A bit of AI - S02 - Ep 8

Hi everyone, and welcome to the eighth episode of season two of the A Bit of AI show

show we talk with Buck about his life in AI. Hi everyone, my name is Henk and I'm a white

male with brown hair wearing glasses and today I'm wearing a grey and black shirt with the words

this the other side expert on it and this is one of our old relics from ignite the tour

when we were still traveling the world i know very good t-shirt that is a blast from the past

you know how i feel about the word expert though hank never always a fan of it but we can talk

more about that and uh hi everyone my name is amy boyd i am a female with blonde wavy hair

And today I'm wearing a black t-shirt and it's got lots of colorful spots on it

So once again, everyone, welcome to the A Bit of AI show

And this show is all about the story from the people behind the AI systems

Because in our job, Amy and me met so many people from around the world

And we learned that there are so many different skills needed to create one of these AI solutions

from pre-sales to consultancy, to creating the deep neural networks itself and running them in production

So in this show, we talk with people from all over the world that are professionals in the AI and data space to have a chat

and just talk about what they actually do during their week. So this is the eighth episode, and we couldn't be more excited

to start talking to our guests and learn more about what they do

As always, all the links and all the information can be found

on our website, abitofai.show. Wonderful. And you know me, always on the logistics

So after the show, if you are new to our show, we have something called the A Bit of AI Cafe

Cafe Experience is anyone who's watching can come and chat with myself and Henk

And what we'll do is we'll share resources. We'll talk more about the things we've spoken about on the show

and feel free to ask us your questions. So if you want to join us, it's straight after this show and go to aka.ms slash a bit of AI dash cafe

Or if you're on our website, you can actually just click a button that says cafe on it at the top of the screen

So perfect. Let's get started and invite Buck to the screen. Hello, everyone. I'm Buck. How are you doing

I'm doing great. I'm a principal data scientist here at Microsoft. I'm an older gentleman wearing glasses and a white Windows Azure shirt, not Microsoft Azure

I was on an early version of the team that created Windows Azure at the time and have been here at Microsoft for about 16 years

And I'm coming to you from my offices here in Tampa, Florida

Amazing. Oh, my goodness. why didn't I not get the memo about like the retro t-shirts

I mean I feel like I should have been wearing one we should have spoke

next time we need to call it Hank and I had a quick chat

about it and we just didn't feel like we wanted to loop you in on it I know it was probably when I was

putting like my thousand lights up to make this happen but there we go

it's all good but thank you Buck so much for joining us and to hear a little bit

about number one, you've been all over the tech industry. So we're really excited to speak to you anyway

just about your perspective, but also the amazing stuff you're doing. I nearly said Windows and Microsoft Azure

at the moment in the data space as well. So can you tell us a little bit more about who you are

and what are you working on right now? Yeah, so I'm working the Azure data team

And when we first started the projects there to have artificial intelligence at Microsoft, a lot of it, we've always had it in Microsoft Research, but a lot of it came through the SQL Server product where we put the R system from Revolution ytics that we had purchased

I had been on the SQL Server team when I first joined Microsoft

And so when we brought that in, they asked me if I'd join the team and work with those folks

And we began that process of embedding machine learning into SQL Server

Now, of course, it's spread out now. And SQL Server is now under the umbrella of Azure data, which includes not only SQL Server, but Synapse and Purview and all kinds of other products that we have

And of course, we work quite well with machine learning in multiple ways

everything from Spark to Azure Machine Learning. Yeah. Oh, no, wonderful. Interestingly, Spark, Python, all these things get brought up on this show every single week

So it's fascinating to see, again, someone who sits behind with the team that builds the product

and is able to kind of, yeah, tell us about how it works

here at Microsoft around it as well, which is cool. But one of the things about this show is we want to kind of

almost debunk a few myths where it's like everyone does the same thing every day or there's only one job

in the space of AI. And so we kind of ask, I know this is super hard

because, I mean, your day even today hasn't started probably its average day

Then again, you do a lot of filming. So thank you for joining us. But also, all jobs in the industry are different daily

Can you describe or break down? What does your job look like

Are you coding a lot? Are you on email a lot Like how does it work Yeah I actually do a session called A Day in the Life of the Data Scientist And it starts off with there is no such thing And basically what I do

is I divide the clock up into the different tasks that a data science person might do. Now, data

science is a big term as well, and has gotten bigger and then smaller and then bigger again

as time has gone on. But in general, I do something very similar to what other data scientists do. We

start out by trying to define a problem we wish to solve or a question we want to answer, which is a

little different than, say, a business intelligence, where I provide you a set of queries and you can

come up with whatever questions you want, and then you can access that data. In data science, it's

slightly different because one algorithm usually equals one prediction. I can't use a regression

an algorithm to do a clustering, and so on. And so the first thing I have to do is define the

question, and then I need to locate, verify, vet, and so on all of my data sources. Then I need to

figure out how I'm going to clean them, which is 99.99% of a data scientist's job. And then after

you clean everything up, you actually pick an algorithm, which is the easiest part, to work on

that data. And then you compile a machine learning algorithm into an experiment. You run a few

experiments, you get your stuff. And then out the other end, you get a prediction or score

And then I work with developers and I've done the development side as well, but I work with

developers to create an application that will actually use that prediction in a way. In addition

to that kind of thing, I'm also a manager here at Microsoft. So I have those kinds of tasks

I'm also on the product group. So I'll sit in lots and lots and lots of meetings where we plan what

we need to be working on and when things release and how things release and personnel things

who needs to do what and so on. And then I also teach, lecture quite a bit and develop courseware

things like that. And then yes, I do a lot of public speaking as well. From time to time I

sleep. This is actually a tattoo. It's not a shirt. That way it saves me the time to get dressed

So I can just shower off and start the next day. And you just drink lots and lots of coffee

I love that efficiency. That's always always the key, isn't it? To these busy, busy roles. Gosh, yeah. So that is that is quite a vast array of different techniques you have there. So from managing people and all the skills that come with that to being deeply technical to teaching

So thank you so much for coming on our show. But also a lot of our guests work with the community just like you do

And obviously we always just want to say, like, thank you for doing that, because actually that giving back element allows everyone to learn as well

How do you find that? How do you find sort of the teaching side of it? I've seen you teach

In fact, you taught me very early on when I was at Microsoft, which was amazing

And I learned a lot from the way you kind of like workshopped and worked with people

Are you still doing some stuff like that today? Absolutely. Both in Microsoft and then privately, you know, working for I taught at the University of Washington for a while

And I also teach at high schools, volunteer at high schools and things

I have a data literacy course that I teach that brings students up through, I would say, the basics of data science, learning how to vet data properly and make sure that people employ critical thinking when they hear information and learning to deal with information in that way

I'll send you a link on that. Maybe we can post that on the show of the data literacy course

it's on GitHub. It's made not only for me to teach, but I would love for anyone to teach it

anywhere they can. And it's suitable for teenagers. And I even made it simple enough to where a C

level, a CEO or a CTO could understand it if you slow down. Nice one. I'm so glad you thought of

that CXO level. That's, they will be very much helpful. so

Buck, you've been doing this work for a long time so I'm wondering

how did you get into AI ok, yeah so I grew up very poor

didn't have a lot of money in my family really poor and I lived on the space coast here in Florida

during the space race years and And there was this interesting show that came out that came out during the late 60s called Star Trek

And I became fascinated with this show. I would watch it on a little black and white TV every night, just stunned

And there was a character on there, Mr. Spock, who was the science officer

And they actually dealt quite a bit with thinking machines, which, believe it or not, those concepts go all the way back to the 50s

And even before that, where people were thinking about, OK, there's these things called computers

I bet they're like a brain. And of course, these things went through stops and starts. Right

And so I began to work with technology as much as I could wherever I could find it

Not nearly the programs, obviously, that they have today, especially for kids that are disadvantaged

But I made copious use of our public library and other places, joined the military and worked with technology quite a bit there

lived in the UK for a few years at that time, and got into computing. It was kind of breaking

really big time in the 1980s there in the UK. So I got heavily involved in that. And then when I got

out, well, before that, I started playing with things, now it's called, which is kind of insulting

good old-fashioned AI. So in college, I learned statistics, fairly deep statistics

for, because that's what you did, right? That's how you made your predictions and so on

and modeling, we called it then. So you were called a researcher. There wasn't a title of a data scientist

It was a researcher. And then as time went on, I got into data mining

And a lot of folks are not aware that data mining and business intelligence are actually two separate things

And then there was this idea of what they now call good old fashioned AI, fuzzy logic

and a few other constructs, which were, we like to say now

were really just a series of if-then-else statements, and it wasn't far from the truth

Languages I played with were things like Prologue, which does really well on things like text ysis

believe it or not. In fact IBM Watson uses Prologue to this day on Red Hat Yeah it still in use But at any rate I sort of say that good old fashioned AI and statistics and then data mining got together in a wild party

And they had a baby and it was called the machine learning era that we deal with now

because you have some of those same constructs from each of those disciplines that are now available

But we have a couple of different things. The two different things we have now are large, large sets of data

And the other thing we have available we didn't have before are the ability to distribute the processing in a new way

Hadoop, obviously, this is the beginning of that. Now we see Spark and other things come into play

So I worked in these things as I came up. And I always tell people I do quite a bit of mentoring

and I tell people in an immature industry, generalize your knowledge. And in a mature industry, specialize your knowledge

Meaning know a little about a lot or know a lot about a little

One of the two, but it depends on how mature your industry is

In the early days of computing, you built your computer. I remember building mine, soldering things on

and literally creating my own computer. I had to write an operating system

used assembly to do that, and then write your own applications and so on, and then record those

onto tape is the way we did it back in the early, early, early, early days, reel-to-reel tape

So at any rate, you did everything. You did electronics, you did coding, you did whatever

you needed to do. Obviously, that's not the way it works anymore, right? There are people that do

nothing but hard drive controllers or, you know, SSD drive controllers and so on. And there are

people who write only one kind, one part of the stack. And data science was like that. When it

first came out, you sort of had the data scientist. She was sitting back in her office. She had a big

gigantic system, and she ran R with a lot of memory. And you would take her a bunch of data

and you'd say, I want to know this thing. And she would say, go away. And then a week later

she would come out with two tablets from the mountain, thus saith the data. And then you would

do that from the Oracle, and you'd move on. And now, of course, that's changed, right? We have

data engineers and we have people that do UIs and we have people that do the algorithm itself

and so on. So as these things have broken apart, I've sort of been dragged along with it. But where

I started was more of the end to end. So that was my path. I think there's like three paths to get

into data science. One is a formal degree. And I was actually on the degree program for the

University of Washington building a program to say you're a data scientist. And then the other

one was you were an yst or a researcher, somebody who had those disciplines, but you also

had the coding skills that were needed. And you were brought up in an era when the distributed

systems were already out. And then there's my path, which is just a little at a time for a very

long time. So I like to say I'm not, you know, good at driving. I just know where the potholes are

So I tend to stay away from some of the things, the old mistakes that we used to make. So that's

kind of my path. That's how I got here. And I think everybody's path is valid and different

You do not need a degree to be in this field at all. There are things that can be done right now

regardless of your background, meaning if you're a database administrator or perhaps a BI developer

or perhaps a software developer, there are places for you to get in

There is enough data cleansing that I think we could employ the world. I was going to say, we briefly skipped over the 99.9% of your job is data cleansing

I'm assuming, yeah, that is still the case in some senses. Still the case

The bit that people, yeah, the bit that looks like the sparkly unicorn of AI

which is choosing the models, is, as you said, maybe not quite as much of the actual end-to-end as you would imagine

But I really appreciate your book for sharing those different paths. This show is, I've said before, I would say this show is a success

if someone watches it and says, I don't know, in a few years

like, oh, I watched your show and now I'm a data scientist or now I'm a data engineer

or something like that. And they've kind of got into what they're doing because we're able to share interesting

resources or insights. So I very, very much appreciate that. But this has been like a very, very nice conversation, very positive conversation

But I don't know if you know Hank, but he always asks a very tricky question

Oh, he does? Okay. Well, let's take your best shot there, Hank

Yeah. So what is the most annoying thing about your role? Other than people

Yeah, specifically AI. Okay. No, I have an amazing team around me and a great set of folks I work with

The most annoying thing in AI is one of two extremes. people believe that it's all hype and doesn't do anything that it's all fake it's not real and

just dismiss that and the other end where it can do anything and apparently you know Ultron or you

know the Avengers or whatever is real and so I constantly either have to tell people no that's a

thing or I have to tell them yes no it doesn't do that right it's not it's not omniscient we don't

have these things. Deep learning is to the brain, you know, as a cell is to a body. It's not even

remotely close. We use some terminology there, but it's nowhere even remotely close to what you have

So I think the number one, to answer your question, the biggest pain is going to be

separating out expectations, managing expectations, because it's in every call. I had one time I went into a place and they said the CEO was talking to the team and they brought me in to talk to them

And I said, how can I help? And they said, well, we know you're a data scientist. We need to get some AI

And I said, I said, you do. And he said, yeah. And I said, do you want like a large box of AI or maybe a grande or maybe a tall AI

I mean, what what are you trying to do? And getting them to actually tell me what they they just they heard AI

They're like, we should get some of that. And I and I said I think you need to kind of come up with what you don know And here how here the question I call it the probing question And I keep asking so what So what So what

And the end question, the last question is this one. What would you do differently if you knew the answer to that question

Yeah. That's quite a cliff to leave us on in some senses

But no, I appreciate that one. That is not a lot of people have said, oh, I find it difficult when someone doesn't truly understand AI

But you've broken it down quite nicely to say the expectations are both

I think it does nothing. It's just if then else statements. And I think it does everything and it can solve every problem we've ever had

And we don't actually need to know what our problems are in some senses

um so no i very very much appreciate that before we wrap up i've told how quick has has this gone

i really really appreciate book thank you for sharing your journey thank you for sharing what

you do um but we do have our quick fire round question and we definitely wanted to get you

to participate in this so a quick fire round is six questions me and hank will alternate between

asking the questions and the only thing we ask is can you answer with either one word one phrase

or one sentence at the most so let's keep it short let's keep it punchy um but also we i think

the first thing that comes to mind is the most telling um so book are you ready for our quick

fire questions let's do it i've had a lot of coffee good stuff good stuff um so first question

you've briefly touched on it in one of your answers and i was like oh you'll you'll love

this first question what was your first computer i built it myself

so like all the different pieces were just like all different wow fabulous there we go so that's a

that is not an answer we've had before i don't think exactly see if this one is also something

different. What programming language was used in the last project you worked on? Python

That one's often not that different. What is the most useful thing you have learned in AI

To clearly define the question. Nice. Perfect. That's a great one. What is your favorite event on the AI calendar

Strata. Oh, nice. Nice. We'll have to share a link to that one

That's a good one. What area of AI is on your list to skill up on next

So with all those hundreds of things that you're doing, if you had time to learn even more, what's on that list

F sharp. Interesting. And what was the first team project you built that kind of included AI

A medical recommender system. No way. That's quite intense for your first one

Do I dare tell you that it used fuzzy logic against a Microsoft Access database

Wow. I don't know. I don't know. I don't know whether we should dig any deeper, actually

Let's leave it at that. Let's just say, hopefully, everybody that used it is still alive

Let's just say that. Oh, my goodness. Wonderful. We love to end the A Bit of AI show talking about Microsoft Access

It happens more than you would think it does, Buck. But thank you, Buck

Thank you so much for your time. We are going to need to wrap up the show. So 30 minutes is never enough

So to all our guests, all our viewers out there, make sure to follow Book on his social handle

You can see it just below his screen here at Book Woody and then NSFT, which sounds for Microsoft

And yeah, Book, thank you so much for joining us. And we will hopefully speak to you again very soon

Absolutely. Thank you both. Thank you. See you later. Okay, we have like a whole of two minutes, I think, to get our announcements out today

So just very quickly, we have had Microsoft Ignite this week. If you want to check out the AI sessions, either go to our website, find our show notes or go to aka.ms slash ignite dash AI dash sessions

And we've got a nice list of all of those different AI sessions going on

Also, if you want to see more from Book and also his amazing colleagues such as Anna Hoffman over in the Azure Data Space, you can check out the Data Exposed show

You can find a short link with loads of amazing information on like a special series they did on AI and machine learning

So it's aka.ms slash S-O-S-N. And that stands for something old, something new, which is quite exciting

there is there is all sorts there's themed t-shirts there's dad jokes there's math there

is scripts there is azure machine learning in there um if that sounds like your type of thing

uh do go and check it out that was aka.ms slash s-o-s-n which is something old something new

and with that that's pretty much all we've got time for today and as always we hope you've enjoyed

the show if you're new to the show or you've maybe missed a few weeks catch all of our episodes on

demand at a bit of ai sorry let me start that one again um at the a bit of ai.show website um and

join us after in the cafe where you can speak with me and hank ask us lots of questions we'll share

lots of links about all the stuff that we've spoken about today that's aka.ms slash a bit of ai

dash cafe. And with that, thank you so much for joining us. And this has been the A Bit of AI

Show with Henk and Amy. We'll see you soon