Jun 24, 2020

Have you ever wondered about what makes a bot tick?  How do you go about setting up a conversational agent in your organisation and what are the benefits to you and your customers/students and staff?  In today's episode we take a look at how to create one and the considerations when doing so.

Microsoft Bot Framework: https://dev.botframework.com/

Microsoft QnA maker:https://www.qnamaker.ai/

The Turing test:  https://en.wikipedia.org/wiki/Turing_test

Chatbot in action: https://news.microsoft.com/en-au/2018/12/03/cloud-and-chatbot-free-up-it-service-desk/

Qbot in education: https://www.cloudcollective.com.au/bots-for-education/

________________________________________

TRANSCRIPT For this episode of The AI in Education Podcast
Series: 3
Episode: 4

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

Well, hello everybody and welcome back to the uh the AI podcast with me, Lee Hickin and Dan. Welcome back.
Hi. Hi, Lee. How are you doing? How's your head since the last podcast?
Yeah, look, that's my head is spinning from the from the quantum story. I loved talking about it. I loved researching it. It's a an amazing area, but frankly, my head is tired. My brain is need some sleep. So I'm going to hand over the reins to you this week and I want to get you to talk about something that I know that you spend a lot of time getting involved in. It is a really important part of of kind of AI's physical world and the way way which a lot of us probably today experience AI is through this idea of bots. These automated systems that we see we engage on in the phone or in web pages and all of those different places in our world. And I think you might have even talked a bit about this in some of the earlier episodes. codes before before me. Um, but let's start there. Dan, why don't you kind of give us a bit of a viewpoint from yourself on on bots and and and how they how you see them fitting into the world.
Yeah, absolutely. And and I think you're right. You know, we did we did cover this uh recently, but I think what we're going to do today is try to take it down a little and you know, some of the feedback we got people wanted to know, well, how do I actually create these bots? Everybody's talking about them. Um, and what are they? So, so let's just go back to basics for the start. You know, we think about about bots as conversational AI or conversational agents and you know we've seen these explored you know around the COVID environment currently in health websites you know I'm getting I I I this best way for me to order wine through the virgin wine website and um and they've got a bot there and I am going to speak to anybody I can just say I really love Savon Blanc can you send me some more um and then they don't have to mock my accent or my taste in wine either at the same time But I'm interacting with these and we all are interacting with bots really um uh pervasively now within websites within social media sites it's becoming more and more something we are used to um and I think despite being them being around since the beginning of the internet they getting their time in the spotlight now um when we start looking about I think it was in episode four we looked at bots and application and education and other industries as well and obviously when we were speaking to uh people uh like Nicole in in series one as well around the um the police force and the use of bots in certain areas there and health and retail you know these things are exploding um and and when we do look back we see them every day like I said and now you know they're becoming part of our home you know Siri Alexa Cortana and and even Clippy was kind of ahead of his time really an account conversational bot when you think back use any of those home automation? I remember you saying in the first episode that you use them quite a bit.
Yeah, we do. Look, we've got a probably a bit of everything in our in our house and and we do use them a lot and you know,
talk to it and we ask it very very simple questions, but it's and and a lot of the time it is quite simple,
you know, in terms of the interaction today in a lot of the bots that we see Alexa play me songs and, you know, Siri, what's the weather kind of types of conversations, but it' be interesting to draw a bit of a bow back. Don't want to go too back in time too far. But you know the original bot possibly would be the touring test would it not in terms of it
the way we think about communicating with a computer. What are you what's your thoughts on kind of you know is that the inception of the idea of a bot?
I I think I think I'd say it is. Yeah. And it was the first thing that I did when I programmed a computer myself you know in in basic and python and all of those kind of actually Python wasn't around then. Basic and Pascal I think when when when I started playing about and you'd you do the instructions to say hi you know, and you what's your name? You enter your name. Hi Dan. And you start to get that chat element. Going back to our first um episode in this series as well. I think people generally have been craving to develop tools and technologies that mimic the human form, don't they? And the human conversation and want to get that that element. And there was a lot of um psychotherapists. I think it was something called Eliza that was it around in the ' 90s where you could you could ask questions. You know, the there's bots from the NHS you know, and I think the touring test one was a was a good example of that to kind of really say, well, if you can make, you know, the person at the end of a um a program not realize if it's a person or a a computer program answering you, then you've kind of passed the Ting test, you know, I think that was exactly where it started. You're right.
Yeah. Look, and and it and it raises an interesting point that we may get to as well, which is you the idea of the touring test was at a time when we thought it was impressive to convince a human that the computer was not a computer. Whereas, I think we're now in this more enlightened time where we want that experience to feel seamless and smooth and we want to be able to communicate with the bot, but we also have this very fine balance of but we want to be sure that we're not emulate. We're not pretending to be a human. You know, we want the system to be ethically recognized as a computer even though it's super helpful versus a computer that pretends to be a human. When I get involved in responsible AI, ethical AI conversations. That's a a fundamental piece, you know, the the sort of the the Steepford wives challenge, the you know, uncanny valley idea of do I want this computer to be like me or do I want it to be helpful but still a computer?
Yes. The interaction
and that's true, isn't it? Because all of the home automation devices and things and I I I would almost say that I can't live without them at the minute. I I absolutely love running all my music through them, getting all my reminders, my kids, you know, it keeps my kids sane on Xbox, so I put timers on for every hour so they can switch over and you know it's it's it's really kind of um something that I can part yeah it's been really good but I but it doesn't feel like a person you know in in my house and it's interesting to see my kids interact with them as well because that they know that it's not a person as well and and it's it is interesting and I think you are right it's it's about where that sits and when you mentioned the uncanny valley there of of whether it's too much like a human you know do you want to explain a little bit more about the uncanny valley there.
Yeah, look, I think it's an interesting one in that we've had this sort of long pursuit over the history of computing to to, you know, through robotics and through computing to kind of create images in our own sense. You know, that's a very much a psychological human trait to build in their own image and you know, in at the risk of becoming theological, it's also very much tied into that kind of, you know, God created man in his own sense mindset that we want to do that.
But then I think that what we hit a point is is that we realize that as computers became more capable than us in some areas. And this is I think an interesting one where we talk about bots where
bots can help us do things that as humans we can't do you know some of them can understand your crazy accent and you know I have no idea what you're saying half the time but you so so you've got that model where they're really helping us and changing the human experience and creating these amazing experiences
but I think you know and again I don't want to call on our friends at Google but Google made a bit of a misstep not long ago when they kind of did an example of a bot communicating with some over the phone and you made a phone call to book a hairdressing appointment with a hairdresser. The whole thing was sort of built for to demonstrate the humanistic attribute you can create in a bot in breathing, in pauses, in intonation.
And it was a really amazing technologically wise amazing to think that you could create such a humanlike pauses and breaths. But what it created was this sudden moment where everybody went, "Well, hold on a second.
I don't want to be duped into thinking I'm talking to a person when it's a computer. That's too uncanny. That makes me not know if I'm in a world of, you know, Westworld or in the real world and is this a robot or a person. And that's where I think those kinds of moments are the minutes where we trigger and and humanity kind of collectively goes, "Oh, hold on a second. No, no, no, no. That's not how I want this to be. I want to be recognized for being a human and using the computer for the valley." So, I think that's where that that uncanniness is that when it looks so much like us that we kind of go,
I don't like that that that that sort of almost
dehumanizes me by creating something that is almost a reflection of me.
Yeah. No, absolutely. That's a good explanation. Yeah. And and so when I'm looking at these and we we did mention this like I said in in a previous uh episode, but you know, there's lots of examples of where we can use these conversational agents, you know, customer service, retail, um lots of the audio and speech analysis stuff for for people interacting with um social welfare systems, surveillance even. Um have you got any examples? You you you've got a plethora of kind knowledge across the industry in Australia and globally. Are there any ones that jump out at you or ones that you've thought an interesting application of bots?
Um look look there's quite a few yes as you say that we've been doing here locally and globally. Um and you know I think when you look at the data it is where you probably would expect it. So you know those industries where uh there's a high engagement with this with population high engagement with with individuals and citizens um and the opportunity to sort of streamline that customer service experience. experience. So retail retail huge amount of activity in retail um a lot of customers um around the world using it and and using it to
kind of triage I think is the right word to kind of simplify that first step customer engagement process. If you're looking for a very simple question you know I want to know where I can buy this from or I want to know if your stores are open on these dates or I want to know some basic knowledge about how to apply for a loan from a bank for example. Those are really good bot examples because they can be done through an interface that's the user. So bots can be voice and written and they can manage language differences and all those kind of interesting cool things. So we see a lot of that. We've seen here in Australia our government um you know publicly we did some work with uh uh Department of Human Services last year who are using bots to help automate and triage a lot of the inbound uh requests for information. And the idea being in that scenario is two things. You know you've got just a numbers of people calling in who you want to be able to help more quickly and a bot can do that by scale. and also those people that don't necessarily have English as a first language or those people that want to communicate uh in a different way to, you know, to to to to perhaps um more traditional mechanisms. Bots just open up this freedom to allow people to engage in systems that weren't there before. So, we're seeing a lot of that across the I'm interested your your point of view given your education background.
What's is there anything happening in the education space? Because I haven't seen anything myself.
Yeah, there's there's been several there's been um uh we've had several schools creating bots for help desks for IT. I think they've come out of they've been born out of the IT element because they still I suppose in geekland, you know, where where I, as an ex-teer, I'd like to see bots more used to support kids studying. But I think um you know, we'll start to see more of that when teachers, you know, you take the likes of um David Kellerman in in a Sydney Uni, he's he's kind of uh stepping in the IT and the education sphere, but a lots of the time the ones that I'm seeing in in education at the minute are based on, you know, it processes. So, I need to change my password. Um, where do I download Microsoft Office from? How do I get hold of Adobe software through work? Um, I need a new mouse, I need a new keyboard, you know, and they and they've actually seen significant uh savings for people's time for first line support. So, that's where I've seen it go at first. Um, but, you know, I'd like to see it push the boundaries with where we can start to interrogate student learning, bring data in and and I suppose one of the things for this and we'll step through some of the processes here, but it's also about where we get that data from. And I think the more mature education and retail and and all the other industries, the more mature they get with their data strategies, it'll feed into those bots and allow us to get better um uh you know, experiences. So you take my wine example earlier on, you know, that that was I was answering questions roughly about, you know, my order, but it wasn't really doing anything. You know, it didn't really go through my order, add some machine learning and say, "Hey, you really liked uh this New Zealand particular wine, you know, why not try these?" It wasn't doing any of that recommendations. It wasn't adding any other intelligence. It was pretty much um like you said, just triaging my order. So, in Edu, we we seen the first foray into there. You know, I've seen several being used, but mainly from that angle at the in the first instance
and I think that's probably because it is a it's an easy angle to get into know and I you know help desk type services are are an easy target I would say in some ways
but I guess you know so Dan I'm I'm interested because I think you know it seems quite simple in some ways when you think about it
you just you give it some data and you ask some questions it gives you answers why don't you talk us through
because you've done it you've created bots I know why don't you talk about how do you create a bot how does it work
so so the first thing to think about is your is your knowledge base. So there's a couple of steps to this. So step one is this initial knowledge base. So you need and like we mentioned previously, you know, expert systems, you've got to have the the actual questions and answers that we need to be able to um allow the bot to actually take a question and then feed an answer back. Um from from a Microsoft services point of view, we've got a service called Q&A.ai. I'll put the links for that in the show notes. Um but essentially, it's a way you can create the knowledge base because ultimately you've just got to have a question and answer pair. Um, and you can create those from frequently asked uh questions pages from websites. You can put text files in. You can type them in manually. But ultimately, you need to think about your questions and your answers to get that that kind of initial u knowledge base together. And that's what I was saying earlier on. You know, you lots of people have got these already on websites. So, you can use Q&A maker to point to those websites or some people might want to create those themselves. You know, I've done this with kids and kids who create a a Minecraft bot, for example, and they they've put in into Excel uh one column with all of the um uh um what they called what you call them inside Mods or mobs, sorry, inside Minecraft like creepers and zombies and all that kind of stuff. And then the description of what they do and and and things like that. So, you know, they can type in, you know, how do I um generate a creeper or how do I kill this particular thing, you know, all all these kind of cool things and they they kind of think about the knowledge base and it's good because it gets people thinking about that computational side of it. So step one is to create that knowledge base and and and I suppose where what I was saying earlier on where we seeing that being pushed even further is actually start to think about well what data have we already got that we can use and can we even use it on data that we can automatically generate with machine learning or whatever. But but actually Initially, most of the ones I see, you create the knowledge base. Nice and easy. So, once you've got that, you've got your knowledge base in place, then what you've got to do, you've actually got to um set up the bot service itself. So, you've got an knowledge base. I've used, for example, in this particular instance, Q&A maker, but all the different um uh technology companies out there have their own tools to create their own knowledge bases. Then, what you need to do is actually create the bot itself. So, the framework that you interact with. So, the code in the back end. Microsoft's got a bot framework that we kind of utilize, but it is as simple as going into our Azure portal and creating a bot. You spin up a bot. There are several different types of bots. You can you can spin up the conversational bots, the kind of main one, the that we use to to kind of share with educators and um and you know, all kinds of different industries. So, you create that and inside that bot in that framework,
you know, from Microsoft point of view, we've created lots of nuances inside there. So, there's lots of things inside the um inside the bot framework that allow you to have a sensitive bot, for example. So, you know, there's lots of I suppose conversational elements to bots that that are kind of really important. So, for example, if you've got uh a bot that's in health service, you might want that to be sensitive. So, you might want not to laugh and make fun as you're going through, but you can have a funny bot because you can have some bots that you might want to entertain the out with when you you've got on your website and you might wanted to add some fun stuff, you know, cuz don't know about you, but every time when when I was a kid or whatever or developing these bots, you type silly things in and you do it to Alexa and home automation devices as well. They've got their own little pieces of personality in there. So, you set up your bot service in your cloud and you add, I suppose, any nuances you want to that that come out of the bottom such as sensitivity or whatever. So, then you you kind of got the bot in place. Then what you can start to do is you can start to think about any AI services you want to connect. So the the the first element once you've created the as you're creating the the bot service itself it says hey tell me first of all where your knowledge base is and then what we do we point there to the knowledge base that we've created. So uh there's a couple little steps in there but it's pretty straightforward. It points at the um the knowledge base and then it says ah okay I know where my knowledge is. And then you can do you add the nuance like I said uh if you need to and then you can also point it to any other AI services that you want. So for example you might want to bring in Lewis which is language understanding and intelligence service. So the reason for adding that service would be that um rather than typing in exactly the the text in the question you know if I'm a consumer and I want to know how do I change my password that's fair enough but if the question is something a little bit more vague. Um then uh you might want to put in some other intelligence to that such as Lewis service or cognitive search or translation services if they're people from different languages. So you can add intelligence to that based on premade AI services or what we call cognitive services in Microsoft.
All right. So let me ask you a question then. So that's a because there's a lot of data that you've gone through. I think it's a lot of construct but
it's basically you know in it very basic sense. You've got a knowledge base of data and you're giving that to a bot engine that's going to either automatically go figure out questions and answer pairs or you're going to teach it kind of reinforcement learning or or supervised learning versus unsupervised learning but it's going to learn how to answer questions and I think that
when you talk about the embedded nuance potential there because they're talking about now almost giving it a character giving it a bit of life um which is a fascinating thing so it is more than just simply a Q&A process it actually It can I guess it can do a whole bunch of stuff like translate languages or or or try and understand what people really mean when they ask a question in a particular way. Is that the kind of stuff we're getting into?
Yeah. Yeah, that's right. And and also it's about that inclusivity as well because we we got to think about people who are interacting with bots and uh you know the standard thing when I think of a bot is that I'm going to be typing in a question somewhere on a website. But that might not not be the case for everybody. You know for disabled people, people with visual impairment, people with with um other disabilities then you know we need to be able to interact with uh everybody in society. So we need to think about how we'd interact and it might you know the bot might not even be needing a question. It might be that I take a photograph of something um and it and it gives me information on that. So there's a whole raft of artificial intelligence like you said machine learning models that you can also apply to that in the back end to help it learn as it goes as well.
Well and and an interesting point you just made there which is because I think probably because certainly for me it was the thought process when you think about a bot your immediate thought is it's a a computer system that you type answers to and you talk you talk to it through the keyboard
but of course bot framework from a an idea it could be something you visually you know you take pictures as you say and show it a picture and it can use AI to figure out what the picture is you could talk to it and language translation so it's pretty broad when you when you think about a bot it doesn't have to be just that kind of very you know, help desk kind of online service. It could be voice, it could be images, it could be video, it could be a whole bunch of stuff. AB
absolutely and that's where it gets really interesting and that's what I think we starting to see as more and more of these services get created as well. We're going to start to see that connection in with machine learning which we'll talk about in an upcoming episode as well and how machine learning works. But then also like you're saying some of these other elements, you know, which which a make it useful but also mean that there's tools and technologies that you know might help other people such as your translation features. So once you've created your bot and you've got you've pointed it to your knowledge base and you've got your your own characteristics in there if you want to set that up and added any of the other AI things, you know, you don't even need to do that, but you can. Then all you've got to do is think about where do I want to surface that bot? Where does that actually sit? And you know, I've I've seen bots inside Teams for example. So that's there's lots of different bots you can they're already created from Adobe and things like uh which is inside teams, there are bots inside Facebook, there are bots inside websites. So you got to think about where you want that third party service. Where do you want to where do you want your your bot to actually sit? Have you seen any bots in any interesting uh situations across industry?
Uh you know that's a really good question. If I think about it now the first thing in my head is I'm thinking okay did I have I interacted with a bot or was it a person or did I even know the difference? Um
I look I don't know if I'd call it interesting. I certainly see them to be more prevalent in almost every kind of business that I interact with. So I mean you know to give a a very broad example I mean certainly when you when I go and log into um you know government services and I know I would greeted with a a help bot that will ask me questions and often can help me just navigate towards what I'm trying to do more quickly. But to the same way after I was away at the weekend I was in uh in Baitman's Bay for the long weekend going there sort of, you know, up in the community. Good. Oh, great. And we wanted to order some food. And so you log on to a sort of local pizza company and
look for your order and a bot pops up on the website of a pizza company in a regional New South Wales town saying, you know, can I help you choose?
And again, it's not, you know, it's probably not a super complica complicated bot. You know, it understands the 22 pizzas they make. I tell it what I like and it'll recommend two or three for me.
But it's a very, you know, it's kind of a it's an unusual place to see. I think in some ways, but it's just it expla shows how how bots are so simple, but they can really simplify that process when I don't know this pizza shop. I've never been there before. I don't know what their pizzas are like. And I happen to know that I
for me I only like pepperoni pizza. Pretty basic, but that's all I
So when I just write pepperoni, give me lots of pepperoni.
For me, it's a very basic example. But
it is what and it's it's landing where the people are going to be as well, isn't it? I suppose in and I know we keep mentioning COVID and things and I think this is you know and whatever other pandemic comes out next or whatever it might be. Unfortunately, you know, I think governments are looking at bots and the way that you can interact with people to give large groups of society advice on things and and they try to embed those into platforms which are being already heavily used or platforms like Facebook and all the social media platforms are are giving other companies uh ability to kind of connect in via APIs or whatever it might be to embed bots in into their um in systems. So that so that is an interesting one as well, you know, like I think if you're in a school for example, where are your parents going to be? Are they in Facebook communities? Are they going to be in teams? Are they going to be on your website? Where do you want to interact with those people? So I think the point of where you put the bot is really really important. And then the the final part I suppose is once you've published that bot and it's all up and running, one thing that we forget about is maintaining that knowledge base. And you can either do that like you said earlier on, you can have machine learning to kind of uh actually learn as you go and as people are putting questions in tries to work out the answers and ask questions through reinforcement. You know, is this what you need? Does this answer um you know, solve that question and makes its own um knowledge and question answer pairs? But maintaining that database and uh or that knowledge base is really really important to kind of drive that further. And when when we when we then looking at it from an enterprise architecture point of view and You know it sounds oh here we go Dan's going to talk about enterprise architecture. I think it's really important because you know it does open things for us. So when we start thinking about where our bots are you know it's important to be thinking about any security we might have there any kind of um processing that we want to have done because you know one of the other things that I worry about and there is you know I don't know how much research done is done on this but every time you know Lee we we do the thing in Microsoft don't we where we can um uh work with schools on e safety and things like that through the think you know program. And one of the key things when you're talking to students about you know really um important issues about social media and grooming and things like that online is if you do mention something and it creates a reaction then you have to catch that somehow. And I think one of the things with these bots is you know especially if you're talking about health and things like that um or or anything really. You've got to think about then your logic applications and your functions that actually happen after that. So if if the user says something that that provokes a sale, for example, then what are we going to where are we going to drive them to? What are we going to do to keep that going? You know, I I I was on a bot on Facebook the other day about trainers from a from a US website to buy some trainers. And you know, it kept appearing and kept asking me other questions. Hey, Dan, you didn't buy the trainers from me kind of thing, you know, you really want them, you you really should have them, you know, and he kept giving me more advice about why I should, you know, it got irritating at the end, but at least it did follow up with that. Um, and it did take that to the next level. And I think the the um the kind of integration with a flow back into your sales process or into your pastoral care if it's a school, you know, if you've got a bot, for example, to support psychological well-being of of students, Um, and then you give them a bot that interacting with and then they do mention something, you need to get that back into your pipeline so that somebody can deal with that particular process or discussion or question so you don't just leave them in a bot and in an unanswered state. So, so there's kind of um really interesting things you can start to add to this bot and and think about for the future.
Look, it it is and it brings up a really to me a couple of really important things that that we should probably talk a little bit about one of which is you know when you we talk a lot about bots in the way that I'm going to ask it a question it's going to give me an answer and I'm going to need something and it's going to tell me what I need to go to do
but really what we are talking about and we use the phrase often is conversational AI and conversational AI implies that there is a two-way engagement and and as the bot is helping me I am helping the bot
you know in its very raw sense and so that it does and it opens up this conversation of okay so when I engage with the bot and it asks me questions about what do I like or where do I want to go or what am I trying to do. It's also going to be learning from me, you know, based on who I am because it knows me as a customer in some way. I've obviously had to be recognized.
It knows who I am and it's going to start to make some machinelearned decisions about well people in, you know, in that category that you are
are always looking for this kind of information and they don't really know the right word for it. They always they always call it this and they, you know, they that what they really mean is is that and so you have this scenario where the bot is learning from you and you are learning from the bot and it's kind of that you know to get back to the ethical and the uncanny valley conversations it's about do you really it's important for bots to be I think at least it's important that bots show disclosure that you know that when you engage with a bot that it is using that data in order to improve services that you are talking to a machine and I think you know in the example you gave say pastoral care in students in schools If a student is engaging with a bot and they're asking it questions and the bot can sense because it's been taught through the AI services you talked about, it can sense anxiety in the language. It can sense uh you know that that student is maybe going through some uh some personal crisis or some other mental state either through you know we can do this in voice bots where we can hear the tonality in the voice or the use of language.
You kind of have an opportunity there as you say to to help that individual and that's That's where bots and AI services can really come into their own because that that can often be missed by humans and we're often very busy and don't pick up the nuances. Computers aren't subject to those challenges and bots will hear the nuances a billion times over and over again as if it's the first time they've heard it, you know, in that very fresh way.
So, you've got that opportunity for bots to be more than just a means to learn, but they are a means to actually help and provide emergency services in some cases or just simply, you know, that offer of help, you know. It seems like you might need this. Can I almost like the clip you seem like you're typing a word document, but can I help you?
It it is. Yeah, it is. It is interesting, isn't it? Because like then, you know, when you were talking, I was thinking, well, you know, if I was in a medical situation and there was a um, you know, an emergency and sometimes we've all been caught up with this when we ring up it support, right? Haven't we? And and you get really frustrated with a person on the end and then actually some cases it's it's interesting because you'd think, well, I prefer to have a person on the end of the line if I'm ill or there's an issue or I want to speak to somebody. But actually, if you can get to the same point with a bot, you know, it's an interesting psychological point of view, isn't it?
It is. And and look, and I'm not a in anyway medical professional, but I understand there are some scenarios where there are those individuals who are differently aabled who have challenges being with verbal communication or or are less or non-verbal in some cases or just struggle with communicating. But bots offer an insight or an opportunity for them to communicate in in processes you and I take for granted
in a way that suits their particular needs. And so it it you know it's a it's a it's not an easy one to answer because there's no right and wrong. There's no you know bots are bad because it's automation and killing people's jobs and people aren't in you know you're losing the human contact because for some people in society you're actually creating a new contact
and and I think you know there's lots of people who have you know when when you look at computer games and things like that where where students or kids or adults with ADHD and other things like that are kind of communicating and connecting in with bots inside games and and you know when when we start thinking about the home automation kind of side of it where does you know a home automation device sit in terms of a bot is it really a bot is it conversational AI would you say
look for me I I see it as a bot I see it as an automated process for me to connect outcomes with asks and and I And and in that exchange of, you know, when I asked the devices to play me my favorite 1980s music, because let's face it, all the good music happened in the 80s. Um, I know that at some point it's also remembering that I happen to listen to 1980s music. And if it happens to be that that serves me up an ad later on for the latest 80s festival, I get that connection. I understand why that's also because a lot of people
that that is a step too far in the ethics and the kind of modality of what I expect from a bot.
But for me, I get it and I see it as that. And I don't see Yet today I don't see those devices as being conversational. You know, I see, you know, I talked to Alexa in our house. I shouldn't say that too loud because she'll start listening to me.
I know that's
I but when I say that, I know that she will hear what I say and I can follow it up by saying thank you and she'll say thank you back. But I also kind of know that that's not really a conversation. That's an automated process.
So I think we're not quite, you know, conversation is something that is quite different to the bot experience that I see we largely have today.
We always interact with bots when we searching online as well. If we're using Bing or Google, you know, it's all about answering those questions and then personalizing that response. And I suppose the more people are using bots, if we creating a bot, if somebody listening to this podcast today goes out and goes, "Right, I'm going to go and set up this bot for my school or for my company or for my organization or a charity I work with," then that's great. But then thinking about it from that ethical point of view is really important as well because people are starting to,
like you said, be discerning about how they're using things and how much data does or is used to personalize that service as well. So, that's kind of a really important thing to think about.
It's a contract we all enter into whenever we engage in any kind of service like that. Sometimes we know it because it should be implicit. When you buy a device like that, you you understand that. Sometimes it's not so explicit as in when you're on a company's website or a phone system and you're being navigated by a bot, but you don't
and that there's an ethical quandry there. I think, you know, Ethics is a is a challenging mindfield and it fundamentally comes down to I think two key things is that whenever you are integrating a bot into your systems whether you're trying to build a conversational bot you're trying to create a knowledgebased bot or just a fun bot for people to experience with. You know you need to have a level of transparency. You need to be clear about what that bot is and isn't and what it does and why it does it and how it does it and what it does with the data it uses. And then you need to be accountable. You need to give give individuals the opportunity to interact with a an actual person, not a nameless face system or or an automated tool when they want to talk about their own privacy, their own ethics, their own concerns. Accountability, transability, those two things are fundamental whether you're talking about bots or anything, but in this world of bots, I think they're they're pretty key.
Yeah. No, that's and that's a brilliant point to end on because I think that's that, you know, that's really good food to food for thought for everybody that's um that's thinking about this. We'll put some notes in the in the show notes around some places where you can go to get some of the learning for this. Uh there's a lot of stuff on the Microsoft learn website and on uh GitHub and there's a whole heap of uh partners out there that do lots of really cool intelligent based bot tools for for schools and for industry. So we'll put all of those in the show notes. So just finally before we go then Lee, what what are your thoughts on bots into the future? Where do you see them going? Yeah. No, I'm glad you asked me that question because I have a I have something I wanted to share and get your view on as well.
So, we didn't talk about things like Tabot. We didn't talk about that and and that's probably just as well because you know like that's been talked to death. But if we think about the future, it would be kind of a bit remiss of me because you and I both work for Microsoft and I think you know there's
Siri and Alexa and and all these sort of tools out there and we had a product called Cortana as you would very well know that we that we built into that market and we're taking a different tack with.
I'm not saying Cortana's there yet today. But I I use Cortana myself. But there's a this is where I think bots are going today. My interaction with a device, a bot or a system is typically uh non- conversational. I ask a question, I get a response. I'm ask a series of questions, I get a a response. But it seems to be it's almost kind of that almost decision tree knowledgebased kind of way. Ask me enough questions, get answer.
What I like about what we're thinking about with Cortana, and I don't know if you've played with this yet yourself, is
it becomes conversational. What I don't say to Cortana in the future is Hey Cortana, what is my what am I doing at 1:00 today? I say to Cortana, hey, how busy is my day? Do I have time to go and have a coffee with Dan later on today? And what Cortana is working towards doing is pulling out all of the bits of data and applying a sense of knowledge to that data to understand what I mean when I say, yeah,
is my day busy? Can I have a coffee? Well, that's asking so many deep questions about what do I consider to be busy? Is my day there? Where is Dan? Where am I? Where would we go for coffee. Where do we like to go for coffees normally? And when Cortana comes back and says, "Yeah, you've got a half an hour at 2:30 and Dan's actually going to be around the corner and there's a coffee shop that he really likes, so you should take him there."
That's the intelligence of a bot that I want to get to.
Not the one that just says, you know, here's 14 coffee shops nearby. You go figure it out yourself.
Yeah. Yeah, that's a really good point actually. I love I love that point and it's good to see how Microsoft pivoted slightly differently around the Cortana engine. Um Yeah. And and I like the way from from our point of view the of you know working with it in terms of productivity you know I think I feel more that the other agents around the home are more obviously consumerbased whereas the when I'm using Cortana it's you know I don't really think about Siri or Google as as work based things they're kind of things I do at home to listen to music and and cook and things whereas like yeah Cortana can kind of search my um files so I use that quite a bit the search inside inside uh Windows 10.
A lot of people may not even realize this. Good to know that you know Cortana is alive and breathing obviously. It's very much built into Windows and and our Windows platform,
but it also hooks into Alexa and Google and uses them and pulls those data. So you can get
you can ask Alexa to connect you to Cortana and then have a conversation with Cortana that way. So we're not you know this I think is a bit of that new Microsoft that we kind of look at and go well that's what people have got so let's find a way to make it valuable to them and bring our tools to There they are. So anyway, I think
Yeah. No, absolutely. That's brilliant. Thank Thanks, Ali. And thanks again for a brilliant episode. See you next time.
See you next time, Dan.
Cheers.