where the lecturer is drawing out the mathematical concept on a whiteboard. It's a digital whiteboard, but you can see and engage and that's I think why they've led to this scenario where students now feel much more engaged with the teaching practice
and with their community of students. I think I think the results were all of the students or in that first year course were responding and saying that they felt a more integrated part of the learning community and what was especially true was for international students because sometimes if you're an international student you just hear a lecture you maybe can't comprehend it at that time being able to ask a bot a question and then go back and revise a lecture or a piece of material. For them, it not only connects them to material, but it gives them a multimedia way of seeing things. It's not just listen and watch, it's listen and watch and read, potentially translate, potentially see information in a number of different ways to help them to embed that knowledge.
Yeah.
So, that's a it's a good example of using AI to be able to help other people. There's a bunch of case studies around it. There's a really great short video of David Calaman talking at our Inspire conference in front of 25,000 people, which is a big lecture. Um, and just explaining in 10 minutes what they did. So, yeah, if you go if you go and search for UNSW Inspire or Microsoft Inspire, I'm sure you'll find the video. It's on YouTube. It's It is phenomenal and we'll put it in the show notes.
Um, so when you're talking about, you know, you you mentioned already being, you know, that's a good use of AI. I suppose that'll be good for another podcast. Talk about AI for good.
Well, we've done the AI for good schools competition. We've got the global grants program. So why don't we talk about how AI is going to help us to solve some of the biggest problems in the world next time around?
Fantastic. See you then.
Thanks Dan.