Sep 9, 2021
Today Lee and Dan talk about GAN'S, Games and other goodies from using AI to upscale games and 80s TV shows to GitHub Autopilot.
Links:
AI breakthrough could spark medical revolution - BBC News
This YouTube channel is using AI to 8K-ify classic game intros and cutscenes - The Verge
Think, fight, feel: how video game artificial intelligence is evolving | Games | The Guardian
Microsoft-powered autonomous beach-cleaning robot is here to clean our shores - Roadshow (cnet.com)
Collections - JenLooper-2911 | Microsoft Docs
________________________________________
TRANSCRIPT For this episode of The AI in Education Podcast
Series: 4
Episode: 10
This transcript was auto-generated. If you spot any important errors, do feel free to email the podcast hosts for corrections.
Hi Lee, welcome to the AI podcast. How are you?
I am well Dan. Lockdown in Sydney like your good self I'm sure as
well. But uh despite the lockdowns doing fine and keeping myself
busy at home.
Ah yeah no absolutely and uh you know I've got I've got kids on
teams meetings around the house. this morning and you know it's uh
it seems to be different this year though or this time uh I think
teachers like previously it was more of like a World War II spirit
and kind of crack on and you know keep keep calm and carry on
whereas now it's like uh proper learning coming through. So you
know I'm I've designed a jewelry box this week. I've uh
uh wow uh yeah I know there's been a heap of stuff there's been
teams meetings on about Egyptian uh kind of pharaohs which is quite
interesting. So yeah I'm running between meetings and one minute
I'm talking about Azure signups and the next minute I'm doing a
jewelry box. So so hopefully I'm I'm talking about the right
content to the right audience.
All all learning is good learning. I um I mean it's it's kind of
funny. I was tweeting my kids um about this because my kids are
obviously working from home as well. Yeah. Uh we our school uses
Google the Google platform so use Google Meet a lot and we did you
know we've done all the classes set them up. Um but I said to them
you know I think they said you know we have to work work from home
like you dad and I said yeah well you know I think about it. I've
now been working from home since March 2020. So, we're now it 16 17
months of constant. So, when you just said then, you know, the
second time around, it's I mean, I know for a lot of people it's in
and out of these lockdowns, but it's kind of become the the new
normal.
Um, but my kids are loving it. I don't know how your kids are
coping. My kids are actually finding the working from home, apart
from missing their friends, they actually find the the day
structure to be um pretty good. So, yeah, so far so good.
It it is. But I've read a couple of art though and I suppose it
does permeate into the AI space because we've we've talked a lot
about you know some of the like in the last episode or so we talked
about the way AI can be used for mental health and things like that
and you know it was it was starting to be used in the first element
of the pandemic and now
you know it's coming to the four quite a lot I think rather than
being a sort of fringe technology people are I'm starting to hear
more and more people talk about the way they put it in structure in
the daily commute, the way they're exercising, the way they're
doing mindfulness, the way AI is supporting them and looking at
what they're doing. And you know, there was a article um I'll have
to dig it up. I can't I I haven't prepped it for the for this
particular episode today, but there was an article talking about
what kids are missing out on because if I think about my son last
year, he missed out on his entire year six experience almost
because, you know, normally there's a lot of ramping up, there's a
lot of things they do, there's the prom, there's um you know, all
these things they do as as year sixes to end the school year and
graduate almost and and my daughter now has gone through almost two
years in year last year two and year three where she spent maybe at
least 60% of her time at home. Um so there's a lot of connection
being lost and it's just talking about you know and and some of
those year 12 students you know it's not just about the content
it's about the experiences and and students you know um uh my
partner's daughters are in uni's doing midwiffrey one's doing um um
uh English and they've gone through two years of courses being held
on Teams and Zoom. So, you know, there's all of the the internships
around that that people miss out on. There's all those experiences
and on-site things like uh groups, community groups, uh university
newsletters, university interest groups that people don't connect
with. So, people are getting through content, but they're not
getting through social community.
Yeah, it's a it's a really good point. I mean, I My kids are a bit
older than yours. I'm in year 10 and year five. Um so they're not
quite at that point yet where it's the you know those big
transition phases like year six or of course those big important
years like 11 and 12 and all years are important but those you know
the points where you kind of finishing up school was sort of in the
middle. But you're right that kids are it's not they're getting the
education and I'm see my kids are kind of still getting the
learning done and as I said quite enjoying the pace because they
can learn at their own pace I guess but they're missing the
experience is so what it'll be interesting to see what the long
term impact is on kids that spent the last you know those formative
years of school not having the experiences of the camaraderie but
the connection of being in that community how that changes their
impact you know kind of change and we've talked a bit before about
the imp digitization of the youth today and how our kids are all
growing up in a world that's so different to what we did and
they're using technology all the time and they live in devices and
all that kind of stuff that you know we think that's going to have
this big impact because they just connect differently to
individuals you know by son's all of my son's friends who he's very
close with are online that he plays with you know through games and
through discord and conversations
whereas we went out to the playing fields to play with our friends
because we're that old
so you know just with that and then now with the intensity of co
creating these kind of lockdown moments it'll be an interesting
long experiment to see what that what that impact is
yeah and and and you know don't want to riff on this for ages but
like the the a use of AI in things like some of the parental
control areas because you know there's it's similar to the security
things that I see technically in in in businesses where they use
multiple tools to manage their security right so they got you know
they got their firewall software they've got their defender or
endpoint software they've got they all and it creates gaps and
that's what I'm seeing as a parent now because I'm thinking
well
what is Megan doing on her iPad on Roblox what telemetry am I
getting back as a parent from that well it's minimal I've got some
parental controls in the Apple ecosystem then when they're on Xbox,
you know, I've got the Microsoft consumer credentials on Xbox. So,
I'm managing that. Okay. Um, because I know what games they're on,
how long they're on for, and that's also connected to the consumer
account in Windows, so I know what browsers they're on and what
they're doing generally and the AI is feeding me back some of that
data. It's not necessarily AI, but it's just giving me a good data.
That's
intelligent. It's intelligent data. Yeah. It's trying to make me
out to be the AI to analyze that data, you know, but but Then
you've got the the other aspects where people are starting to
really try to police kids mobile phone and data for older kids
possibly, you know, like some of the Optus apps you can get now and
the Vodafone apps that can manage data plans and turn kids data off
and mobile phone data off so that you can focus them in. So there's
like there's these gaps appearing where you've got to keep looking
for different tools, you know, who's managing the iPad, who's
managing the Xbox, who's managing the
Yeah, it's it's I think uh I look And it's it's an interesting one.
And I think maybe to put a cap on that and get us moving forward, I
think I think we should do a a session a podcast on uh parental
management and parental controls and technology because I think
we'd I think we'd explore it a lot of interesting ideas and
thoughts around that. So, let's put let's put that one in the in
the show notes for the future.
Yeah, definitely. So, what's been going on in your world then, Lee?
Today's podcast, we're going to be looking at some of those some of
the the the things that are happening in the news at the minute,
and there's been a lot of big announcements around our our main um
Ignite conferences and so on, inspire and things like that over the
last six months. So, um we've thought about bringing some of those
to the four. What's been what's been happening in the news from
your side?
Uh yeah, look, that's really good. Yeah, and I think yeah, there's
a lot of stuff out of Inspire obviously that was last week I think
now is as of recording today on the 27th I think we are. Um
so look, yeah, obviously I mean inspire and again one of those
virtual events and generating a lot of news, but as you know Dan,
we've talked about this before. My world is um largely looking at
where Microsoft's technology has a a social or a national impact
and I look at kind of the big big markers around technology like AI
and quantum and others. So AI has been top of mind right now and
and I think we talked last week uh last time on the podcast about
copilot and the outcome of the the GitHub copilot solution which is
built on that uh uh the pre-trained GPT model that we've bu been
building with open AI for some time.
Yeah.
But look let's rather than talk about some of the specifics and
there's a lot of interesting stuff going on. Something that I've
noticed more more often than not now in the AI world is we're
seeing a lot more what I would call that democratized access. So
basically very complex rich AI tools being productized and then
made available to people or individuals or you know anyone really
kind of enthusiasts and modders
to to take those tools and do interesting things with them. And of
course the challenge you have with that Dan is when you democratize
a tool like AI, you know, and we can, you know, AI is a very broad
term, but let's say something like, um, the one that's really
popping up a lot at the moment is GANs. Um, your general generative
adversarial networks, and we'll talk we'll talk a bit about
that.
The more and more those kinds of models, uh, it's a machine
teaching model become available, more and more people can use them.
And there's always the use for good and use for bad. So, I'll talk
a bit more about it, but what, you know, what, you know, do you see
this? Are you finding that, you know, you're finding AI tools are
just kind of more open and available and almost in everything these
days.
Yeah. Yeah. Absolutely. And and what's what's peing my interest is
I'm seeing lots, you know, I think when I was doing uh talks to
students and to educators and to general, you know, the the
technology community, I I'd use like
possibly one example every three or four months. You know, it was
snow leopards. I know I know one of my colleagues used to like uh
make a joke, oh, Dan's talking about snow leopards again. But now
it's turtles, it's fish. It's birds. It the the the use cases of
people using it uh and using these technologies are becoming you
know huge. Yeah. I'm seeing so many examples now and it's it's
almost like which examples do I pick? The police force is another
one. You know the medicine medical area. It was just you know I did
a talk last week and it just kept going on and I think which which
one do I actually pick here because there's so many good examples
now
which is good. I mean it's great to see AI sort of finding its way
in good ways into so parts of society. But there is that negative
side. I you just thinking we talk about generative adversarial
networks and we kind of say that phrase. It's probably best to to
explain how that works because it's an interesting you know people
we talk about AI models and we are most people kind of get the idea
that AI is about kind of behaving like a human decision process.
You know we're trying to emulate some of those human cognitive
services.
But when you think about it a a g because the way I was like I went
learning about this and thinking about GANs because GANs get used a
lot in um things like uh upscaling or or image improvements or kind
of rapidly iterating on something to to create either an enhanced
version or an artificial version something. So if you the way a GAN
works in simple terms if you didn't already know Dan is you kind of
got this idea of well you let's say you have a a piece of data a
real piece of data like a picture for example a picture of the Mona
Lisa it's a good example picture of the Mona Lisa what you then do
when you build a gen a generative versal network again is you've
got one side of it which is and this is the adversarial piece of
it. You got that generates things. So basically it looks at the
real data and creates lots and lots of fake versions of that data.
You know just slightly modifying the pixel colors, the density, the
layout and the kind of and you end up with all these lots of fake
versions of the Mona Lisa. And then you so that's the generator
side and then the other side of it is the discriminator which is
this other tool that then looks at the images that the generator
creates and goes ah that's fake. Uh oh that looks kind of real. Oh
no, that one looks like the real thing. And what's happening is
every time the discriminator looks at it and says, "I see that's
fake." The generator goes, "Okay, ignore that one. It's not good
enough to fool the system. I'll create another one." And it creates
another one. And the discriminator goes, "Yeah, that looks more
real." And what you're doing is this really rapid teaching process
to get to a point where you've created something that is looks like
the real thing. So, you know,
great idea when you think about So, some of the things I've seen in
the press this last couple of weeks, um, people are using this
technology. There's a company out there called Topaz. Uh, Topaz
will build a product called Gigapixel A I and it does it's a it's a
GAN based service that lets you upscale images. Now I'm a bit of a
gaming nerd as I know you are too. Um and a and a science space
nerd. So there's uh so NASA's using this technology to upscale
images of the sun that is taken through space telescopes.
I know it's incredible.
And the same technology is then being used by a bunch of YouTubers
to upscale gaming uh assets and gaming videos and gaming images
from games from the '9s and stuff. So you
I saw those phenomenal. It's absolutely real. Yeah.
Well, I was actually this morning I was watching one just to kind
of get myself on it. And you know which one I chose? The Masters of
the Universe He-Man trailer from the 1980s.
It's so good.
It's so good.
It's perfect. It's pixel perfect. But it's the same technology that
of course we, you know, that it works in the deep fakes and starts
to create, you know, you know, versions of Tom Cruz, you know,
doing things that he shouldn't be or or Donald Trump. So, you can
see where you get this challenge that something like adversarial
networks can create these amazing experiences for humans to see old
content or or old things in new ways.
But it creates this challenge of well if you how do you make
something available like that but it only gets used for good and of
course you you can't you know you put a technology out there
somebody somewhere is going to find a nefarious use for it.
Yes.
So now you have this problem of you need laws and regulation to
build the what we call the guardrails. And this is this is kind of
what I'm you know to get back to your original question. What am I
seeing going on in the AI world right now? This is the big topic at
the moment is this challenge of
building the guard rails. So AI needs regulation. No question about
it. AI can't just be allowed to run free because it is both a very
powerful tool and an incredibly dangerous weapon.
So guardrails create that kind of safety barrier. Yeah.
But in order to build those guard rails, people who build the
guardrails need to understand the technology. They need to
understand where those risks are. They need to understand how how a
deep fake is made, but also how you know medical imagery can be
improved for better diagnosis. and understand the mechanics of
that. So it's simple process but deep context and then as a
society
we all need to define what those boundaries are. We can't just let
lawmakers and governments say well that's now allowed and that's
not allowed. Society needs to shape that
and in order to do that we need to then empower our lawmakers to
make those laws which then again requires us as society to
understand that tech to value that technology in our lives and to
trust that the system is going to work for our good. And so you see
this is the conversations that are going on right now in AI
circles. We know we need guardrails, but who defines them and how
does society have a voice in that? And do we trust it? And do we
all have it the empower do we all feel empowered to be a part of
that conversation? So, it's a big topic, but that's kind of what's
happening right now. I don't know. What are your thoughts on on
that?
That's huge. What are my thoughts on that? I'm like, yeah, it's it
yeah, the the you know, we've we've interviewed and spoken to
people and we know how important the ethics and the and the the
guard rails and that that governance are and and you know I'm in
awe of people like yourself and the the folks that are looking at
this at the uh in Australia at the regional and national and
international level because um it's absolutely where we should be
it what it's what differentiates good use of technology to bad use
of technology and you know if we if we disregard it all which you
know I do worry about that polarization of society at the minute
you know we being polarized with vaccines we've been polarized
things like controversial things like Brexit and Donald Trump and
blah blah blah. You know, things like AI are very easy to polarize
and just ignore and then we miss things out. You know, I was
looking um uh this week at the proteinbased scientific
breakthrough. I think that was through Google um but there was
there was an AI um uh model built to predict the structure of all
proteins in the human body which is massive.
Yeah, that's that's phenomenal. Um you know and and that that's
really available to like supercharge some of the discoveries of new
drugs, new uh you know, it's it understanding that jigsaw puzzle.
It's almost like cracking well it is like cracking the genome. You
know, they've just managed to understand how all these proteins
kind of connect together and and that that's that's that's
absolutely phenomenal. Um
uh and and the AI has to be used to do that. You know, similarly,
you know, we did the genome manually to a certain extent and now
things like proteins which are almost impossible to do are being
cracked and Without AI, if we discard or disregard certain aspects
of technology, then you know we've got big benefits. You know, you
got to jump in that curve.
Look, and this is the thing, isn't it? We don't want to sacrifice
the value we could get in the future from the things we're building
now, the big scale models, the huge data sets, the increased
complexity of machine learning and AI tools, but at the same time,
the democratization of it. We've got to make that happen. But, you
know, at the same time, we've got to ensure that you know that
those that seek to do harm through those things are not are not
free to do it and you'll never stop bad things happening. I mean
you know it's kind of the eight old adage you never you can never
stop criminals being there but you can build
societal norms that make criminality wrong and in the same way we
can build societal norms around AI that recognize that the way some
ways of using AI are bad and some ways are good and it's you know
and you got to look this in the context of of the of your human
journey on this I mean we are just
so early on if we think that you know if we generally accept that
AI was you know was as a concept was formulated in the in the 50s
and really
built around this the 80s and then the 90s and we went through
those two winters of AI and we're kind of really only now just
coming to the point where we're actually building something that is
actually promise delivering on the promise of what AI was
always
capable of doing. Yeah.
So it's almost like it is it's like you know it's a baby it's a
tiny baby which And it's and it's the ideas, right? And I think
that's one of the things and and you know, I know I've been biased
here from a Microsoft point of view, but it's about creating the
platforms for other people to develop on, you know, and then
sometimes those things are mind-blowing and sometimes you don't
even know that they developed on our platform or whatever other
platform it might be. But then there's still those those odd ones
though. The one that jumped out to me is the is the beach uh the
cigarette but beach cleaning robot. And there's there's always one
thing that you sit down and as as a member of the public, you sit
down and think sort of good idea but somebody can go around and do
that as well. It's kind of like so some of these things you know
you you see some of the technology like and when when when the
technology is deployed say to with the turtles for example when
it's looking at feral pigs um and and managing that the technology
sounds romantic and then when it's picking up cigarette butts using
a drone on a beach you're looking think ah similar technology but I
don't know if that second one is really like worthwhile developing
and you know it's fantastic the the the approach I think it's the
team of tech ticks they're called I think it's uh two two guys that
kind of set that up Martin Lucart and Edwin Boss uh like Dutch
engineers who set that company up and they've got this um uh
startup going around which is fantastic and it's picking up these
cigarette butts and things on the beach and and you know it's a
good idea um but then you know it's kind of exploring well where
can I be extended is that a use case yeah it's there's kind of like
quirky things that happen which are which are serious but you know
as people in the public think there's a is it easier to go around
pick up
yeah I mean yeah it kind of plays that is is this the best use of
technology but I think what it kind of highlights to me Dan is this
it kind of gets back to the very core of it that we have this tool
that we don't really know yet what it can do you know AI or machine
learning I mean let's I think we we again we've had conversations
this keeping those two separate but
the ability for machines to sift through data and create and
identify and then predict patterns that we can't see. And then the
application of that learned model, inferred learned model into an
AI system that can look at the world and perceive the world like we
can and do things that, you know, like picking up cigarette butts
on a on a beach. It's a step on that journey towards teaching
machines to identify things in our world that actually do need a
solution to it. You know, if you scale it out, cigarette butts on a
beach to plastic on a beach to plastic in the oceans to debris in
space. You know, think if you can kind of draw a bow between, you
know, an automated vehicle on the beach doing that to one day
launching an automated system that can go up into space, identify
space debris versus actual satellites and clean up that. That's the
promise of AI, but that journey is hard. Most people aren't going
to see that journey and go, "Okay, that's where we start." But if
we don't have like if we're not allowing the innovation to happen
today, so if we don't give those two Dutch engineers and the
Microsoft people they work with the license to go crazy. What a
great idea. Build it. It's not economically viable. It doesn't make
a lot of sense in the short term.
But if we don't learn that bit, how do we get?
So I get true.
I get your point. But it is it's
it's a learning process and that's I think one of the challenges
that AI has today in society is that you know your every person on
the street might see some application of AI and go well that's a
waste of you know taxpayers money. You know what's the value of
that? But you got to look at in the steps of where it's taking us
towards and how we're getting better at understanding how to apply
AI which and you know keeping on that tone of learning how to do AI
better um something else that I'd seen before because I'm
privileged in my job to get around the teams inside of Microsoft
that work on society and ethics for AI but something they've just
released uh I think it was last week or the week before is
something that I've had had my hands on in the past but it's called
the hacks toolkit so it's essentially the human AI experience
toolkit and what it is It's a set of cards, but it plays like a bit
like cards against humanity kind of cards. It gives you these, it's
designed for for engineers, for UX user experience, for software
developers to have these cards and to play through the things that
they might not think about. So, if you think about AI in the
context of uh you know, I live in my world, I do AI like this or I
I do software development like this.
Yeah.
This is about giving them the opportunity to ask those questions
and we'll put a put a link in the show notes, but
it asks you those questions around and the application of AI. So
that you might look at it and go okay so what do I need to think
about before I build this what do I need to understand about the
interaction of humans in the system and then once I've built it how
will it evolve and humans interact with it and what are the things
I need to think about
yeah
it's really smart thinking about not just the technology but the
human aspect of that technology
you mentioned the GitHub I saw that in action on on a video and
then suddenly that was a bit of an epiphany moment to me can you
expl This is very bizarre. Can you explain like in a couple of
sentences to the podcast listeners what that cool pilot authoring
is because it's pretty amazing.
It is. It's pretty amazing and and I think we can probably we
should talk a bit about the you know the the the challenges of of
the way in which it does what it does. But look, Copilot is a um
it's a a trained model built on the the constructs of the GPT model
which is a it's a it's a language prediction. engine. So
essentially what we we we show it thousands and thousands millions
of examples of of language and in this case language that was
derived from GitHub code submissions. So it's built on open source.
So we're not stealing anyone's data.
GPT just acronym busting generative pre-trained transformation
algorithms.
Correct. Yes. It's a nice fancy word for a a type of model. But the
beauty of the GPT transformer is its size and the amount of data
that can operate on. So I think the latest version GPT3 175 billion
parameters. So if you think about a parameter in a data set being a
point where the model was able to make a decision about what it saw
and then learn something. So it's 175 billion neural snaps in your
brain going decision decision smarter smarter. Anyway, the upshot
of this is co-pilot in GitHub will watch what you type. So you know
as you expect the the the tool is watching what you type both your
comments and your code. So what's really smart about it is you can
write you could write a bit of code that says in comments um the
next bit of co you know this next bit of code is a recursive model
to generate random numbers from a pool of data from you know the
pulled from a a data set X for example.
Yes.
And then you write the first code declaring your write first line
of code declaring your function and then co-pilot will see first of
all that you've declared this function. So that's the function
we're going to operate within. And it reads your uh your or notes
and says okay what you want to do is you want to build a automatic
random number generator very simple example and then it will
automatically fill in the code now that filling in of the code is a
combination of have I seen examples of a random number generator
before yes probably millions of them inside of GitHub do I
understand the constructs of the language you're writing in i.e.
you know if I'm writing in Python or something like that and how to
write good structured code to do a random number generator and what
you end up with is a combination of past proven code fragments and
predict predictively build new code that didn't exist before and it
gives you m it gives you multiple versions of so essentially I mean
there's a great little you go look at the videos of it you'll see
somebody typing in the the notes writing the first line of code and
then copilot just fills it in and and as the programmer you can
accept it take one of the other examples or completely remove it
and write your own amazing idea and you know same thing about when
we've seen GPT3 doing things like language prediction so you give
it a a 10 page document and it will give you a one paragraph
summary of that by understanding not just words but context and
association and semantics and the language structure. So it's very
clever.
Yeah.
But of course here's the challenge when you look at it. It is
unreal.
It is. But then what you've got is an issue and this is gets get
back to that issue of sort of ethics and uh kind of the right thing
to do question is if it gives you a fragment of code from somewhere
else Are you stealing the code? Did you is the code under license
that allows you to use it? What are the terms of that license?
Because the code is now presented to you outside of the constructs
of its licensing model. Now, you know, with that licensing model
could be any number of different open- source license models.
And if it builds new code,
then
whose responsible and whose IP is it? And so, but I think again,
it's one of those learning points a bit like uh if you remember Tay
the Taybot we did in and and and multiple other examples of not
just Microsoft and other vendors where we build something,
it doesn't always behave exactly as it predicted. That's the beauty
of AI. It doesn't behave as predicted.
And instead of going, oh my goodness, the sky is falling. It's
terrible. Shut it down. We it can't work like that. Let's never do
that again. We have to kind of look at it and go, okay, so it's not
perfect, but it's learning and we can learn from it and it can
develop better code. And from a from a coder's point of view,
I can write faster code. I'm not actually having to write the
code.
Exactly.
In theory, over time, my code should get better because it's going
to learn from my
code as well. Yeah.
More efficient coding. And
that means as a coder, I'm less spending less time writing code and
more time working out what that code needs to be like and working
with a business and we get a better connection between code
development and operations or in fact something like better
streamlined DevOps model. So, it's an interesting
Yeah, it is for all. Thanks for explaining that because I did you
know that that is that is phenomenal. I I I can't describe it
anymore. It's very difficult on a podcast, but when I saw somebody
just putting in a couple of comments, you know, in starting their
code, putting a couple of comments in saying what it does, what
language they're using, and then it's developing thousands of lines
of code, uh it's it's so good. Um uh but wow, you know, that's that
that that does start to kind of blow blow your mind a little bit.
So, anything else then from the the book of news for from Inspire.
Anything else to update the the listeners on? You know, there was
lots on general technologies around things like Windows 365 coming
out and and and some of the general Microsoft technologies, not
really connecting into to AI or anything that we we talking about
on this podcast. Viva, I remember that was mentioned as well and
that was great because uh uh we had a chat about that with one of
the engineers a couple of uh episodes ago which was fantastic to
see that through there.
We did. Yes. So, no, you're right. We did mention we were going to
talk about Inspire. We probably been talking for a long time and
not even mention it. So, look, you're right. There's not a lot of
AI that came out of it. The only thing that stood out for me, there
was actually one thing in particular, lots and lots of cool stuff,
and we should obviously share some links,
but Windows 365. So, yes, obviously big announcement there, the
ability for you to be able to now deliver Windows 10 and 11 um ser
desktops essentially uh streamed apps and services directly from
the cloud to your PC and kind of you know, giving us giving our
customers the opportunity to really experience the best that
service is on on any device, on any hardware in a very managed
environment to maintain security controls and all those things are
really important to a lot of business and enterprise customers.
There is at the heart of that and I we and we should probably go do
some details on learning on this but underneath that actually is
actually quite a lot of AI in our system in the back end about how
we optimize that experience for for the user. So if you think about
this being streaming a desktop,
it's not like the old days where you know streaming a desktop meant
streaming down the entire thing onto their machine and then it ran
locally. We're talking about really streaming now. So making sure
that what we deliver to the desktop is what the customer needs at
that time and then predicting what the customers or what the user
is going to do based on the current experience. You know what
they're doing at the time. So it's a lot more lot more of a
uh intelligent experience. So that's probably the only thing that
stood out to me is that you know maybe maybe people aren't thinking
about it but a lot of AI goes into building the capability to
stream something as rich and complex as
Yeah. One thing that jumped out to me And I suppose then just to
kind of clear the this this section off would be the the
sustainability element. And I know I keep talking about that, but
my conversations over the last three or four years about using
cloud technologies generally and and AI and all all the kind of
things that come with that has usually been about cost and people
always get caught on this onrem manageable cost opex, capex, all
these kind of things. Whereas the conversation now I make sure I
add into my conversations that it's about moving to a more
sustainable approach with your carbon footprint. So the
sustainability calculators and the you know and again not trying to
be biased about it but the Microsoft approach to um ensuring and
working towards that net zero emissions um targets that Microsoft
put in place. You know I think me for me I feel really passionate
about that element and you know I like to know that when people are
using the technology that they're actually um thinking about the
that we are investing in those techn ology so that it's as
environmentally friendly as it can be and going forward to be to be
net zero. So that that you know I'm I'm glad that we keeping that
uh progress going and there was lots of uh discussions about where
we are towards those targets which which is great to see.
Look and it is important uh and it's and it is again a lot of AI
goes into the background of kind of understanding the modeling of
the distribution of and the measurement of some of those carbon uh
carbon challenges that technology provides. and you know to be
non-bias you know yes we're doing a lot of stuff but I know that
there's a lot of great work going on at Google as well around their
carbon commitments and carbon sustainability and it's important
that not just us but all of those absolutely kind of large players
in this industry are on top of this problem and it's great to see
that it is
an industrywide thought process you know we're all thinking about
that and I think for us we're thinking a lot about how do we help
our customers get on top of that sustainability it's not just yes
we talk a lot about what we're doing to ensure our footprint is
reduced and we're doing the right things, but we know that a lot of
our customers are also saying, "Well, that's great, but how do we
do that in our downstream business? How do we use your knowledge?"
So,
and that's an area where, you know, where we want to get into and
we want to teach more more people. Um, but no, super important
thing.
The other there was another thing that cropped up this week wasn't
an inspire thing, but I wanted to share it before we close out this
this show for this week.
Okay?
Because I think it's really a really important one. It's the kind
of thing that I want to see more of.
Um, I we'll share the link, but what we one of the challenges I've
seen when I go out and talk to customers around these kinds of
technologies and I faced it myself as I've sat down and thought you
know what I need to learn a lot more about machine learning I need
to get more hands-on with the technology I need to get programming
again it feels like a really big hill to climb it's not an easy
domain to get into there's lots of techn courses out there to learn
Python and learn coding but machine learning not just about the
technology but also about the constructs the ideas behind it the
methodology behind machine learning is something that's really
important to teach. And so actually what we've got is we have built
a uh a course um one of our um wonderful engineers in the Microsoft
uh business in the Microsoft machine learning business has built
built a course called machine learning for beginners. It was
published just in fact just this this week on the 25th um we'll
send the link out there and it really is a real but it's what's
interesting and this is what caught my eye. It's kind of all the
learning steps how to create models how to take your first steps in
Python and build that using tools like Azure machine learning,
understanding natural language processing, but there's one of the
learning paths which I've got to get to when I get there. Yeah.
Predicting rocket launch delays with machine learning.
Really?
So the minute the minute you talk about rocket launches, I'm like,
yeah, I'm in. I might want to get that.
That's fantastic. So need to share.
Great course. I think we should share that. We should definitely
share that with the listeners because it's uh everyone should learn
a little bit about coding and machine learning. I think it's an
important skill for the future.
100%. Well, thanks for sharing your uh knowledge and your kind of
insights this week. Uh Lee, it's been a great great to kind of chat
again. And um I think in the next episode we're going to try to
speak to um a couple of interesting people. We've got two fantastic
uh ladies lined up that we're going to speak to. One around
Minecraft uh and music and another one around governance. So she's
going to be fantastic. So uh absolutely look forward to speaking to
you then.