Oct 23, 2019

Dan and Ray talk about Artifical Intelligence for Good - things like using AI to count fish stocks in Darwin without having to risk your life from the crocs; and the use of AI to detect and prevent poaching in the national parks of southern Africa.

Now AI can recognise text, speech and images better than humans, there are lots of scenarios where it can help us to achieve more. 

Ray talked about Microsoft's AI for Good competition for schools, and the ideas that the teams pitched at the national finals in Sydney. More information on the finalists is here.

The 2019 competition is over, but it's not too early to register for the 2020 competition here: https://aiforgood.com.au/

Slightly off topic, Dan mentions https://www.how-old.net/ which uses AI to tell you how old you look; Ray talks about apps to help detect autism

Ray puts Dan on the spot about having a house full of smart speakers and how he uses them, which Dan uses as an opportunity to try and re-direct listeners' Alexa devices (warning: if you're listening to this through your Alexa, you could end up listening to Baby Shark halfway through!)

TRANSCRIPT FOR The AI in Education Podcast

Series: 1

Episode: 5

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 explores numerous inspiring examples of AI for good, highlighting how artificial intelligence can address significant societal and environmental challenges. Key applications discussed include using AI-driven "fishial recognition" to monitor fish stocks in challenging environments and employing sound and vision technology for anti-poaching efforts in Southern Africa. The hosts emphasise the growing capability of computers to excel in recognising images, text, and speech faster and better than humans, opening up a plethora of practical use cases, particularly in humanitarian and sustainable development areas. Importantly, the discussion showcases the remarkable innovations emerging from student-led competitions, such as apps to reduce food waste, diagnose sight disorders, and manage diabetes, demonstrating the boundless potential when people are supported in applying AI to real-world problems.

Welcome to the AI and education podcast with me, Dan Bowen,
and me, Ray Fleming.
And today we're going to look at some really good uses of AI. AI for good, Ray.
Oh, awesome. Does that mean in a couple of weeks we get to record the AI for evil?
No.
Oh, I think we should then.
Oh, there might be. Maybe we could cover that today as well then. Squeeze it in.
Let's do a little bit of it. So, okay. Thinking about AI for good, Dan.
Yeah.
What's your best case study? What What have you read? What have you seen?
There's some great ones, isn't it? I The ones I love. Oh. Oh, actually the one up in Darwin recently was really good where we actually got AI that can actually uh check fish stocks inside the kind of harbor in Darwin. And obviously for people who don't know Australia that well, there's lots of crocodiles in the sea in Darwin. So, you don't want to go and count fish yourself or um and and you know there's a lot of uh you know things around fish stocks and uh and you know a security issue in there with jumping in and counting them. So so actually uses AI to do that facial recognition to kind of check the faces on the fish
fishial recognition Dan
fish shield recognition. So what is it doing? It's got a camera looking underwater
looking looks underwater and kind of identifies the different types of fish that are there and kind of uh gives them some data analytics to be able to kind make decisions on fish dogs.
So, in the same way that people use facial recognition to like look in a crowd and count the number of people there, they've trended to look at fish.
Yes.
And recognize different breeds.
Yeah. And how happy they are and how sad they are.
How about you, Ray? Have you got
uh I think probably the one that's closest to my heart, which links to, you know, when I was lucky enough to take my long service leave in Africa, is they're using AI technology to help with anti- poaching in Southern Africa. So, we have this AI for good grants program which is a global program and in southern Africa they're using it to help them to detect poaching. So one of the ways they're using it is listening for sounds across the national parks and so they can listen for things that are signs that there are poaches in the area. So for example to be able to listen for animal alarm signals to be able to listen for gunshots
and be able to put more proactive anti- poacher into place. They're also using it with uh cameras, so to be able to set up hidden cameras and spot if people are coming into the park on poaching routes so that then they can send out the rangers to intercept the poachers before they've actually attacked anything.
So, you could use drones as well. I suppose
there's all kinds of things that you could do with that. And it's just sitting there and having a the understanding of the problem and b the application of AI is well, how could we solve it? Just like the the fish recognition in the harbor, it's like the the problem we're trying to solve is how do we count the fish successfully? And then the application of AI.
And to segue those two together, there was another case study on snow leopards, wasn't there, where they would where where it took them a long time to count the snow leopard population by analyzing hours and hours of video footage from videos in the field and actually connecting the dots together and using the analytics behind that. So there's a lot going on with with AI for good then. Right.
Well, because we've now got into this scenario where computers can recognize images better than humans can and certainly faster than humans can. They can recognize text better than humans can. They can recognize speech better than humans can. And so you've now got into this kind of scenario where all of these different use cases suddenly become possible. So for example, I I remember at the Edgitech show much earlier in the year, I was showing an example of looking at images and in preparation for it, I picked up some of my pictures from last year and I picked up a picture that we' taken on safari and some elephants. And I'd forgotten that the picture was from 5 years ago in Sri Lanka. And so it looked at the image and it said this is a group of Indian elephants. And I thought, "No, hold on. That's from Africa." It's like, "Oh, no, they're right. I'd miss muddled up all my pictures." But it understood the difference between an Indian and an African elephant and was able to label the picture for me. And so that's the bit where it's like this stuff is better than a human cuz it took me longer to work it out than it did for the tech.
And it's knowing knowing the tech. And I think one of the things that I've been doing the last couple years is is is showing kids and teachers and you know CIOS and business decision makers just some of the examples of you know we call it cognitive services don't we? So so from our point of view we we're looking at the services that Microsoft use in inside our cloud technologies to recognize images and things and they show lots of examples you know it goes right back to the first one I remember used to demo it which is how old.net then that that expanded across to kind of um what emotions you know you're happy or sad and what age and what gender and really interesting things and then then you know emotional intelligence like I said not intelligence but emotional kind of features that that you might be doing and crowd insights and then and then fun things like what dog am I and what celebrity do I look like and and and all of those kind of things and I think the interesting thing for me is that when we show people the technology and say well this does this then it's the application of it so it's then saying well what can you do from an education point of view yes this is This is a um an image of this. This tells me how factually correct I am looking like another person like a celebrity or whatever. But how can you kind of augment that into an education setting for example for plagiarism in an exam or or to compare you know for the department of immigration people coming in and out of the country and things like that? So once you show people the technology then they come up with really good ideas then especially kids kids come up with fantastic ideas.
Yeah. And at the AI for good schools competition earlier in the year We actually set a challenge for years 7 to 9 and 10 to 12 to look at AI services, think about a society's problem and and often when we're talking about society's problems, we're talking about humanitarian problems. We're talking about things that link to the sustainable development goals from the UN and said to them, okay, well, look at look at this scenario.
Yeah.
Think about how you could apply AI to help people to navigate the problem. And it was amazing seeing the results that they came up with. You know, we had groups of kids that came up with an application to to stop food waste. So, what's in your fridge? What's going to go out of date? Suggest a recipe for it. It's it's a really chatbot in the last session, but it's a really simple example, but it's, you know, it's it's a deep society problem because we waste a third of the food that we produce. So, if you can reduce that, and that's that's a big goal. They looked at other things. So, they um they looked at being able to use AR and an AR headset and a neural network
AR
uh augmented reality the big horrible goggles thing you put on the front of your face but using that to be able to diagnose uh sight disorders in young students because it's really difficult for medical professionals to do
but they were able to do that
using the vision services
and and so this is like young kids coming up with those
just just on just on on that one I remember doing a session with some uh kids at at Microsoft in Sydney when some students came here from his local school and I gave them showed him some cognitive services and talked about that and one one boy I remember he he said to me now this isn't scientifically accurate I will disclaimer for our podcast but but he actually said to me his brother and his sister were quite severely autistic and um I showed him the emotional APIs the emotional services that we have to to di you know to to allow you to see if you're happy or sad and he said that his observations of his and this is a kid a primary school kid was telling me this. He said that cuz I asked him to come up with some applications and some a million-dollar ideas and really got him excited and he said that um based on his experience with his his uh brother and sister that if you had the emotional APIs and the camera on on a phone application and you ask questions around love or films and things and music, he said they get a a um almost like a bit of a vacant look. on on on their faces from certain emotional questions. So he said he suggested he could create an app to kind of almost like detect autism which is kind of pretty phenomenal. I it's obviously not scientifically correct but there are must be elements in there that might be kind of
that's amazing that they would see that potential because that has actually been turned into a commercial application.
Oh wow.
In the medical uh sphere. So it's gone through testing and basically what you're able to do is early detection of autism amongst children and again difficult for medical professionals to do it to the quality and volume that you're able to achieve through AI and so um I think the scenario is that they show very young children funny videos you know the the things we used to laugh at as kids you know somebody walks into a rake or whatever and then they're looking for emotion you know
um so they look for an emotional reaction and so if there's an absence of an emotional reaction then it's an early
that's interesting because he came up with that in in you know in in a session. I just didn't make myself.
Well, let me talk about some of the things I saw at the for good finals because I wonder which ones of these we're going to see in an app soon. So, there were a bunch that were using drones. So, there was one group of students that were keen to see the bilby and they were using drones and
a bilby is a
Australian animal,
right? Okay. Like, okay.
Um, and so they were using the services of a drone and vision recognition to be able to look at populations of billbees. Another group were doing exactly the same thing for agriculture. So their idea was you fly a drone over a field and look for areas where for example the crop isn't growing as strongly and so you know you need to water that area and so you then apply irrigation in the area that's needed rather than across the whole field
saving water at the same time. Fantastic.
You were looking for infestations. So they were looking for certain types of leaf discoloration and so they know that they would only need to apply pesticides and insecticides in certain parts of a field. Whereas today you fly over the whole field and spread inspe insecticides everywhere.
And drone technology, I know you've used drones quite a lot because I was following your Facebook with your fantastic African safari, but that drone technology is pretty good now, isn't it?
Yeah, it really is.
And so drones have become consumer grade from a price point of view
and they have very good cameras on them. So if you can link that camera to some AI services that then is able to analyze the image, then you got great potential. M
other ideas we saw there was a group that was using AI to manage diabetes for kids. There are medical devices that are able to take readings from kids all day long but being able to connect that to live AI services so that then it was able to predict what's going to happen next and do interventions or recommend interventions.
Wow.
There was a team that created a watch concept. So very similar to smart watches today but it was using the vision services. It was using the sound services in order to be able to do things like use the microphone to spot when a student is in a very noisy environment. And they were specifically targeting on children with autism and saying that, you know, if a child is in too noisy an environment for too much of the day, then their performance is going to drop off in terms of their ability to cope socially.
Yeah.
And so they said, well, we can use these services to listen to the environment and provide nudges for both the student and advice for the teacher and parent. And and it's presumably, you know, you can use a lot of these things in in technology you've already got. So like an Apple Watch has got a audio microphone on it already, hasn't it? I believe and you got your sensors on it, your heart rates, you can actually bring all of those together in application processing in the cloud. One of the things that when I've always had that debate around ethics and and the use of, you know, when you get into the medical element to it and people kind of say, well, I wouldn't share that data and I would share that data. You know, I I go back to my my own daughter who when she was born She was born with a defect in her heart to do with an electrical pathway. And she used to have heart attacks like every day, like several of them. And it was really kind of disturbing. And it was before Fitbits. And I always think, you know, if I had a Fitbit on her, cuz you you you can sense heart attacks and things like they've done with the Apple Watch now where you can sense heart attacks before they're going to occur in a lot of cases and definitely after the case as well because it appears in your ECG. So I always think about that because we do worry about, you know, we had a big debate in Australia around health records and things like that and some people sit on one side, some people sit on the other side and and you know the stuff around Alexa and home and we can touch on that a bit later but you know it is an interesting when we talk about health because you know the the the use of AI and being able to predict if my daughter was would have had a heart attack would have been phenomenal. What other examples did they come up with?
Well, so there was another group called Foodie Impacts and uh they were from WA and they were using the vision service
to be able to look at food items you bought. So they were suggesting that could provide a provenence backwards so you pick up something on the shelf and it can say well this is where it came from so that you as a consumer have more insight into the way that your food is produced
like blockchain almost for for your your products.
Exactly. Right. But but straight from an image service which is well just pick up the packet and we'll recognize what the packaging is of and tell you more about that product. That's
and help you guide your decisions about food you ate. The other things were there an app for mental health on your phone to help use AI to help understand patterns of emotions and triggers for certain emotional behaviors. I think we see that often in a discussion around how much you're using social media apps and things like that, but they've extended that way out into other scenarios about being in noisy environments, being in super bright environments, getting quiet time during the course of the day, all kinds of different scenarios.
That well-being element's fascinating, isn't it?
Well, what what fascinated me was these are year Seven.
Oh yeah. Think about it like that. 10 to 12 kids. You know, this isn't university researchers that are seven years into a PhD solving this stuff. This is, you know, kids that have had a few weeks and some training to understand what things can do for them and applying it. Often I found in most of the presentations was a scenario that was related to them. You know, they had a brother or sister with a problem, they had a parent with a problem, they had a a village with a problem. And so they were applying it to scenarios for them
and that's interesting as well when when you know I did a lot of work on trying to get girls in computer science and STEM and and things like that and that was one of the main drivers around project based work was very much around the intrinsic value of what what was actually in it for them you know what's the actual output is it something to do with a greater good was it was it a good representation between girls and boys in the
yeah it was there were 584 teams took part and uh you know we only saw in the national finals we only saw just a few of them. But it was a really good representation. I'm trying to remember the two teams that won. One was a team of girls for something they did around an AI enhanced midwifery support in Africa. The other was mixed team. So it was really, you know, it was it was really fascinating to see there was a real blend of different age groups, different types of schools across the different states, you know, different gender mix. They they were genuinely engaged in problem solving. It wasn't geeky stuff.
Yeah,
it was a genuine society problem.
And I suppose, you know, there's lots of grants and things that that Microsoft do for example on the air for good um projects as well. So some of these might be, you know, some of these projects which some of the and the ideas these kids come up with could be pushed into reality really because there's plenty of capability around that.
Yeah, we've got an AI for grants program for researchers which is mainly aimed at universities and research institutes and they're doing things as diverse as scanning the Great Barrier So that we're supporting drone flyovers of the Great Barrier Reef and then looking at the footage taken in order to be able to look at particular problems in the reef, look at uh um the proportion of reef that's been damaged or dying. Um we've got the same happening over in WA. There's lots of different projects like that. I don't think that any schools have applied for it, but having heard some of the ideas from these to be able to turn them into reality would be amazing as a global program.
Yeah. And it's interesting from just stepping back from it you know there's all there's that element of creativity that's needed for these things and to a certain extent it's not they don't need to be able to code a lot of it but they do need an appreciation what the technology does when I used to teach I remember when I was doing mobile phone application you know um uh we used to do mobile phone app development that that's really you know in in it's not really fad anymore is it it's kind of a bit of an old school thing to do but when you're teaching kids to program and make make software for mobile phones what I found was the more they knew about the mobile phone, the better. So, I used to teach this module for years. And I was thinking it was the the output wasn't great. It was they were trying to make games essentially like Flappy Bird and all of this kind of stuff that was already in existence. They were struggling to conceptualize it. As soon as I spent the first three lessons, like literally 3 hours worth of time talking about what an accelerometer does, what the GPS does, how that works, what does Bluetooth do, what's the range of Bluetooth. Once they know all of the things in the kit bag, the ph phenomenal applications that they came up with after that was great. So I suppose it's about us being able to support kids and teachers and schools being able to support kids with that technology and say well this is what what's available now the ideas are yours right
you're right you have to understand how the technology works because then it unlocked that potential of being able to do other things you remember Seymour College
in South Australia.
Oh yeah they did a fantastic project there with us around it was around the Azure machine learning which sits inside my Microsoft Excel and they did a a project to predict breast cancer. So they're using data to kind of allow you to kind of select your gender, your your age, if you smoke, certain things like that, and they actually predict, you know, something real real time. There's a girls school in in South Australia that came year 10 kids, I think.
Yeah, that's right. Yeah. Who knows?
Yeah. I wonder what they're going to do when they turn up at university. They're probably at university now and they've been used to doing machine learning predictable.
Yeah. Yeah. Yeah. Exactly.
Fascinating.
Absolutely. So So Dan, I'm always fascinated. We've had this conversation in the office a few times. So let let's do it when you're staring at a microphone because when you understand the potential of technology to be able to see things, understand things, listen to things.
You've got Alexis at home.
Yeah. Yeah.
And so my perspective is so you've got microphones plugged in listening to you all the time, but you love it. So tell me about that.
Um well I I suppose thanks for putting me on the spot here, but yeah, the um you know For me, I suppose when I'm using home automation tools, you know, I suppose it's it's it's kind of a perversive way into your your home into your everyday environment. It started off with me for getting the kids ready for school, setting alarms and things like that. And then what I do now, I use it to police Xbox. So when the boys I I got two boys and a girl, the boys kind of kind of argue over time on on Xbox. Yeah.
So they go up to Alexa and they, you know, they haven't even got to go up to Alexa. They just say from another room, they say, "Alexa, How long have I got left on Xbox? You know, how long's the timer left? And then it'll um when I've just said the word Alexa, now everybody listening to this at home, Alexa, stop um stop the podcast. That's always quite a good little thing to do. Alexa, turn the lights off just in case somebody's got home automation listening on podcast.
You just lost us 10,000 subscribers.
I know. Uh but doesn't matter. Um but at the end of the day, I think it's been really useful for me, you know, when I'm cooking, reading recipes. I mainly use it for to the news. So, you know, I ask it for my daily briefing every day. I ask it what I've got on and it tells through into my uplook calendar and tells me on my day is going to is going to kind of start and I know we've had tools and services in Cortana to do that before and I know Siri do those things. But I've got that connected at home and being able to kind of have that listening in play like this morning I was playing music in the background. My daughter just completely just her standard line now is like Alexa play Baby Shark on Spotify. and it's like and everybody else we've got Alexa in the background now is going to go crazy but she she you know she'll kind of listen to music she'll talk to as if it's another person in her house
but again you know there are things with all AI where you know and I know there's been a few things recently in the news around the fact that through dialects and through you know you mentioned earlier on in in a in a podcast so aquatic signs through an aquatic center and things like that there's certain nuances in different languages and and you do need that human ele And I think there's there's that element at the minute that some of those things are being trolled through by developers who are listening in on audio clips to check the accuracy and to improve the AI in the back end. And you're right, you know, you I there's been spooky things that's happened to me in the past where I've talked about things that I wouldn't normally talk about on things like Facebook Messenger or in rooms in conversations and without doing any kind of conspiracy theories or anything, but you suddenly see an advert for it coming through. And I know I've been talking that it was a particular time where I was talking about a bizarre European country that I'd never even heard of and suddenly I get like a flight deal for it and I'm like that is way too spooky for for for my liking. So, you know, it is it is an interesting one where you think about AI for good but also AI for evil. Yeah.
What what's your perspective on it then?
Well, you know, I don't have any of those devices in my house, but I don't know if it comes from the could somebody be listening all the time or if it comes from I think we're in the beginning stages of home automation. And so the knock on effect of having all these devices connected together just might mean I spend even more time troubleshooting technology issues. So you know for me it's like well there might be a time but I when I think about the scenarios I'm maybe just a bit just a bit I'm much more old-fashioned than you maybe you know I was born in a time when I was the TV remote control when my parents would go can you go and change it to be
I remember that
but it didn't take but the good news was there were only two TV channels, so I wasn't able to change it too much. And so I look at some of those scenarios like using voice control for lighting and stuff like that and go, "No, you just go and turn the switch on and off."
And so, you know, I kind of am in a mode at the moment whereas, yeah, at some point in the future, but, you know, I I get these occasional things where I bump up against technology trying to be too smart. Like, you know, I'm an addict to Fitbit.
You know, I've been wearing a Fitbit for six or seven years. You know, fortunately, I've got out of the But now I've taken the dog for an unnecessary walk at the end of the day to to get my 10,000 steps. Some days it's okay to not hit it. But I've got a Fitbit connected scale and you'll remember last year I had a terrible knee injury and I couldn't run for a few months and I'm still recovering. So as a result I kept eating like I'm running but I've therefore put on weight and I got my Fitbit scales recently for the first time in a long time. And
Fitbits they connect this scales connected to your Fitbit app.
They connect on Wi-Fi through to the Fitbit website. And then I can see my results on my phone. And and great, it charts for me my decline or or or gaining of weight. But the last time I got on them, I hadn't maybe got on them for 4 months. And it refuses to recognize me because I've put on a bit of weight. Brilliant, isn't it? And so now I've got to reset my Fitbit scales in order that it knows who I am.
And it's like that that is it's like that's that's AI for evil. That is technology that is, you know, punishing me.
But but it but it like when we looked at all of those, I think is the thing is there's all where technological change I think we were listening to something in our kickoff by uh Dr. Jordan Yen and he's done a lot with disabilities and students and kids with cer Pauly and a lot of kind of really severe ailments and helping them to kind of drive cars and pretty amazing stuff but I suppose he he was saying that with every innovation comes that downside you know and you know we've got the technology there that could quite easily you're talking about the drones and the crops it's quite easily you could quite easily coded drone to go across into somebody's garden and explode or take somebody out or whatever it may be based on the color of their skin, their gender.
Yeah, cheerful
English rugby supporters.
So, let's think about it then because we've got this potential for AI to be able to help us, you know, the AI for good idea, the to be able to make things more accessible, to create a more inclusive society. I mean, there was a a group in the AI for good schools final, the one of the Tasmania teams that c created an application that was designed for people with limited mobility or limited visibility to be able to use public transport because you and I don't think about it but public transport can be difficult because you've got a timetable that might be published electronically these days but how does a blind person use it? How do they know how to get somewhere? And so they created an app that integrated bus timets, live traffic information, taxi information so that they could help people with inclusivity need to be able to use public transport and And it was great because they went to their local bus company and said, "We're thinking of doing this." And the bus company was like, "Yeah, we'd love to kind of help you as you as you start to prototype." So, I think there are all kinds of ways that we could be using technology to promote inclusivity rather than to kind of create the halves and have nots of society.
I mean, maybe we should spend the next podcast looking at inclusivity then in AI.
Okay. Well, in that case, we should see if we can get Troy. So, Troy is our accessibility lead in the educate team. Um, now he sees it from the the needs of individuals. So maybe we can have that conversation and we can have a conversation about how AI is helping with that inclusivity bit next time.
That would be good.
Okay, let's do it then. Thanks.