Dec 4, 2019

After 10 episodes, we head back towards the beginning, with a discussion about the origins of Artificial Intelligence (as Dan gets a chance to use his teacher voice), and then go a little deeper into what one of the common terms - Machine Learning - means in plain language. Ray gets to confess to having produced 27,000 brochures with the wrong spelling of 'literacy', before discussing a good approach to your own learning on AI, and how he uses a combination of courses and experiments to further his knowledge.

TRANSCRIPT FOR The AI in Education Podcast
Series: 1
Episode: 11

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, featuring a discussion between hosts Dan and Ray, aims to provide a refresher and deep dive into the origins and core concepts of Artificial Intelligence (AI) for their listeners. The hosts move beyond recent guest interviews to explore the history of AI, mentioning early influences like Alan Turing and the coining of the term machine learning by Arthur Samuel. Key technical terms are defined, including algorithms (a sequence of instructions) and Big Data, which is characterised by the three Vs: volume, variety, and velocity. The conversation ultimately emphasises that the practical application of AI, moving from simply generating reports to driving action and better outcomes for individuals, is what distinguishes it from traditional business intelligence.

Well, Dan, how are you?
I'm great, thanks. A AI. I I nearly called you AI.
I've been called many things. When I order coffee, I'm often put down as Brian instead of Ray, but I have never been called AI before.
Well, welcome to the AI and educ podcast today, Ray.
Oh, thank you, Dan.
We don't even need introduce ourselves this time.
Now, last week,
yes.
Well, in fact, the last couple of episodes have been really interesting because it's not been just you and me. We had that fantastic interview with David Kellerman talking about what he's doing around personalizing learning for his students in engineering at UNSW. And then Sulab came along and Zap was talking about the development of the technology behind it. So, that was really nice. But I felt as we got to the end of the last podcast, we probably ought to rewind a little bit because we talked about the different types of AI in episode one, episode two, but probably a good idea to do a bit of a refresher, but maybe go back even further than that and kind of look at the origins of artificial intelligence.
That's a good idea. Let's do that, shall we?
Okay. Well, Dan, you're the ex-teer, so did you used to teach computer science?
I did. Yeah.
Oh, excellent.
I loved it. It was brilliant.
Over to you, Dan. I'm just going to relax now and and listen to the history of AI.
Great. Well, you know, when we start thinking about it and when I think about the history of AI, we think about where robots are in the future and science fiction and lots of the things you know even I suppose for me my passion in AI came from when I was watching things like the Wizard of Oz we've always wanted robots and things to come alive and we've always wanted that person or that thing to be talking back to us that we can create you know when you go back to Frankenstein and Mary Shelly is fantastic and I suppose AI has been kind of in inbred in soiet the society's psyche for quite a while through movies and through popular culture When you actually do look back on it though, it's been quite a new science as I suppose. Um, when we start unpicking that, Arthur Samuel coined the term machine learning when he was working for IBM and he was in gaming and AI, I believe in in these late 50s early 60s and he actually coined that phrase machine learning and that was kind of the birth of kind of AI and that kind of element. But before that, obviously through um the second world war Alan Turing was doing a lot of stuff with algorithms which you'll touch on a little bit later around what algorithms are cuz there's lots of technology around AI, I suppose, and and computer science where they're quite specific and they sound quite posh, but they're actually not really that complicated, but we do need to understand what they are.
But there's a difference between maths and statistics.
Yeah.
And AI, isn't there? Because sometimes they get confused cuz Touring's work was in the kind of early days of computing, wasn't it?
Of course. Yeah. And and that's it. And we we started to go from Turan's work into there there was a kind of big explosion of AI. There was a thing called the Dartmouth Summer Research Project on on artificial intelligence in 1956 in the US.
Were you there, Dan?
Yeah. Yeah, I was. No, but it's got a famous element because because essentially when you look at the history there, you had this all these mathematicians got together and they said, "Hey, everything can be done by maths. All human um thought and kind of movement can be used. You know, you can you can pair it back to perfect maths and if you can if you get an input and an output, which you can in maths, you can do that." So they they they were under the kind of thought process at that point and the faster the machines came the faster and more they could compute then the more likely they'd get to get human parity in that element but but then after after kind of the that project got together 8week project that they did and there was a bit of boom in AI suddenly went all quiet and they call it like the AI winter and then AI kind of went into the underground and then people started to emerge through that by doing what's called like expert systems and I remember doing this when I was teaching when we used to create like expert systems to to guess the type of disease or guess beer, you know, is the beer brown, is the beer yellow, is it white, what what is the liquid in front of you?
The computer version of 20 questions.
Ah, absolutely 100%. And they're expert systems and that's basically a rulebased system that started to become quite popular. And then in the 80s we
and again just a question that's not really artificial intelligence, is it?
No, not as we see it now.
Not as we see it now. Right.
Yeah. Exactly. But it was the kind of basis of the kind of rulesbased arch ures of things we use. So data scientists do have a good background in some of those algorithms. But but then what happened was in in the 80s we had IBM's deep blue if you remember that and that was when they really brought a lot of computing power towards those expert systems and really start to bring it together and then eventually it led to Deep Blue beating Gary Kasparov in in the late '9s at chess which was like a big kind of change and and there was a lot of things that kind of appeared around their buzzwords around neural networks and things but they were still linear statistics and they weren't really machine learning. But that was about something that we thought of as an innate human skill with a computer having the ability to outf fox a human by outthinking them.
Yes. Exactly.
So I know when when I've
I used to play chess when I was younger, you used to think about how many moves ahead could you be and if you could be thinking more moves ahead than the person you were playing, you had more chance of winning. I realize now I've played chess so infrequently that if I can think one move ahead, that's pretty good going. But I guess that's the bit what were doing was putting the logic into the computer so that it could work out all the moves ahead and I guess the more compute capacity that you've got the more steps ahead it can think of.
Yeah. Yeah. Yeah. Correct. And and I think that's where where it's kind of started to explode now because it moved from there you know when people were doing almost like pet projects after that AI winter and the boom started to come back in where people started to think about it more computer processing power and memory were kind of moving on really really quick and speeding up and therefore lots more processing could get done and now we're in a phase where almost almost as if by magic. I think I mentioned in one of the other episodes that suddenly Elon Musk is using using a lot of that technology to suddenly get a self-driving car going. There was a competition one called the DARPA Grand Challenge uh in the US between Cariegi Melon Uni and Stanford which is about the self-driving car which um Stanford won over Cariegi Melon in mid 2000s something around there but Google's now charged in ahead with some of the autonomous automobile industry stuff with Elon Musk and some of the other players coming in there like Microsoft and and bringing things together with much more complex and much more kind of integrated uses of some of these technologies which we think
presumably oh sorry and presumably better training of the AI as well because I know that a lot of AI systems have to be trained on the past in order to be able to make predictions for the future and the big difference between now and maybe 30 years ago is the huge volume of data we've got available just just a question have you trained a self driving car this week cuz I have.
Yeah, I think I have as well actually. Yeah,
because every time I go to a website and I have to prove I'm a human, that capture box that comes up which used to be house numbers and before that was objects. Have you noticed that now it's can you spot a bus? Can you spot traffic light? Can you spot another car?
And what you're actually doing is training an AI system because you're giving it data as a human. We can spot cars better than a untrained system. So they're actually using us to trying this technology that's going to replace it.
That's brilliant. I when when you mentioned that I was thinking my car myself because I there are elements when you're driving these days in your particular cars where you specify you know the distance you are you train not necessarily train the algorithm inside I don't train inside my car but I'm sure if you got an expensive car it probably does learn a little bit
but it's interesting if you go back to that capture thing where you know prove you're a human maybe six or seven years ago everything you were asked to enter was give me the number that's on this picture and the number was actually the house numbers off Google maps and the other people that were doing those road mapping because they couldn't read it terribly well so they used humans. So for for years you and I have been AI trainers and I'm pretty sure all of our listeners will be as well.
That's brilliant. So we
we're smarter than we think
we are. Well so in that case for for today's the point of today's podcast is to pick out some of those key terminology that we've got around this. So what's our first question?
Well I think we've already covered AI and clear we're talking about artificial intelligence.
Okay.
So then probably the next thing is the concept that I hear more often is ML. That's the other acronym which is machine learning. What does that mean
at this basic level? I suppose machine learning is allowing computer systems to actually learn from our inputs and from the data it's collecting itself. So forming almost like a closed loop. So when you think about the definition of it, it's it's a kind of scientific study of algorithms and models that uh computer systems can generate. to perform a specific task without ex explicit instructions from the user. So it does it on its own.
Right? So that would be something like my phone deciding that I'm going to the office every day. Now at the moment it's my phone that draws a map of where I should go, but I guess in the future it will be my car that I'll get into my car in the morning and it'll just come to the office without me telling it.
Yeah. And what you're doing there, you're setting up what we call training data for that system. So training data is a big essence of a machine learning system. And you've done some of that stuff in Excel, haven't you? With AutoML.
Yeah, I've been playing around with it because I'm fascinated to see what you can do with it, but also how much easier it's becoming because not that long ago it was the domain of really really powerful data scientists whereas now it becomes a lot easier to to use it. It reminds me of the the days of desktop publishing where before to lay out a document was something very very specialist and then laser printers arrived and then suddenly everybody could do it. We're going through the same world at the moment with artificial int igence and certainly with machine learning. I actually did a machine learning course 18 months ago where I learned the very technical way to build a machine learning model. So the one I built was forecasting which video you'd like next. The kind of Netflix picker thing, but it was oh 7 week course and it involved a lot of heavy coding and then
that was one of the books that I started and then dropped out of after
statistically you're likely to drop out of a mukdan. I'm I'm the statistical freak on this one then. Recently we've started to see the emergence of auto machine learning. So autoML that does the calculation for you and and I had a go with a data set that is actually it's very relatable and much easier to understand. So do you remember the Titanic Dan?
Yes. I wasn't on it.
I didn't think you were that old. But what was really interesting is the Titanic is a very classic data set that's used by people to understand how machine learning works because a certain number of people survived the Titanic and a certain number of people went down with the Titanic. And it's used as an exercise to say is it possible to be able to predict your chances of survival based on what actually happens. So what they do is they use the data from the Titanic as a training set to be able to build that model. So let's take one of the factors. What's very clear is the lower the class you were on the Titanic. So first class passengers had a higher survival rate than second class and then third class.
Yeah. And
you would have been okay then.
I doubt it, Dan. But it means that that just take that one factor that you are three or four times more likely to survive if you were a first class passenger. But then you add on top of that what else would be a factor in helping you survive the Titanic? Dan,
uh
what is it they always say when a ship's going down? And then
women and children first.
Women and children first. So if you're a woman, you're more likely to survive. But what we probably couldn't work out easily is, so if you're a woman in first class or maybe a man in first class compared to a woman in third class, what are your chances? of survival. So machine learning projects are about taking that data, feeding it into a computer and the machine learning system working out what are the most important characteristics and how do you weight the importance between those characteristics. And so there's this very uh famous project from Kaggle. Kaggle is a it's really a site about running competitions and and training about learning how machine learning works. And they give you the data set and you use it to train an algorithm to work out whether you'll survive. or not. And what is really interesting is that it's not perfect because it's not as binary as these three factors you survive, these these factors you don't, but it's a game to see using the training data, which is the historical data. How accurate can you get in your predictions? And so typically you get about 85% accuracy because it'll predict some people will survive and the majority will, but maybe 10% of those it predicts won't survive,
right? So you'll never get a 100% accurate model, but you'll get something that's good enough for a scenario. And and I always think about it in the context of student attrition or student retention, say up to year 12. So that you know, can we get a ch a student to graduate? And how many bits of information do you look at in order to decide whether somebody is going to graduate or not? And is it important to be 100% accurate? No. If you can get 70% 80% 90% accurate, then you might identify 100 students in your school or a thousand students in your university that might be likely to drop out and therefore go and intervene with them to help. And so you need some training data which is all your data in the past about students that have dropped out. You pop it into an algorithm. It then works out how to be able to predict the future students and what will happen to them. So it's take the data from the past, train itself to be able to use that to predict the future.
And then and then you mentioned another word algorithms there. So let's unpick that a little bit in a second. So So we get these machine learning algorithms that are used in like a wide variety of these applications from computer vision to email filtering to every everything really in the in this day and age. You can bring lots of data sets together which we look at later. But when when you create an algorithm when when we look at what an algorithm is all that is really is sequence of well-defined computer implementable instructions. So uh a series of statements that you can kind of break down a large problem into, you know, if the doors open, then set the alarm off and things like that. So, when you get when you put all of those together,
if the email is for an African prince and it's promising you millions, put it in your junk email folder.
Oh, no, I didn't. I've already rung in. But yeah, I you know, so when you get those rules together and you can kind of create an algorithm which is a series of those rules together, then uh you can do a lot of interesting things around AI. So when when you were talking to people about algorithms and trying to explain algorithms. Have you got a way that you like to explain algorithms to people or have you got analogies that you use?
Possibly one of the best ones I I've seen is um and sadly this is an American baseball one. So when you're a coach in American baseball, actually it's probably true in English and Australian cricket as well.
Okay.
So the coach sends signals out to the team to say, you know, at this point I want you to do the following maneuver,
you know. So they'll say to the they want the bats to, you know, throw it and run.
Is that right? Yeah, probably.
Um,
yeah.
But they'll, so they'll give them signals. Now, what So, what they actually do is they touch parts of their body, you know, maybe they touch their nose and then their forehead to give the signal that this one's going to be a home run. But it's not as easy as that cuz you don't just watch the coach to try and work out what he signal he's giving because they also put negative signals in there. So, for example, they might say, "If I touch my elbow before I touch my nose and my forehead, That means don't follow the next instruction. It's like Simon says and without the Simon says. And so you think about that to to do the first bit which is to spot the signal that says go for a home run is easy until there's data that points in the other direction that you don't know about. So it's really complex to try and spot stuff like that. That's stuff that computers are really good at spotting the patterns in a complex set of data. Yeah. So think about student dropout. It might be if this person hasn't turned up to lectures this year and hasn't used this the canteen, then they're likely to drop out except if they're an online student,
right? Yeah.
Yeah. So, it's that complexity of things and what you do is
you give that to the data to the computer and it works out what the patterns are rather as well. Yeah, that's fantastic. So, so we talked a lot about data in the last and I think one of our second or third episodes we talked about data being the new oil So now we know what machine learning is and where AI comes from and then what these algorithms are. Now we got lots of this data together and then one of the terms which we use quite a lot across the podcast is big data.
Oh yes
big data. So what what are your thoughts on big data
or big data?
Big data. You say potato. I say potato.
Yeah. A lot of people talked about big data. It's it's less so now but it was such a big topic a couple of years ago
ironically being big data and being
but but what was fascinating is that many of the conversations in education where people were talking about big data wasn't really big data because big data's got three characteristics to it. The first is the volume. So are you getting so much data that you cannot possibly deal with in a human way? The second is the variety. So big data isn't just one stream of data that looks the same. It's multiple things. It's things coming in by video. It's things coming from your IoT sensors. It's things coming from your learning management system. So that variety of data and then the third one is also a V. So there's three V's to big data. Volume, variety, and velocity. And so by velocity, it means is this data coming to you so fast that you cannot possibly deal with it in another way apart from having this approach around big data. And so my example of that would be every time a Airbus 380 lens they download half a pabyte of data from it. So velocity of data you know in 5 minutes they receive half a pabyte of data. Variety of data well it'll be everything from the heating and air controls to the controls to the fuel information to what's going on in the engines. And then lastly the volume of data. Nobody can make sense of half a pabyte of data especially if we When you consider, you know, if your Emirates, you've bought 100 new Airbuses this year. So, you're getting 100 flights, maybe 200 flights a day of half a pabyte of data. So, absolutely huge volumes, velocity, variability of that data. And somehow you've got to make sense of it.
Wow. Yeah. And and I suppose that's one of the reasons why the self-driving car kind of um elements and and I suppose the stuff which Elon Musk is also doing around is um SpaceX projects at the minute that seems to be getting a lot of velocity because of those three Vss. It's really good to remember those three Vs. So, it's volume, velocity, and variety of data. Yeah. Got it. So, you can you can almost apply those things to any of those situations you're looking at, whether it's education or any setting really.
Yeah. So, so then if you think about a typical education scenario, what you've got in your student information system isn't big data.
Probably doesn't have the volume. Might have a bit of variety, but not really because most of is very structured data
and it doesn't have the velocity. But if you add your student information system data together, you add your learning management system data together, you add your web logs,
your attendance data
suddenly and then you've got IoT sensors all over the school plotting stuff and you want to make sense of all of that, that's when you have to take a big data approach because you can't possibly by the time you've integrated all that data,
the year's gone.
So you need to think about it differently. And I think that's the big data message is from a human perspective, we probably can't deal with that ourselves. That's where we need some intelligence to help us deal with it.
Got you. That makes complete sense. So, so brings us on to another thing that that kind of uh we put this out on Twitter to ask some questions and somebody did ask about data warehousing and what that means. So, if we are looking about looking at data warehousing, then you mentioned earlier on the there's an element of structured data and I suppose you can take that big data in two ways. One one is your captured data which sits in usually what's called a data warehouse and usually uses tools you know in Microsoft terms we call something we've got something called Azure data factory which essentially takes that data in puts it in a nice form um and then puts it into your structured data of a data warehouse so somewhere where there's structured databases where the data is in columns and rows and it makes sense then you've got the other element of big data over here which is um unstructured data so data that's coming in from everywhere so you got your unstructured data, your your structured data and and you using I suppose some tools to kind of analyze that. I know we released a couple of weeks ago something called Azure Signups which looks across uh unstructured and structured data. Most of the time the things that we're looking at and some of the tools out there from various providers are usually looking at structured data only and not bringing the the stuff together.
Yeah, I think my background has often been in structured data. If I think about before I joined Microsoft, I worked for student information system companies
and that was very very structured data. You you got down to a a fixated point where you go well that postcode isn't in exactly the right format so you're not allowed to put it in. So that was hugely structured data. But then unstructured data you know is everything from a document to a video like a lecture recording that's mostly unstructured or it is when you get into the data sense of it because it's just a jumble of words and images and things like that.
One thing I'd like to ask you then so once you this is an area you've been playing with it know that over the last couple of years. But once you got the data in wherever it might be when when you get that out, you got to visualize it and we've talked a lot about different tools in the podcast. So visualization and and I think that's you know one thing when I was teaching and teaching some maths you'd you'd always students would always just pick the prettiest thing you know I'll do it in a pie chart. And sometimes you're showing data in incorrect ways. You know you you might be showing some linear data over time and you're showing you end up showing in a pie chart which doesn't sense because it's you need percentages really for a pie chart, you know, so it's it's a part of a bigger problem. So what what are your thoughts when we when we talk about visualizations of data, what what are your thoughts around that?
Well, I'm going to thank you for that question and then tell you you're on the wrong podcast,
right?
Well, because you're starting to stay on a little bit about BI in education and we're supposed to be talking about AI and education. So here's a point. The old world was always about what report do we generate and how do we make this look?
I think where we're moving to in an AI driven world is how do we turn that data into information and that information isn't about here's your report it's about here's the action
AI is often about action if you think about it's jumped to the world of retail because everyone knows it when you when you go to a website and put something in your in your shopping basket and it says people that enjoyed that also bought this people watched this movie also bought this that isn't about a visualization or whatever it's about the action you do with the data and the action is I'm going to recommend this so you buy it or I'm going to adjust your insurance premium because you're a safe driver or I'm going to put it up because you're not a good driver
Dan so so it is is a different delineation in the AI
well AI when you think about big data massive volumes of data it isn't about oh let's v visualize that data to make sense of it it's about going all the way through to how do we use this information and so a lot of the scenarios we've talked about in the last 10 podcasts have been about making a difference at the youth level. If you think about all that conversation we had with Troy about AI for accessibility, it was about how does it help you to comprehend information better? How does it provide access more equitable access for students? If you think about the conversation with David, it was about how do we use this data to simplify and personalize learning.
You know, the AI step is about taking all of that and turning into things that drive action as opposed to turning into yet another report.
Yeah. And BI meaning business intelligence to really look at that as a report or whatever it may be. Yeah.
Look, there's still a place for all of that. But if we think about genuine digital transformation, the point of all of this work with data in an AI world is to go through and make an a difference to an individual.
Yes. Yeah. That's good. I I love that. That's a good analogy actually because we've kind of wandered into that area and seemingly into a trap. uh but the visualization of that data is important as an end user goes but before you even get to that point you may have made decisions using AI algorithms using machine learning before we get there.
Yeah.
So it's really interesting. So if we were going back to basics again and then people are just tuning into this podcast where do people start you think where would what would your top tips be of where people would start with with AI?
Well I I guess the starting point is the learning journey. So for For me it was some of the reading around what was possible then thinking about the data types and then what dragged me into it is what's the business problem that's we're trying to solve. So in my case when I first started working with AI I was actually working with NAPLAN data. The problem I was trying to solve is could we predict nap plan results based on other data. And so I then did a little bit of training on the techniques and and I used some of the open muks for that but then I got my own data and I started building models because I find that is what makes training much more relatable to me is I'm understand I'm trying to solve my problem not some theoretical
and and that's what we've been doing with lots of schools recently using their data to do that and similarly you know I you know I think what one thing especially because these these technologies are emerging one the one aspect that I did and you alluded to it earlier as well was the fact that I did a lot of learning around this I think you can't do enough learning because the more you know the more things and tools in your mind and the the kit bag you've got of ideas and and you know you look at cognitive services which you talked about which are visual recognition and facial recognition and speech recognition all this kind of stuff once you've done and and you understand some of those concepts then you can start to apply it and start thinking well how can I do that and using real data and real problems
well do you remember that workshop that I did at Edgitech last year my starting point was the the Azure cognitive services website and I stepped people through each of the different services, the image recognition, the OCR, the speech recognition and literally on the website we went and put in a piece of data and saw what the computer did with it and saw the data coming back. So for example, if you put a block of text into the text recognition service, what it shows you on the right hand side is here's how I've interpreted that text. You know, here's the words, here's the language, here's the key concepts that I found within it. Here's the recognized entities, which sounds a bit geeky, but it kind of says if you've got the word London in there, London is a place.
If you've got the Brad Pitt in there, Brad Pitt is a person. Now, buried amongst a jumble of words, you we as humans can spot that stuff, but computers aren't great. So, using that cognitive services website, I was able to take people through to say, look, I know we say that computers can see as well as humans. Let see what it does. And that's, you know, you load an image and it provides you with labels that tell you about the image. So, you know, a starting point is well, going there and I guess we'll put this in the show notes, won't we? Y
to go there and actually try each of those before maybe then going on a course, you know, an online course that steps you through the process. We'll provide some links to those because there's a a heap of courses in the Azure training courses. But then probably the next step on from that is actually build something yourself and the courses tend to go to that step as well.
Yeah, absolutely. And and I did one, you know, one of the things I did was a chatbot. I did a chatbot at Edite Tech actually. We mentioned in a previous podcast, but uh I created a chatbot. I used a couple of services to do that. It takes about 20 minutes. But also, as well as doing online learning, I actually found somebody who' done one and then I sat down with them and within 10 minutes they' shown me what to do.
And then once you've done it once, then you can apply it and think about it and And you know, it's just showing people really
and can I just divert a little bit bit into why this conversation is so important because I don't think I'm ever going to be a data scientist. I'm never going to build a missionritical algorithm that is used by somebody in education to predict something and I'm never going to be called on to predict survivors of the Titanic. But the reason that I do that training is first of all to be able to understand the concepts to it. But the second thing is in my job Knowing that empowers me to have better conversations. So, God that sounds awfully let me let me put it into plain English. For years, I was a marketing person. I was a marketing manager in the UK, but I understood technology because I started as a developer. And so, as marketing went from producing brochers, and I'm famous for producing 27,000 brochures where I misspelled the word literacy, as we migrated from paper to digital in in which case you can fix your mistakes much quicker. I was a much better marketing manager because I understood what was possible in a digital world. And so when I was meeting with people that said, "Oh, we can do this on your website and we can't do this." I would know some have some sense of what was possible or not. And I think we're in that same stage now when it comes to using data and artificial intelligence to transform our future, whether that's in the airline industry, in the retail industry, or in education. And So we need to understand the business applications of this stuff and some of the technical concepts because it makes us a better educator or a better principal or a better leader.
Yeah. Yeah. That's that's great. Yeah. Because I think you're right. Those conversations are really important and the stories around these, you know, that may be good for next podcast. Maybe some of the stories around AI, you know, we've tried to weave that into this podcast as we go in as well.
Well, it's interesting. There's a lot of reporting about AI at the moment in the media, especially about AI and education. And I look at some of it and go, well, the person that wrote this doesn't understand the topic and so if they understood the topic more they might be asking more critical questions or they might understand more the value of it. So that would be really interesting next time to go go through something
and it's not all about technology is it? So we've talked about some of the technological terms here around artificial intelligence but when we for example this Monday coming up we've got a day of learning where we can spend as much time as we want through that day picking up different areas uh looking at resources online, doing exams, getting certified in certain technologies, but it's not all about the technology. It's about the actual business outcomes that are put in place. And one of the things I'm going to do on Monday is called AI business school. I haven't done it yet. I know you've led a lot of those courses. What have I got in store for the AI business?
Are you really going to enjoy it? So, you're right. I I I turned out to be the first one to have done it in Australia, but really but the reason I did it is because I think it is more than just technology, you know. Yes, we've had the ethics convers before, but it's also in what areas could you use this technology to make a difference to outcomes, you know, whether it's education outcomes or business outcomes or outcomes for an individual student. And so the AI business school that you're going to go through is a course that looks at AI and does some of the basic concepts, but then it goes into the book, what difference does it make in an organization? And so in the education one, it looks at personalizing learning, it looks at predicting student outcomes, It looks at a whole range of different ways that you could use artificial intelligence.
But you know what I what I what I'll also like to do though because you mentioned there's an education stream, but I also like looking personally at the other stuff as well because I kind of get I kind of sort of think I've got a bit of an idea of AI and edu. But then when you then look at it in retail or in manufacturing, which other topics are in there?
Um government, manufacturing, financial services, retail. And you're right, they're they're fascinating when you
film tech might be interesting.
Well, it is because when you think about financial services industry. They're doing amazing things with artificial intelligence in their customer relationships. The same in retail, but they have less data on their customers than an education organization would have on its students. They might be using six or seven data points. You you've been to the website, you browse these particular pages, you put this thing in your shopping basket, and out of that they're able to influence you to do something else. But if you think about education, the huge amount of data we have on on an individual student and on the flow of students through the organization would give you so much more potential. So it's really interesting to see those scenarios in the other sectors and then apply it into gosh we could do that you know I've
and even even as even as like a person in uh you know who's living in society the government want to be interesting as well because you sit there and see all the inefficiencies of the different websites and
well you think about a smart city and traffic lights and you know all of that stuff But if ever if I can jump from a retail example to an education example
or from or or from let's say uh let's go from Netflix to education shall we? Let's see if we can make that jump. So you watch a movie in Netflix and it says did you like it and you do thumbs up or thumbs down
and based on that it then says I recommend this for you and it's pretty good at recommending things that you're going to enjoy. When you've ever done an education course you've watched a video say in a in a coursem Has anybody ever asked you to say was this helpful or not? Did you understand it or not? But imagine if you collected that data how you would be able to help a student make a journey going forward.
Yeah, I'm I'm I'm sitting sitting with uh one of my friends kids at the minute and they um they selecting a uh university courses and they're doing a BA honors kind of degrees and they got they got 40 50 options per semester and they're picking them now and there's no connection. It's not it doesn't and say, "Oh, you did really well in anthropology semester 1. These are the courses." You know, you could pick criminology in in semester 1. You could pick Pixar studies in season 2. You could pick psychology in, you know, semester 3 or whatever it is. However, they all work. It's no, there's no connection with AI at all.
I think you've illustrated why we started this podcast because the potential upside
Yes.
is huge out of this. Okay. So, next week,
yep.
Are we going to talk about news stories.
Oh, let's do that.
You bring along your favorite four news stories. I'll bring along my favorite four and we'll talk about them and we'll look at the story behind the story.
Fantastic.
Great. See you next week, Dan. Thank you.
See you next week.