Jul 14, 2021

AI developer Tom Resign talks to Dan and Lee about his work around AI development emerging technology.

From Design principles and inclusive design of AI to GPT3 and even Minecraft Music! 

________________________________________

TRANSCRIPT For this episode of The AI in Education Podcast
Series: 4
Episode: 8

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

welcome to the AI podcast. Good morning, Lee.
Hey, Dan. How are you? How are you? It's a very cold morning for Sydney. I think it's the coldest morning we've ever had. It was like four degrees this morning when I woke up.
Yeah, and it it was very cold. I haven't checked the guinea pigs this morning. Hopefully, they're not blocks of ice. Um But I went into the the new uh Sydney Microsoft office yesterday. Have you been in yet?
I haven't. In fact, I'm heading into our new building about an hour's time from now for the first time. So, I'm to So, don't tell me any secrets. I'm going to get in there and discover myself.
I had a customer meeting in there yesterday and I didn't know where anything was and it was the most embarrassing thing, pointing to rooms that I didn't know where they were, walking around aimlessly. But, it's a fantastic building.
I think I'm going to be doing that myself.
There's lots of AI in there actually in the actual building. Some know the intelligent environments around where people sit and the way they work and things. So, we'll have to possibly get somebody on to talk about buildings and and AI in the future. Absolutely.
Yeah, it's really good.
We absolutely should. We should too. But we should get to today's topic and I guess this one's an interesting one Dank because you know you and I have well we've been talking about this for a long time and uh the story behind this uh today's uh topic is you know we've talked about tools that we use here in Microsoft quite a few times. We talked about a product called Viva. We've talked about Microsoft Workplace Analytics and we've sort of looked at the good and the bad Dan of you know how these technologies help us do our job you know from the point of view of AI better understanding your well-being your work experience you how you work and what things are interesting to you um which has been like you know Dan you and I both using today yeah correct
yes exactly I know I I and I think in the last one last podcast we recorded or one of the last anyway we I I'd explored some of the insights tools insight teams and there was you know the meditation and the the the kind of virtual commute and was absolutely stunning. But it did ask me to close my eyes, if you remember, while I was driving and and visualize my day. And I was like, "Oh, I better I better not do that."
No, there's a time and a place time and place. So, um, but so, look, and that the great thing about having these conversations and it's really a little bit humbling for us is to know that there's actually other people in Microsoft who listen to what we talk about. And so, we had a gentleman uh by the name of Tom Resing is and reach out to us uh and sort of ask us about the topics. And so, we said, "Hey, Tom, If you want to talk about it, let's get you on the show. So, um, welcome, Tom, to our show.
Yay. Well, thank you for having me. Yeah, I know. I really enjoy the show. I, uh, it's amazing to hear co-workers from, you know, thousands of miles away talking about things that you're interested in. And, uh, I really appreciate what you're doing. It's a very enjoyable show, I must say.
Thank you. Thank you. And it's kind of for us it's it's it's interesting that, you know, hey, someone in Microsoft listens to our to our show and what we talk about. But when we think about these tools like like Viva that we're going to talk about today and some other things perhaps we often forget that there are people behind that people who are building that and and so you know it's great to have you here to be able to share that with us. So just a little bit of an intro you know Tom's a senior content designer here at Microsoft also I believe a certified SharePoint master which sounds incredibly formal.
Wow. Um yeah but so Tom why don't you tell us a little bit about yourself who you are what you do and and kind of how you came to be in the role you are today. Yeah, absolutely. Yeah, content designer is kind of a new thing. Um, it's like an evolving area in product design. I'm I'm really lucky I think to be here at this at this time when it's emerging. Um, and I came about it I think a kind of a fairly securist route. I um I've been in interest in computers forever, right? I had a computer science degree out of college and you know was doing Java programming, net programming and leading engineering teams. Um, and then I got into consulting and that's what led to that Microsoft certified master. I think I after I did my first uh SharePoint implementation in uh in 2007 basically as far as consulting gigs went, it was like every new gig was a SharePoint gig from the growth was just phenomenal
and the demand for anybody who knew who could spell SharePoint uh was was so high uh and that led to that that certification which I what I have always appreciated about Microsoft certifications they're very practical they're very base like because I'd been working in SharePoint for so long it it was that way I'm it was hard that Microsoft certified master program they've discontinued but it was basically you may be familiar with the rangers program
yeah it was basically like an extension of the Rangers program but like So I was the only external uh person in that training right everyone else was deep technical experts from inside Microsoft. It was amazing opportunity a really highlight of my career so far and I still have friends inside Microsoft from being in that program. Right.
That is so good. A lot of this it's an interesting SharePoint's one of those tools that still underpin everything about the tools and technologies we're doing today. You know it's I I'm still having conversations daily uh about the way the people set that up and governance and compliance and you know it's still you know it's one of those structural foundations that that is a gateway to so many other things in the Microsoft world.
I was just listening to Jeff uh Teper talk internally today and he was highlighting something he said at build which I don't even the numbers are just ridiculous. I don't know what he said. I think he said we're adding I don't know 100 terabytes. Customers are adding 100 pabytes. I don't know what he said. It just sounded ridiculous like a day, a month. I was like,
the growth is phenomenal still right now.
Yeah. No, absolutely.
It is crazy.
So, so, so let's so so the so the certified SharePoint master is kind of sounds like that was something you did, but now you're this content designer and I think well, you tell me you tell us what team you're in today, but you know, obviously Viva's in your purview and the work that you do today.
Yeah. Yeah. So, I um So I've been working in this area like technical writing and content development and we were calling it UX writing this part of the product design on the one drive and sharepoint team for about five years and about two years ago uh this newly formed team called content services and insights inside one drive and sharepoint team uh came to our team the you know the content team and said hey we really need somebody to help us you know get the language right for this these products we have in development and we didn't know it at the time in fact the naming is so so hard I it's so like that's a marketing branding thing it's outside of my uh purview
uh I mean I I gave some input but the names came to us basically from uh from others uh so they ended up being called SharePoint syntax and uh Microsoft Viva topics uh these two uh products that this this team developed at after years of uh of work. Um, and they're both AI related. So, uh, that's one of the reasons I've been listening to your show.
Awesome. Uh, well, look, we I think we'll definitely get into the the Viva stuff. I think that's going to be really interesting to understand kind of what that is and how it works and get some of your perspectives on that. But I have to say because I think when we spoke previous to you coming on this, you mentioned that you obviously have a connection to Australia as well, don't you? What's your connection to Australia?
Yeah. Yeah, that's right. Um, My sister Joanne moved to Australia to get her PhD um many many years ago and uh her daughter L was was born there and uh lives in Melbourne. And my my sister and her partner actually are uh they just built a house outside of Melbourne and they're going to uh retire. Uh they are from they're from they're in Karens right now. So
wow. Oh, they got to go from hands to Melbourne to retire. That's that's usually the only way in Australia.
Yeah,
it's about like I think it's a 2hour train ride from from Melbourne where they're going. So, uh somewhere in the country K is about a 55h hour train ride probably.
Exactly. From Melbourne, it's a long way.
They wanted to be closer to their daughter. Yeah.
But that's awesome. No, that's that's awesome. And see see now you you got a now you got a reason to come down and visit us sometime when when when borders open again in Canada.
I I on the find as well. There's a connection with Minecraft as well. Tom, tell us a bit about that.
No, it's really s um when we get into like one of the reasons I love your show is is when you talk about education um I I want to know how to educate people about our AI features, right? And so when you think about te teaching people how to use AI, that that comes into it. And and when you think about like the youth and and how the younger generations are thinking about AI and technology. It's amazing. So, my niece L um there's pandemic. She's produced an album. She's written all the songs. She's recorded most of them. She's done some videos and then lockdown, right? And so, she was going to launch it, right, like you normally do with a tour and, you know, and at festivals and she couldn't and she just kept putting it off and finally she decided uh They did splendor in the grass, you know, famous. They did it in uh in Minecraft because um they needed to do it virtually, right? And so she reached out to the group that had put pulled that off and uh got some collaboration and they started building out a world in Minecraft uh to release her album virtually.
So So she So she not only built a Minecraft world to create a virtual concert tour or essentially a virtual event and then she wrote all the music to go into this concert as well that she wrote an album worth of songs.
Yeah, she released it. So like you could find the songs in the world like you would like open a chest or something, you know.
Oh wow,
that's phenomenal. We have to share the links to that in the show notes.
Oh, we'll have to get some details. Yeah, for sure. Like does like I mean does the Minecraft team know about this or do a lot of people do this? I've never heard of anyone do this. They do. Uh they interviewed her. There's a YouTube video. I'll send you a link. Uh
dude, please.
Yeah.
Wow. Awesome. That is.
And is it in the same sort of style as the music in Minecraft, like the music that C418 does for the rest of the game?
No, it's just because, you know, like we we talk about how there's digital natives, right? Like I mean she just grew up in this stuff and so you know her roommate was running a Minecraft server and was when she was in college. out of their their, you know, apartment in Melbourne. So, um, it just came to her. She s and she saw that Splender in the grass thing and she was like, "Oh, well, that was neat, right?" Like, I could do something like that. And what the best part is is that
it wasn't just her building it. It was like she built up this community of the the people that had been going to like concerts with her and they came in and they built things. I even spawned like 50 llamas which they had to deal with later time. That was my contribution.
It's like there's just two things. I mean we should get on to Bieber as well. But when I think about that like just the when we think about the way the youth looks at technology today and when when kids and young adults look at technology they don't they just say well I can do anything you know like nothing is a barrier. I can just do whatever I want to do. There's no well you know I really shouldn't couldn't do that. And then the collaboration the fact that you do these things and it doesn't matter who you're working with that person the team could be anywhere. I think it's just We could have done a whole episode on this D. I don't
Yeah, I know. I know. I'm fascinated.
Definitely.
It is. In fact, my my 13-year-old daughter buil built a house uh with them and and it's in the it's in the world, too. Uh and it's in some Queensland style. I I forgot what it was, but yeah, like a Federation style thing, I guess. I don't know.
Well, so Tom, one action out of this is we're going to have to get your niece and your daughter on.
My niece will do it. My daughter is much more shy right now. So,
okay. Anyway, we should get back on topic, shouldn't we? We should get back to to what you do.
Um, that's awesome. So, uh, look, hey, Tom, I mean, we're really keen to get your view on on AI, but why don't we start with maybe give us just a couple of a brief introduction to what actually is Viva and the platform that you work on. How does it work and what does it give to to users of it?
Yeah, absolutely. Um, so you you touched on it in the previous uh show, especially Viva Insights. Uh, um you know it's an employee experience platform and uh basically we've launched you know these four modules learning insights connections and topics and the team I work on focuses specifically on topics and uh I think you mentioned topics you got an email asking you for some input on a topic one of you did and um that was the first thing that caught my my attention because right now those emails are only uh only Microsoft employees are seeing those emails. It's not actually a feature of the published product. Um but it could be potentially but um but it's part of the idea that humans are are very necessary in giving input and training AI models. Um and that's also the other product I work on SharePoint syntax is a it's it's really a model building experience. We're bringing things that you could do, you know, with like Azure cognitive services and um AI builder and Power Platform into the modern workplace, right? And so that we're empowering information workers. So with both of these tools, the idea is right there's just so much information and AI can help you kind of control it and and take uh you know, claim some more power over all this terabytes. pabytes of information you're you're you're adding to Microsoft 365 every day, right?
Yeah. I I think
Yeah. I think from from my point of view and I'm looking at, you know, we've done a couple of episodes on this as well where things you know the simple stuff that just happens you know Lee Lee alluded to earlier on where kids take things for granted and you know they're aware of of this thing you know things happening for them you know simple things you know I think my first like epiphany moment when I was looking at AI was a simple thing when And it's not really AI, but I suppose it is. When I was attaching something to an email, you know, about four years ago in Office 365 and then, you know, the system I didn't have to go browsing and hunting for documents. I said, "Hey, Dan, you're working on these things, you know, last pick from one of these." And then that that saved so much time and you didn't even realize it was AI. And I suppose that's the that's the interesting part that we could unpick as well. So, you know, if you were developing it in the back end, how do you you know, you know, I suppose it's like a giant sweet shop of tools and technologies that you can do to make make work the workplace better. What like where do you start, you know, when you're looking at these things?
Yeah, it's amazing. I we've uh we have a great uh customer research uh group within uh the product team that's working on this and and at Microsoft in general, we do a lot of great uh customer research. And so um we really you're right, Dan, there's just so many tools and this is I think one of the things that SharePoint's always done well as a team and so this has extended out to this team is yeah said like oh we are a platform for bringing in all the great tools Microsoft has to offer and and integrate them and and allow people to build on them themselves right always make sure there's extensions either for like these model builders who are like advanced information workers kind of makers like my niece the Minecraft maker but in the in the information world uh or if they're you know developers right to have APIs to extend it.
So is it so so it's interesting so is it a so you're saying it's almost like a feedback loop so uh because there's there's there's a thing you don't know what you don't know right so it's like workers like I didn't know that I needed to have that drop-own list of my attachments um so uh in product group do you have a you have customer input and customer voice to kind of feed back in to think about where the AI is used or you do you sit there as well and go well the customer doesn't know they need this at the minute but this is a really good use of the technology well I think there's two things right uh first we we talk to customers about what they need and um so this one of the scenarios like for Viva topics that's pretty common that our customers told us about is onboarding uh so you come on either you're new to an organization or you're new to a team and they use a lot of terminology that you know is is foreign to you, right? And and it it can be very confusing, right? And so this is one scenario that we heard from customers early on that they said, you know what, this kind of thing could really help with that. And uh so that's an example of listening to the customers to develop before we develop the product. Right? When you say what we could choose all the things we could do, all the different Microsoft research technology Azure co cognitive services power platform work we could bring any of that in to Microsoft 365 but what do we focus on and that's that's how we start right
yeah that's fascinating
and yeah look it's when when I got those emails you know the ones asking me to kind of provide feedback on a topic or a subject and I apologize if I let anything out the bag I didn't realize that that was in
No it's not I just wanted to let you know that it's only you see that right now it's but
for sure so so I can see how that works in a test like let's call it an internal testing uh you know a dog fooding kind scenario where I'm I'm helping to as an employee to kind of teach the machine and give input to the models um like in a customer scenario is how how are they going to teach their own models and learnings does it have a similar mechanism externally or do we look at that kind of differently because that's you know requires a different nuanced mechanism when you're talking in different customer scenarios
yeah and that's the next feedback loop I think that Dan you were also you know, kind of leaning towards is um you you can train the models to learn about your organization, right? So, uh so we've kind of got it uh built out in a couple different ways and we have found internally that the email's been a a great way to collect feedback. Um we just haven't built that out into a way that um it's extensible that works outside of Microsoft because we have a different kind of contract with you uh than we have with our customer. Right. Um and so in product what you see is a couple different other fe feedback mechanisms. Uh one of them is if you hover over a topic that you see uh in SharePoint right now and then later in other apps as well uh or in search um there's sometimes there's questions on the bottom like um basically asking if we got it right. And what that'll do is that'll tune that model to to your organization, right? So, what does this mean?
You know, what is Minecraft and education? Does it have a specific meaning for us or you and your or it might even tailor it, you know, more to you uh differently than it tails tailor it to me because the people that you're talking to to about that and the documents you're collaborating on about Minecraft for education are going to be different ones than than I am and I might not even be able to see the ones you you're working on. Yeah. And and I suppose it's interesting you mentioned all those big numbers that Jeff was talking about in in a in a call today like we need you know when you look at SharePoint we need this support as well and and people uh with the abundance of data out there without AI it's almost impossible you know Delve you I know we're talking about Microsoft here but things things in multiple platforms as well but when we looking at Delve that was a bit of an epiphany moment for me again where where I'm sitting down and then I'm being served served up the things and the content that my boss and my manager and my co-workers are are bringing to me via AI. You know, there's so many documents out there, especially in SharePoint. You know, going back to my customer stories there, you know, they're always talking about, well, I've got all these documents. I'm moving to SharePoint online, for example. What do I move? What compliance and privacy and governance policies I need to put in there? I don't know what I've got, you know, help. And without AI, it's almost impossible, right?
Yeah. And I what you're talking about both with in terms of like most recently used documents right which I think you mentioned basically with the outlook scenario or um you know documents around you like people you collaborate with often these are all really parts of the Microsoft graph right and um the Microsoft graph is basically input to this to this AI right it's it's part of the big data for the machine learning right and uh that is one of those well of information, right, that uh that the AI is kind of trying to tap into and also kind of narrow down for you, personalize to you, you know, the right information like you're saying at the right time like an outlook there.
That that's it's funny you say that point about the right time, Tom, because I was as I was listening to you talk and thinking about it like obviously this is going to get better over time. That's the purpose of any AI modeled system. It's going to learn more about how you want to operate and going to get better and better at servings up serving up content that matters to you. What I'm finding to be more important and Dan mentioned it earlier is the the timing of that and the place in which it's served to me. So, you know, I'm when I'm in I've noticed a little tricky thing now like when I'm in a teams meeting like I am now and I've got a word doc or a powerpoint deck open the two are connected and it says hey I see you're talking in a teams meeting do you want to talk about this presentation which I know is a little bit clippy you know we could get down that path of the kind of the hey looks like you're typing a word document but I love this idea that it's not just about getting through the noise of information, but actually thinking, oh, this person's going to want this information right now. This is the point where it's going to be most important to them. Is that a part of your thinking around the Viva model as well about kind of perfect timing?
Yeah. Well, absolutely. And uh I know a little bit later we wanted to talk about the um guidelines for human AI interaction. Um but this is kind of a interesting segue to that because um you know, one of the guidelines is really about context, providing the right information at the right time. That's something that Microsoft research has determined um that is very important and they they co codified it into a guideline for anyone developing uh AI applications, not just Microsoft uh developers, but this is shared publicly. Anyone can go look at these guidelines and apply them. And in fact, you can see what my favorite part about those is you can see how we've applied them in our most popular products like you know the office products with uh like PowerPoint designer or I don't know if they have one for that particular one that you're talking about Lee but I love that one honestly what I love and I think what avoids the clippy situation is when we get it right right because the problem with clippy it wasn't that it was the right idea right and now this is one of the things I love about Microsoft too right
controversial thought there yeah very controversial yeah go on
it's you iter we just you just got some you have to try not be afraid to innovate, right? And uh and learn from your mistakes, right? And and continue to iterate, but don't don't let that initial setback stop you from continuing to try to figure out what the right way is to do this. Um and I think with some of those insights, I at least what I'm seeing is like they're are providing me interestingly useful like how they make those connections that Yeah. I'm having this meeting. Yeah, you're right. Like I was looking at that document yesterday, but I closed it, right? Maybe I should review it before the meeting starts.
Yeah, it's I find that that and just, you know, that little insights in Outlook that knows that, you know, I've had a conversation with Dan and so my history of conversations with Dan or people associated with Dan, he's here. You know, it's it is super valuable. But of course, and I don't want to get too far down this path, but you know, it gets into that world as I think we touched on when we brought this subject up of, you know, AI is really good, but it's also starts to kind of go, well, hold on. It's listening to an awful lot of stuff I'm doing. It's kind of, you know,
what is that boundary, you know? So, I don't know what are your thoughts on, you know, where AI kind of steps over that that mark. I mean, I'm not saying we I'm not suggesting it does in this product. I'm just kind of getting your views on is there a point where AI kind of needs to back off a bit.
Well, I mean, it's a common question and as we were talking to, you know, as we always continue the conversation with customers about uh what they'd like to see and and what their concerns are, I mean, that that comes up for sure. Um there there's two points I'd like to make about that. I mean, one is Microsoft as a company is dedicated to um e ethical AI and and really setting the standard for ethical AI and that's part of this guidelines for human AI interaction, right? They're part of that. I know you heard that ether uh effort um responsible AI. It's part of that. Um but but the other part is really about the Microsoft 365 hive story, right? And and a and bringing capabilities to customers for security and compliance uh that will help them understand what's in their you know what their c what what their company is uploading to uh what they're working on and how they're sharing it and if they're sharing it safely and allowing people the right tools to share it safely. So I'll give you one example. This one's from the other product I I work on SharePoint syntax uh but we recently or we're adding support um in June for uh sensitivity labels and and Daniel, you're probably familiar with this because uh you see it in Word documents and on SharePoint sites and in Outlook email messages. Microsoft trains us on how to correctly apply sensitivity labels. It's it's common in some other organizations like I used to work for the US Army. They have their standard sensitivity label labels that most people are familiar with like classified, uns classified, secret, top secret,
right? We use different ones here like um they used to be called MBI and LBI, but now they're um
yeah, something else, right?
But uh any rate, what syntax will do is it can say you can train a model to say, okay, this is a contract and contracts generally have uh potentially have uh sensitive information about customers that we don't want to share broadly, right? This should to a limited audience. And so I will automatically apply a sensitivity label that makes sure that this document is treated in a secure and compliant way so that this customer in sensit customer information doesn't get out.
Yeah, it's a really important point actually and I'm glad you brought up syntax because we sort of talked a bit about Viva but what syntax does and particularly in that example because we know that that's always been a challenge. I work a lot fair fair bit in public in public sector government that Tom. So you know the the issue of data classification, data assurance, data risk, data management, you can't always leave it in the hands of individuals because, you know, we people tend to either forget or don't recogn realize the context of something they're doing or just, you know, kind of miss some of those markers. Whereas a tool like syntax is going to help our customers or help anyone kind of manage the constraints they need to think about around say, you know, privacy, security, content, context, and all of that. So it's it's really interesting to see and I've seen it in action and and it's, you know, it's pretty accurate like it recognizes certain things. It can see that you're typing certain types of words or certain phrase phrasing or certain things and kind of goes, you know, recognize that there's you're right on the edges of this sounds like this is classified data or this sounds like this is something that might be sensitive. Yeah.
And and and those things are actually important, right? So like when we looking at say education or in a corporate world, you know, information loss, you know, you can get big fines now and we know, you know, I I work in the educ sector
and you know there's There's there literally is data in there that's life or death and I know we get that in defense and things as well but you know there's very um important data on kids well-being um in there for example so you know there's there's a lot of stuff that that that is that is around and I think taking that away a lot of the breaches when I look at the or you know the notifier breaches in in Australia for example which they release every quarter you see a lot of that being human error you know because those tools haven't trapped them before you know, caught that they were sending that credit card detail or that personal information to somebody else via email. So, you know, being able to catch that early for people is very very helpful. Right.
Well, and the thing to bear in mind, sorry Tom, we're taking over your I'll let you talk in a second. Um, but it but it we have that thing here in Australia D which you know of course which is the duty of care that that is incumbent on uh Department of Education, schools and and education providers
for their students. And so when you've got tools that help you,
I I tell you the first thing that drew me to your uh podcast was you were talking about Imagine Cup. Uh I don't know how long ago this was. It was probably a while ago. And as I listened to the episode, I was thinking,
"Oh, this is not the Imagine Cup I was thinking of." Right? And I had to actually go search and look for I was like, "Oh, this is there's like an Australian edition or or something, right?" And and um and And that kind of eyeopening experience is what I what I was looking for because you know we don't want to build products that only represent like one culture or one perspective. It's a challenge right we have a worldwide audience and we need to uh we can't understand all the cultures right um but we do need to keep our eyes open ears open and learn from different perspectives so it's interesting to hear you know just as you're talking through that right now duty of care like I think that has a very specific meaning in in your uh context that I don't understand but it's it's really interesting to hear that conversation.
So so you know you mentioned earlier on uh Tom about the human AI interaction guidelines that you're working through. Can you enlighten us a little bit about that that work as well?
Yeah, absolutely. And I'll tell you uh just um in general like my work as a content designer and and anyone who's writing content for Microsoft like we're this is a pretty common thing for us to see these types of guidelines and apply them. Right? So, the the biggest example is the Microsoft writing style guide. I'm not sure if you're familiar with that, but um
yeah, it's you know, it's a key uh to all our work. I've got I'm looking at my desk. I've got the Chicago Manual of Style, the you know, the giant reference um on my desk and I need to look up things like um what was it? An illision. The other day I was looking up uh which is an omission of a of a concept anyway but there's all this uh so I just see the guidelines for human AI interaction is just yet another one of those things that I can use uh as I'm going through and uh making some of these initial decisions be you know uh that we we're we may get customer feedback that something's unclear but then we can look at these guidelines and we can apply them and say, "Oh, you know what? I didn't follow this one where uh I'm I'm using I'm presenting the information in the right context." Basically, right? That's one of those ones that we're talking about. And it's like, "Oh, okay. Of course, the customer doesn't understand what's going on here, right? Because I did this." And then I can go in and I can see their examples like this is how we did it in PowerPoint designer, right? And or this is how we did it in editor for Word and It's really helpful
and if I'm not mistaken, Tom, correct me if I'm wrong, the the the guidelines for human AI interaction. Uh it's sort of born out of our work we do in society and ethics under the ether work. So it's kind of all tied into this. It's not just about, you know, kind of building good UX experiences and the right tools for customers. It's actually doing it in the in an inclusive way, in a diverse way, in an ethical way. Is that like kind of that's also got to feed into your thinking models? Yeah.
Well, and and some might say those two uh goals are maybe one and the same. Um that uh when basically, you know, people people know when you're doing things uh that don't make sense, right? You get this you get this feeling, right? Like it's like the clippy thing, the clippy thing. It wasn't right. You get that feeling, right? And that's not good software, right? So, um you a lot of times some of it, you know, these ethical decisions, they're also they're just what people want and need to see out of our software at the same time, right? Is uh to do it in a way um like basically the the system has to be ethical because I'm ethical and moral and I don't want to deal it's hard for me to deal with people that don't follow the morals and ethics I do. Like I have to I have to figure that out, you know, and I don't want to have to figure that out the computer. I don't want to have to question if it has questionable morals or ethics, right? I want it to have similar uh principles uh to me.
So when you're talking when and when you're talking you know I I I reminded me then Lee when you were talking about the inclusive stuff I remember we we released I think it was three years four years ago even the inclusive design principles document which I sent out to schools and things as well because it was about what Microsoft were doing around inclusive design at that particular point just generally uh and I sent that out to schools just so that they could think about the content they were creating online as well cuz it was generally useful. But I suppose my question would be that um you know with with the interoperability of of say Microsoft platform being a platform where people can build upon then do like how how do we then engage other people to follow our human AI interaction guidelines to to make sure that you know if they're using an addin for SharePoint or Teams that that also is is in the same fundamental guidelines. I know we can't force it on anybody but but I suppose that's that's the purpose, right? To get everybody to try to develop inclusively.
Absolutely. Yeah. You know, I think u wow, there's so many um good um topics here, but I I'll give you one example. I can't think of one yet where I know the guidelines are are shared publicly. Anybody can uh go go read them. We're they're continuously developing. One of the coolest things there's a demo now right now where like I've got a deck of cards on my desk where I can turn them over. I can read all the guidelines and I can look at the examples, but now they've got a virtual experience too where you can go and pull the card out and turn it over.
Yeah.
Um, one thing I have seen our customers apply and and I've gotten feedback on is in the writing style guide, we have a section on um avoiding bias uh biased language.
Yeah.
And um and I've done some work in the US we're having um something we call June coming up uh this month.
That's right.
And uh last year uh they they g they told us to take the day uh to spend learning about diversity and inclusion. I decided um I wanted to do a little project on the day and I um wanted to look at how I could empower the community to remove bias uh in our own product language and I was inspired by a friend of mine who co-authored shared a book on SharePoint with me years ago. Uh he had submitted I think it was 10 poll requests. Now I don't know Lee and Dan if you know about this or your audience knows about this but all of all our content on Microsoft Docs right now well most of it um is on GitHub and if you want to make a modification like you see something wrong you can just go in and submit a pull request and you don't actually have to know GitHub or what all that you can just hit the little edit button, type in your thing, and then at the bottom it will say submit a pull request. So he submitted 10 pull requests and he referred to our guide, right, a guidelines in the writing style guide and he said, you know, this uh this article uses the word slave um
and our guidelines say uh to avoid using that and in this case we can avoid using it because we're referring to some UI in the product that uh that has the word slave. So, we have to and so we added these notes in a bunch of places. And what I thought was the best thing here of like empowering our community and like helping them see the guidelines that we want, right? Like
and uh
he actually there were multiple product changes that happened after that and in in quick order actually uh where they said, "Oh, We're using that word in the product.
Yeah.
Or we're not. Actually, in a couple cases, they were like, "No, we're not. This documentation's out of date. We can remove the word slave because we removed that, you know, a month or a year ago." Yeah.
Yeah.
That's um that's a it's a it's a great story and I think it's really good to shine a light on the fact that that capability is there to to make those kinds of changes and edits, you know, whether or not it's Juneth, but just the you know, knowing that that that that kind of community engagement is there. Um the the deck of cards you were talking about, by the way, Tom, I I have the same one. It's judgment call is the game um that we that the society and ethics team, but we should put a link to it in the notes down because you can actually download it and and and use it. And it is a great
might be a new one actually. That might be a different deck. I'm thinking I like the idea.
I'm looking around. I think I had the deck here, but I can't find it. Anyway, but we've got a couple of those resources that you can actually use to ah yeah, I've got that one as well. Yeah. Um uh so yeah, there's a couple of versions of that and and these are things that put out there and I think to Dan's question about how do we kind of get more people to know about it is just about it's like any any message you want to get out there you've got to take it out there you've got to give them give people mechanisms by which they can adapt and use it at their own pace in their own world at their own in their own uh kind of uh mechanisms um and then and then the most important thing is to the GitHub point is provide them mechanisms to make sure that change can be impact change can be affected because it's you you can't have all these desire for change that then have no mechanis for change to take place. And I think that's that's when we think about those AI human AI interaction guidelines, the fact that we don't just post them out there. We don't make them available. We also make you able and empower you as an individual, Tom of a company, but anyone to to hold us accountable to that. Um, and I think those are the hallmarks of good kind of responsible application of these things because we're never always going to get it right, but if there's a mechanism by which you can demonstrate where you're wrong and change it, then you I think you're kind of part of the way there. So, it's really good. Thanks for bringing that up, Tom.
Well, it seems like we hit on a theme here alto together about feedback and and loops and uh and listening uh that I think is really important and I think it does capture a lot of what's going on uh at Microsoft these days with you know we talk about customer empathy and um customer obsession and uh I I think it's no surprise right that these things are going on throughout the company basically.
Yeah.
Yeah, definitely. Yeah.
And and and you know this is fascinating. We we could we could talk to you for hours, Tom. It's really really brilliant and thanks for coming on to the podcast. But I'd like to just uh just touch on also the fact that you know we we moving very quickly with AI and uh you know you talked about cognitive services earlier on and you know every time we do a build we had something else you know everything is moving so fast. Um a I suppose how do you keep up with these technologies that are moving and then also what excites you most? I suppose in the future in next year or so with the products that you're working on, you know, which connect in with AI.
Yeah. Yeah. I mean it's uh a challenge but um I I love it. I think this is what draw draws me to technology is not just um you know learning about how the new tools work, but I think what my kind of superpower is that I've kind of focused on as I've grown in my career is trying to help people understand it because it really that feels really good to me. Right? When I explain something to someone and they say, "Ah, oh, that's what I can do with that." Right? Um that it really that'll make my day, right? I think, "Oh, I've helped somebody figure out something. Um I understood, but I wasn't sure if I could explain it, and I did. And now I can probably explain it to somebody else. I hope." Um yeah, there's so much. I what from this last build I was impressed. So there's a lot of things coming in in Viva topics and uh SharePoint syntax I could talk about. Um but I'll just because you mentioned build the one I was really excited about was that we finally got the GPT3 into uh one of our information worker um uh platforms. You can use it in the power platform now to build expressions. And that's uh an audience that I'm really focused on we call it uh modern work solution enablers is the group that I'm in um because I feel like there's so much people can do it's kind of like uh some people use Excel to do amazing things uh right but we've always had this uh in in Microsoft through Microsoft access or now the power platform um but just this idea that we can apply that large uh model right GPT3 that is amazing. I don't know if you've played there's another game uh you may have talked about. I can't I I'm not sure using GPT3 that it's it's a worldbuing game. I I tried it out. Um
no, I haven't seen that one. No.
So, this this is a language model that um it's so advanced that uh you can have conversations with it and you can build these worlds out and it'll just continue to modify the world. as you're talking to it. So, it's kind of like playing Dungeons and Dragons with a dungeon the AI dungeon master kind of thing.
That's That's awesome. Yeah. No, that I mean, look, we talk a bit about the the GP3, the the generative pre-trained transformer model, the 175 billion parameter model, the biggest model ever been built. But it's a really good that you're talking like to see that come out in context, actually see it getting used
in tools and services that people can actually start to kind of see and feel. That is a really that's a great one to pull out. from the from from build. Uh Tom, thank you for that. So, look, we're pretty much at time, Tom. Um we've been talking for a while. So, um thankful for you to come on board. Uh next time you won't comment on a podcast so quickly because it ends up putting you on the podcast. But but no, your stories are fascinating. Yeah. Really good examples of of of kind of where how we're thinking about these things and where we should be thinking. And look, the Minecraft album, uh your niece and your daughter, you know, we've got our next episode lined up. Um you better Thank you, Tom, for everything you you've shared with us today. Really appreciate it.
Thank you, Lee and Dan. I really enjoyed it. I I will not hesitate to comment on your future podcast episodes because I love it and you've been uh very gracious uh hosts and good good response to my comments. Thank you.
Thank you, Tom.
Good, Tom. Take care. Thank you.