Behind The Tech with Kevin Scott - Sam Schillace: Deputy CTO, Microsoft

Episode Date: August 10, 2022

Have you ever wondered who was behind the creation of Google Docs? Kevin talks with Sam Schillace, one of the people who launched the transformational – and at the time controversial – idea 16 years ago. Today, Sam is an engineering Corporate VP focusing on consumer product culture and next generation productivity who stresses the importance of asking “What if?” questions. Kevin and Sam talk career advice, the fact that there’s always tension in high functioning teams and why we’re on the cusp of a new technological era.  Kevin Scott  Behind the Tech with Kevin Scott  Discover and listen to other Microsoft podcasts.  

. Find the thing you feel guilty about getting paid for, do the hell out of it. If you feel guilty that you're getting paid to do something, it probably means you're really good at it and it's fun for you, and if people are willing to pay for it, just go do the heck out of that. Hi, everyone. Welcome to Behind the Tech.

I'm your host, Kevin Scott, Chief Technology Officer for Microsoft. In this podcast, we're going to get behind the tech. We'll talk with some of the people who have made our modern tech world possible and understand what motivated them to create what they did. So join me to maybe learn a little bit about the history of computing and get a few behind-the-scenes insights into what's happening today. Stick around.

Hello and welcome to Behind the Tech. I'm Christina Warren, Senior Developer Advocate at GitHub. And I'm Kevin Scott. And today we have Sam Scalace with us. He's somebody who was a huge part of the creation of Google Talks, and now he's working with you at Microsoft. Right, Kevin? Yeah, Sam and I have known each other for a very long time.

So Google, many, many, many years ago, bought the startup that Sam and his co-founders started that created the first interactive web word processor, like back at a point in time where people thought that doing that was impossible. And that work became Google Docs. And Sam ran all of the Google Apps business at Google for a while. He went on and ran engineering and operations at Box, another startup until it went public and was back at Google. And now he works here for us at Microsoft. That's awesome. And we're really lucky to have him, especially like all of his experience. And

also, thank you for bringing the word processor to the web browser. That's amazing. All right. So let's go ahead and let's dive into your conversation with Sam. Sam Scalacci works at Microsoft in the office of CTO as Deputy CTO and Corporate Vice President, focusing on consumer product culture and next-generation productivity. Prior to Microsoft, Sam spent three tours at Google, GDocs, Google Ventures, and then Maps and Search. Sam originally worked at Google as a founder of the Rightly Acquisition and led the teams that built Google Docs. He has a long history in consumer software as the focus area lead for consumer apps at Google, and in multiple startups he's founded.

He also worked at Box as head of engineering for four years, where he and I met on the technical advisory board. He holds a BS and MS in theoretical mathematics from the University of Michigan, where he focused on combinatorics and discrete mathematics. Sam, it's great to have you on the show today. It's nice to be here. Looking forward to it. Yeah. So, we always start these podcast conversations by going all the way back to

people's childhoods. And I'm just super interested in how you first got interested in science and math and programming and just what lit the spark. Yeah, it's kind of funny i i'm a little bit of a black sheep in my family like my mom's an artist and my dad's a psychologist and a psych professor and so like it's a it's always been a little bit of a mystery like i've always been a science kid like that was always the thing i've liked as long as i can remember i still have this old microscope my mother bought me as a or my parents bought me as a birthday present when i think i was like eight you know it was like the coolest thing ever that i had this like cool old microscope and

like i was super super into it like all always like i wanted to be a doctor when i was younger i wanted i did a lot of like biology and stuff like that in high school and things like that like when i was about 12 for reasons that i don't even really think i understood even at the time and certainly don't remember. I really got into doing computing and the computer at the time was the TRS-80, you know, model on the Trash 80. Convinced my dad to buy me one, this used one from somebody.

It was like 600 bucks, I remember, which, you know, is like the equivalent of like three grand or something a day, like a huge amount of money. You know, I don't know why he did it really, but he did. And he bought it and like, I remember it being down in our basement and I would like, you know, write basic on it, you know, out of the back of Dr. Dobbs. And the pixels were three by two. They weren't even square pixels, like this chunky little thing that you could do

like really basic stuff on. And if you felt like, you know, not retyping your program, you'd tape it out to a cassette tape and, you know, read it back in later and all that crazy stuff. And then a friend of mine had an Apple II that we played on and, you know, wrote video games. I was really, I like pixels a lot. Like, I like making pixels on the screen. So, that kind of drove a lot of it. Did your high school have computers? I mean, like, this is in the 80s, right?

Yeah. This was, yeah, I graduated from high school in 84. My high school didn't really have computers or really have a computer program. I didn't really think of computers as anything other than a fun hobby, even most of the way through college. Like, college, the first exposure I really got to anything significant other than those early things was, like, the nursing school across the road from my dorm had a basement full of Mac classics that you could go write, you know, use. And I would go write my papers on them because they were awesome. Like I didn't have to like, you know, I was a terrible typist and I never learned to type touch. So, or, you know, to actually type the right way. And so like, you know, that was great because I

didn't have to like white out and backspace and retype all my papers. I would just go do them there. And I even remember even then, like all of my like friends and roommates and stuff, classmates thought it was a weird thing to do. Like, why would you go type on a computer? Like we've got all these perfectly good typewriters. And I was like, no, this is way better. And that's where I sort of got, started getting into them. And then like, I did a lot of like goofy, weird little side programming projects in college, like making like sort of pseudo three two and a half d landscapes that were randomly generated that i like to look at and crazy stuff like that and then you know i took one class in one cs class in college was like the

knuth algorithms class you know first book it was cool but at the time i was like really into math and really into science and thought i was pre-med and I was doing like graduate math classes. So that intro to algorithms class was just super easy. It was like the only thing I ever got an A plus in ever in college. And my math profs really looked down on it and sort of discouraged me from doing computers because it's like, well, there's like math, then there's applied math, and then there's engineering, and then there's whatever the hell those hairy computer guys are, you know, it's like the bottom of the bucket. So I didn't really pursue it. Like, I didn't get a CS degree. I kept doing my math degree. Got to the end of college, and my college roommate, who was one of the co-founders of Rightly, ultimately, had dropped out to go do a company that was also a word processing company, something called Ann Arbor

Softworks, that got bought by this company called Ashton Tate before Oracle killed them. And he got moved out to the Valley. And I didn't know what to do with this weird math degree. And I didn't want to be a doctor anymore. And so I just moved out to the Valley and started writing code with him. We wrote a video game and then ultimately

wound up starting six companies together and just worked together for about 25 years. That is... Just fell into it. I mean, just absolutely fell into it. Well, so I'm going back a little bit. I'm curious, like why math? I don't know. Like I liked math. I thought it was fun. I've always been kind of an abstract thinker. I like math. I like the, you know, I like being able to kind of tinker and build stuff with your mind. I mean, that might be the kind of the through line of all this. Like software is always, you know, you and I are both makers.

We like to make stuff. And I always liked software because it was the most that you could, you know, the most direct connection between your mind and making something, right? Like as little mediating that as possible. Math is kind of the same way, right? Where like you can think really interesting, crazy, cool thoughts, but you can do this in like this interesting structured way where you're not just bullshitting. Like you're actually like, you know, like I did a bunch of stuff with like set theory and the completeness and correctness theorem and the omega number and all this other, like, you know, the halting probability. Like it's just wild stuff. It's like metaphysical

and cosmological, but like it's grounded in some real rigor, which is just a really interesting way to think about the world, right? I just find it to be this amazing tool set for understanding stuff. Yeah, it is an interesting thing. Some mathematicians go off in a completely different direction, but like a lot of what you do in math, and I don't think I really appreciated this enough when I was taking a lot of math in college, is that it is teaching you in a sense, like how to do abstraction really well.

And decompose problems. And I think like there's very little that's sort of really directly valuable in the computer career from my math degree, although there are some definitely some ways of kind of thinking about O of N problems and stuff. But the thing that's super useful about it is it taught me to be alone in a room with an unsolved, you know, seemingly unsolvable or extremely difficult problem and just, you know, not have any excuses, just like work through it, right? And like one of the, like this class this class at one point like there's group theory which most people know sort of like this

abstract theory of algebra then there's semi-group theory and semi-group theory you start to take away some of the axioms like you don't have associativity or commutivity of the binary operator and man is that stuff hard you look at these things these theorems that you have to prove and you're just like well that's obviously true and these theorems that you have to prove, and you're just like, well, that's obviously true. And there's like, but you got to prove it. And there's just like no place. It's like a seamless wall. There's like no place to get any traction on it at all. And like sitting in the room with those things and spending like eight, 10 hours on like a two line proof, you know, it's like a really good mental discipline of like, yeah, well, there's a solution. You just got to dig away at it and you got to break it down into smaller problems and you got to, you know, figure out what you

know is true and what you hope is true and how you get to the things you hope, you know, like all that skill is super valuable, I think, in engineering. Yeah, I learned a similar set of lessons when I took linear algebra in college. So, my linear algebra prof, it was a very theoretical course. And he – 100% of your grade were two exams, a midterm and a final. And on the two exams, they were 10 true and false questions. And they were take-home exams, and you could spend as much time as you wanted doing them. And it was brutal. Like you could take those home

and to get to, oh, I'm confident that this is true or this is false was a massive amount of theorem proving. Yeah. I remember doing one of these midterms like this, I think for like one of my group theory classes where it was 12 problems like that.

And I took it home over spring break and I got one problem done a day. Yeah. Like it literally like could grind through one of them every morning. And then I would go out on the beach or whatever, where we were. Yeah. It's, I think that discipline is, you know, this is the nice thing about math. I think is that it's very easy to fool yourself. You look at something and think you understand it, or you look at something and you think you know how to do it, but like doing it is the proof, right? And math really teaches you that. I also had a, when I took combinatorics in grad school, I had a professor who would occasionally slip an unsolved problem into an exam. And what that teaches you is what

to do when you are faced with a problem that is literally too hard for you or anyone else to go solve? Like, can you get anything out of it? Just the process of making the attempt. And, you know, the answer to that question is like, obviously, yes. I had one of those happen by mistake once in one of my advanced classes where the prof would give one, two, or three stars. One was like, you really should be able to solve this this. Two was like advanced. Three was unsolved.

And he accidentally mislabeled a three-star as a one-star, and he had a very unhappy group of kids on Monday morning who had spent the weekend banging their heads against this obvious problem that was unsolved, actually. So, how did you go from being a theoretical mathematician to, I understand the impulse to go work on some of this software, but, like, why work on software in context of starting a bunch of companies? You know, again, it was a little bit of this like kind of falling into it and a little bit of networking, right? Which is also one of the reasons why I try hard to like network outside of my particular group, because I think networking is why we don't have good diversity in the tech industry. People network with people like them. And so you kind of have to force yourself to go

and, you know, break those networks. And I think it's an important part of increasing diversity. But we were at Ashton Tate. Ashton Tate was, we got hired into the Mac division of Ashton Tate. And they were in the process of being strangled to death by Oracle, basically. So they were a database company that Oracle was crushing. So that job only lasted about a year and a half. And about six months into it, it became very apparent that it wasn't going anywhere. We were trying to figure out what to do next. So my friend and I, Steve Newman, were just bored. And they'd loaned us like a Silicon Graphics machine that we were playing around with the flight simulator on it.

And so we were bored dudes and we wanted to play games. And there weren't any really good games. Like Castle Wolfenstein was the game that was there at the time. And we just wanted a Twitch game. And so this is actually a pretty good set of learnings from this game. We built this thing called Spectre. And it was a very early multiplayer game. It wasn't even Ethernet. It was over Apple Talk.

So you could play against other people in real time, you know, in your land, like within the office. Just a straight up like first person shooter. You're driving around in a tank. It looks like this old game Battlezone. And like it taught me a whole bunch of stuff. It taught me about like really focusing on user need and user value. Like we knew what we wanted. We were the customer.

We were iterating with ourselves. We wanted a Twitch game. We wanted it to feel really clean. We wanted it to be really straightforward. We want stuff to blow up a bunch. And so like we built that, which was really hard to do on the hardware of the time. It was super hard to do with the networking stuff we had to deal with. I learned a whole lot about value in programming is in confronting these really annoying hard problems and getting them right.

We got a patent for how the rendering worked because the hardware was too slow to draw all the pixels and so we had to like erase a bunch of them and there's all this kind of stuff like that so that taught that taught me a whole bunch of stuff and that like wound up paying for like four years of my life like that we we sold it to this company or licenses to this company that just like published it for us and that was like you know good money for me you know as kid, like kind of to live on. And then, you know, as Ashton Tate fell apart, the group that had been part of that company went on to do another startup,

which didn't really go anywhere. But Steve and I went with them and worked on that for that four-year period. And when that failed, we sort of took the pieces of that and, you know, the dot-com boom was starting. And so we turned that, the pieces of that company you know the dot-com boom was starting and so we turned that the the pieces of that company which was like a personal information manager but we had built like a good cross-platform mac and windows app framework out of it we took that and quickly pivoted into

building internet tools one of which we sold to claris so that right there is like three startups in a row it's like bing bing bing you, bing, you know, like kind of like, kind of by accident, but like, you know, again, like if that, in that era, I spent a lot of time, like feeling bad that I was doing something fun for a living. Like I kept feeling guilty that I sort of dropped out of being pre-med to go do this thing. And I literally spent like the first 10 years I was in Silicon Valley thinking that computers were a fad. I needed to go back to med school, get a real job. Like this is just kind of a goofy, stupid thing to do. And eventually after about 10 years and several successes, I was like, well, I guess this is my life now. You know, I've got money in the bank and this is kind of fun and it seems to be continuing to develop.

So I guess this is a career. Let's go do this. And there's actually a lesson in there that I tell people all the time, which is, I think we're really confused. Mostly we have this kind of Calvinist ideal idea in our heads that like work is equivalent to suffering. Like you're not really working if you're not, if you don't hate it. And I don't think that's true. I actually think the place for people to be, this is the career advice I always give people is like, find the thing you feel kind of guilty about getting paid for, do the hell out of it. Right? Like if you feel kind of guilty that you're getting paid to do something, it probably means you're really good at it and it's fun for you. And if people

are willing to pay for it, just go do the heck out of that. Like career growth usually comes from having impact, which usually comes from doing something you love with a lot of passion. Like that's, it's not that much more complicated than that. So that's how, I mean, so I wouldn't, I never started to think, I never started off with like, I'm going to go do a lot of passion. Like, it's not that much more complicated than that. So that's how, I mean, so I never started to think, I never started off with like, I'm going to go do a bunch of startups. I just was like, what's the next interesting thing I want to go work on? Let's just go do that with all this energy.

I always liked working with my friend. So we just kept doing stuff and we kept making money and it kept being successful enough. So like, why stop? Yeah, and I think you've, you had another couple of pieces of interesting career advice in there as well. So I totally agree with what you just said. Uh, like I, I also agree that working with your friends, like whether they were your friends before you start something or whether you like form bonds with folks, uh, you know, when you

join a company and you start working with them, I think that's pretty important. Because on most days, like, even when you've got a thing that, you know, you're passionate about and you nominally enjoy, like, there's just a bunch of hard stuff you got to go do to make anything worthwhile. And so, you do want to be doing it with people whose company you enjoy or in where you feel like some degree of camaraderie with them. Otherwise, it's just, I mean, you and I do this now, right? Like, I don't want to sell the image that Microsoft is, you know, somehow perfect. Like, we got hard things we work on all the time. And then the two of us, like, you know, because we're friends, we will, you know, we will complain to one another.

And, you know, the thing that you try to do, I try to do, yeah, I even do this in my marriage, is like, you want to sort of be grumpy out of phase with each other. Yeah, yeah, yeah, that's right. You want to be complimented. There's all kinds of complementarity that you want in these partnerships, right? That's definitely one of them. Skill sets is another one, right? Like, you want a linear thinker and a nonlinear thinker. And then you want them to be, like, you know, in creative tension with each other, basically, in a constructive way, right? Yeah, I mean, the other, like, the most interesting, like, complementarity advice that I ever got is I had a mentor years ago who told me to imagine a histogram that has five buckets. And on the extreme left end of the histogram, the bucket is labeled idiot. And on the far right end of the histogram, the bucket is

labeled genius. And in the middle is average. And you can take everything that you do and every skill that you possess and put it into one of those buckets. So, like, that's not the, you know, the breakthrough thing for me. Like, the breakthrough thing was this mentor said, if you work really, really, really, really hard, you can move something over one or two positions on that histogram, which means that if you've got a thing that you are an idiot at, like... You might get to average if you're lucky. Yeah. And every minute that you spend trying to get to average is a minute that you're not spending doing the thing that you're a genius at. Right. And, you know, like what you want to do with teams

or partnerships or anything else to your point about, you know, nonlinear versus linear thinkers is you want to like, you're trying to do something together. You need a set of skills to go do the thing. So like, how do you like figure out this thing where everybody's histogram adds up to like above average, where you can have everybody focus on what they're really good at. Well, and the real challenge with this is that, you know, just to keep going with that example,

like sometimes the two, you know, if you've got two geniuses that are pulling in different directions, they neutralize and then you don't have either genius. And so like, and like, in fact, that often happens, right? Like this, I had this tension a lot with my co-founder, Steve, like, you know, when you have genuinely different perspectives on the world and you are genuinely both good at them, those are often, you know, you're kind of blind to the other perspective to some degree. And so finding a way to like understand and respect the other perspective, even if you don't really get it, like you just understand like this person, I don't get their domain. I don't even maybe necessarily fully value it, but I understand that it is valuable and they're good at it. And so I'm like, you know,

deliberately like carve out some space. I always tell founding teams, like, pick somebody you like arguing with. Yeah. That's, you know, don't pick somebody you like. Like, you know, it's easy. Don't just go found something with a friend. Pick somebody that you enjoy arguing with where you have genuinely different perspectives where you're struggling to even find common vocabulary. But it's okay. You're willing to have those arguments to kind of find that common ground in between. Yeah. Because often there's super small overlap in the Venn diagram of even the language of these, right?

Even going back to you know marriage like my my wife is smarter than i am and like a lot uh and yeah there there were a bunch of things when we first got together that both of us were both of us were good at and both of us like doing and eventually we learned that like we we just sort of had to pick on a bunch of these things where i sort of had to say like all right this is going to be your thing and like i'm not going to try to get up in your business on this and and vice versa everybody has comparative advantage and stuff too like for one of the things drove me nuts my marriage for a long time is i come up i have opinions about stuff my wife doesn't as much right

like i'll know what i want for dinner and she won't or whatever. And like, you just sort of decide like, okay, well, that's fine. Like, you know, you kind of get comfortable with those different modes and like you take that role or whatever. But, but I, I think the thing, you know, that we're sort of getting at the meta point is like, you just really have to be thoughtful about how you're trying to do things together with other people. You have to pay attention to what the dynamics are

and try to make things work for everyone. And you have to recognize that that's not, like if you're struggling with a coworker, a team member, whatever, like that's not an anomaly to be worked out. That is a fundamental requirement of high-functioning teams is that you're going to have tension between different perspectives.

And that's like a fundamental critical value of being a mature team member and leader is that you're actively aware of this and managing it. Like you're never, there's no easy, like, you know, if you snap a bunch of people together, they're all the same person, that team will fail. There's always, there's always tension in these teams that has to be, have them be high-functioning. That is a skill and and i think the thing for managers as well is realizing that you can you can realize all of what we just said and still be a victim to some of the stuff yourself and which is why it's really important if you are leading things to have people in your network who can help you go deal through your own short-sightedness where intellectually you should understand how to go deal with it, and yet you just aren't because you're a human being. That's actually a really good illustration of one of my favorite things in tech in general is favorite patterns that shows up in a bunch of different ways, which is this tension between

long-term and short-term, short-term, right? The manifest is like the tension between what I know I should do in the long-term versus what I want to do right now. I always joke, it's like, well, I want to be on a diet, but there's a donut in front of me and I'm eating that thing. You know, like that, that's kind of the easy way to think about it. But like here, right, you know what you should do. You know, you should build and manage this thing in the long term, but it's like hard to do because what you want is a day-to-day life where you don't have to deal with tension among your team members. And so you're kind of biased in this direction of picking the easy thing. And the thing you mentioned of like, you know, reaching out to outside groups and stuff like that is a common solution to that.

You have an externality. You pick something that's an external forcing function that makes the short term kind of equivalently expensive in some way to the long-term or reduces the cost of long-term. So you do the right long-term thing. But that pattern manifests in architecture, it manifests in development tooling. I mean, it manifests everywhere in the tech industry. I see it as one of the fundamental challenges of doing what we do very well at scale. Yeah. So going back to what we do very well at scale. Yeah. So going back to what we were talking about 10 minutes ago, just in terms of career advice, one of the other nuggets that you had in there that is worth calling out is it is very important,

I think, to choose to go work on things that are hard um like usually the meaningful things things that have impact like things that have lasting value uh were hard uh and like if they weren't hard somebody else would have done them and the problem would be solved but like you know the the yeah we we had this you know hypothetical you know frontier that we're all you know as human beings trying to collectively push forward and like getting you know the frontier push forward is just a inherently hard thing uh it pushes back against you um and so the only way you make progress is just with the choices that you make you have to choose to go work on hard things. You have to sit in discomfort all the time with this stuff.

Yeah. Which is hard for engineers, right? Like, the engineering mentality, we like neatness. You know, people like things to work. Our job is to solve problems, and so we don't like things to be unsolvable or to confront stuff where you don't know the solution or whatever. Like, one of the common mistakes people make

is waiting for that moment where the thing seems obvious and easy to do. Cause it's, as you said, it's too late. Like I always think the right time to work on a new problem is when you have this pit in your stomach where you just know that it's going to happen, but it also like, you know, you just know it's going to be hard and miserable to grind through it. Cause it's so early. Like, you know, we can talk about this in the context of GDocs. Like, I always give this example of the early days of Rightly. We're like, you know, I had this, I didn't really, I can't claim I had the full idea for GDocs early on. What I mostly had was just me messing around with JavaScript and content editable on the browser because it seemed like a fun thing to play with, which is a lot of what we would do.

But, you know, we had this idea of doing this word processor and we kind of stood it on its feet and we cleaned up some of the early collaboration stuff. So we kind of worked together without stepping on each other. And, you know, we had this sense, first of all, we had this sense that like this was cool and it was going to be really hard to pull off. And then you kind of had that sinking feeling in your pit of your stomach, like, oh man, is this going to hurt? But also like my two founders said you know this is a dumb idea we're not going to be able to build a word processor in browser that's satisfying and the funny thing is that they were right as well as being wrong like in the moment they were right the browser of the day ie4 and early firefox and all that stuff

like there's no way man like you know they they were really not ready. The JavaScript environments weren't ready. The DOMs were flaky. There were no standards. Like, in a zillion ways, we were not ready. But the interesting thing was, like, we solved enough of the problem. We confronted that hard stuff at that early stage and solved enough of it to demonstrate value to ourselves and to some other people. And that started a flywheel, right? And that flywheel between, like, the browser got better, the app got better, the users adopted it more, the browser got better, you know, like we

had these three pieces, this triangle started to spin. And that dragged us forward to where we are now where the browsers are super capable, the apps are super capable, billions of people use these things. So like, you know, that's, you know, that 100% agree with you, like, that is the thing you have to do is like, you have to be able to see those vistas as they open up, and then you have to jump into them when they're nasty hard to deal with. Yeah, and look, I think it's a good thing to just sort of pause and think about. that like obviously you're going to be able to have a interactive word processor you know with collaboration and you can comment and you can share it with anybody in the world and like you've got all of this stuff when were you first doing this it was 2002 five two two thousand five so like in oh five like nothing like it existed and so everything today that's in your life, whether

it's a PC or a laptop that you're able to buy for not that much money, relatively speaking, that you've got an internet that's got all of this information on it. You got a device in your pocket that connects you to all of that. You have things like the, everything that we take for granted now was impossible at some point in the not too distant past. Right. The iPhone didn't exist when we started doing this.

Like we were right before it, right? Like, you know, I don't know if AWS was a thing then, but I don't think it really was. Like I think like LoudCloud was a thing back then, right? Like it was sort of like, yeah, like all of it

had to happen somewhere, right? Like, and I think Jobs had a quote about this,. Yeah, all of it had to happen somewhere. And I think Jobs had a quote about this. He realized at some point all the cool stuff in the world, somebody came up with it and that somebody could just be him. And so that's why he came up with a bunch of stuff. He's Jobs, but he's right. Everything gets done by someone somewhere. So if you don't confront these

seemingly almost impossible things and you don't get to be one of those people i guess it's kind of the price of admission is you have to deal with it yeah i i mean the other lesson i take from that is you know and i tell people this all the time like i i'm you know i'm not that smart like i didn't you know foresee the cloud or anything like that like we were we were messing around with stuff and we were listening to signal from the market and from users. Right. And that's really the important thing is, you know, this stuff in retrospect, 16 years later looks super obvious. Like, oh yeah, of course, collaborative documents and mobile and all this other stuff is obvious, but like it wasn't at the time it was super controversial. And, and there are a bunch of reasons why stuff like that is controversial.

You know, it's not only that it's just hard to implement, but there are also, like, kind of natural psychological reasons that people push back on stuff, right? Like, you know, you also have to remember, like, in that era, you know, we were mostly on the desktop. The internet was just kind of starting. And, you know, the idea was like, well, Microsoft has won this game. Like, Office is this giant thing with all these features and they have this distribution channel and software is just software comes in boxes and that's just how it is and so like this this disruption of like we're going to distribute differently we're going to focus on convenience instead of functionality we're going to kind of go down this different path like it was super controversial to challenging to

a lot of people's worldview and when you you, when your worldview is challenged, you have this very stark binary choice that you have to make, which is I'm, I'm wrong or it's wrong. Like that's, you know, something either I was wrong about the world or like this, the world is right. And I, you know, or, you know, this thing is wrong. Like, and so usually people choose the thing and humans are great at coming up with stories for why they want, right? We all know this. So if you want to think a thing is wrong, a new challenging thing that challenges your worldview, if you want to think something's wrong, 100%, you will find a reason why that thing is wrong. I told you that my founder said the browser won't support it.

That's a good example of that kind of story. But almost every new technology gets dismissed as a toy or impractical or there's no use for it or whatever. And I always feel like the right place to put yourself is the what if, not the why not, right? Like, yes, you have to ask the question of what if. Like, when you confront this thing and you see a bunch of problems with it, you don't, the why not is like, well, here's all the why not. Here's why it's not going to work, right? Here's the problems, all the problems I can see, the stories I can tell. But the real thing to say is like, what if it works? What if like we transform software into being this zero install on-demand global thing where we can collaborate with people? What does

that world look like? And if that world's compelling, then it's worth doing the effort to break through all the problems. And that's the much better mentality but you have to recognize that in the moment you won't see the vision like you won't see the prize even if you ask the what if questions you'll maybe see some of it but like you just kind of have to like grind through these things like yeah make them happen yeah you know and it's sort of interesting um i mean bezzos has said this a bunch of different times, and I suspect you and I have both experienced it over and over and over again of other people telling you why it is the thing that you are. The thing won't work.

It isn't going to work. I mean, one of my favorite examples, and I won't name names, is when I first joined LinkedIn right before LinkedIn IPO'd, we had a whole bunch of work that we had to go do to make all of the company's development infrastructure better. We had just gotten ourselves into a state where it was getting progressively harder to do the sorts of rapid software development that a consumer internet company needed to be able to do. And I came in and there were a bunch of people inside of the company who knew what needed to be done before I got there. But there were also like a bunch of people who looked at what they were suggesting that we do. And they would be like, Oh,

this is impossible. Like we'll never be able to pull it off. These three people who tried to do this before failed and it was a miserable, embarrassing failure. That's my least favorite excuse. Oh God, it's the worst. I was just like, just because you couldn't do it doesn't mean I can't do it, right?

Yeah. And so, the interesting thing, we did a whole bunch of things that I think are, you know, innovative and like, you know, some of them are now, you know, multi $10 billion, you know, public companies, you know, that spun out of even what LinkedIn was doing. But the core of what we were doing wasn't rocket science. I had sort of done it a couple of other times in different contexts. And it was just a matter of will and determination and enough patience to get it all the way over the finish line. And I remember going into a meeting with a bunch of people and told them what the plan is. And the plan was uncomfortable because it meant you had to,

like we literally burned, we went over this bridge from old build and deployment infrastructure to new and burned the bridge behind us. We literally tore the old system down where there was this three or four week period in December of 2011, right after we'd gone public, where we did not have the capability to deploy new software.

That's pretty terrifying. Yeah, and so like, yeah, I was telling people, you know, what this plan was and like why it was important to go do it and what we would get on the other side. And I had, you know, like a senior person look at me and was like, well, you better, you know, you better hope you're right. Like it was this personal thing that I was doing. I was like, no, no, no, I'm not doing this for me. I'm doing this for all of us. Like we all need to hope that this is right yeah the

proper response is you better too right like yeah y'all better over right um but but yeah it is my least favorite thing is like somebody telling you uh that a thing is impossible i mean one of the lessons the the real lesson for me from the early days of rayleigh which was super valuable was this exact lesson of like then don't listen to the naysayers very much right like and you get this like particularly really disruptive things you know when you when you have a product idea and you give it to people mostly what you get is like a bell curve of interest right like some people are kind of interested they're sort of you know smeared along it's not like there's not really high emotional valence anywhere the really interesting stuff I found has high emotional valence.

You either get a small number of people who are totally on your side because they see the vision and the rest of them want you to stop doing what you're doing immediately or have it die in a fire or whatever. Like they're violently opposed and there's very little in the middle. And we had that with Riley for like a year. Like after Google bought us, we were at Google. I had executives at Google tell me it was a bad idea and they didn't want to fund it and that we should stop doing it. And I was like, I've got millions of people already that think it's a great thing. And

the company's moving over to it. And like, you know, what are you talking? So like, fortunately, like we had enough, like I was not, you know, I hadn't learned the lesson, but I had, you know, a contract that I had to fulfill and I had backing of Eric Schmidt. And we had like a bunch of reasons why, you know, we, we kind of believed in it and kept going and I didn't listen to the naysayers, but like, that was a real privilege of like being given permission to not listen to the naysayers and then realizing after the fact, like, holy crap, like, I'm really glad I didn't listen to them because that would have killed this thing. And it was critical to it coming into existence that I didn't listen to the naysayers.

And that like, that's a's a really deep lesson for me of like, I'm very comfortable being in conflict or in these kind of early chaotic things, like taking these controversial positions. If I have conviction, if I can see a path and I think the problems are solvable and I have a vision for where I think we might be going, it's fine if people object. It's kind of a good it actually is a good signal like you want it like you want controversy you want that energy yeah and and what's the worst that can happen if you fail right well even failing like you know the worst thing that can happen if you fail is you don't learn anything failing when you learn something or that

you learn not to try again you learn not try to try again. Yeah. Like, you know, like I, you know, I've, you know, there's all kinds of things where I've seen teams try stuff and they'll say like, well, that didn't work. And, you know, we can't do it again. I'm like, no, actually you achieved part of it. You just didn't get it quite right. Right. You didn't quite, quite get the, you know, you didn't quite understand the market. You didn't quite understand the problem. But you like, you made progress.

Like you got energy out of the system. Like keep doing it. Like, you know, you'll't quite understand the problem, but you made progress. You got energy out of the system. Keep doing it. You'll get there. I was having this conversation with someone yesterday who had just joined the advisory board for an energy company. And the particular energy company is doing something that's completely preposterous sounding. But it is a thing that

if it's successful, like we would all want it to exist. It would, like everybody would want it. And so, you know, the only question is like, is it a good use of time to go push this thing forward that everybody will want if someone can make it exist. And like, you know, this person I was chatting with is like, yeah, you know, it's like, I don't know for sure whether this is going to work, but I know they're going to make at least a little progress on it. And, you know, even if the person who solves it later, like they're going to refer back to this and sort of say like, hey hey this was like you know an important thing that

helped me solve the eventual problem yeah that's i mean that's that's kind of there's a there's a book that i like by this guy named stephen johnson it's called where good ideas come from i think is the title of it and he talks about like the adjacent possible right and in this principle that like it's very difficult to see more than like one step past the edge of what's known and possible currently. Not many people can see really two or three steps into the unknown. We kind of make these small, we move kind of in small incremental steps.

And sometimes you get kind of concurrent inventions of things when a bunch of different fields advance to within one, each of them is one step away from coming together on some new thing. Like microwaves, I think was the example he gave. Like there were three or four inventions that like, you know, all of a sudden, if you were in like one of three positions, you were one step away from figuring, putting these pieces together. So I think like that's, you know, it's important to make those like those steps forward. And, you know, you just, it's fine to make those steps forward. It's fine to not really quite know the solution to the problem as long as you can have these kind of relatively short chains of problems that you think you can solve, or at least you know the vector that you need to solve them along.

Because someone will get there eventually. Even if all you do is move one step, if three steps were needed, well, then the next person only needs to do two. Yeah. Another really good book along those lines is Arthur C. Clarke's Profiles of the Future. And this is the book where he articulated his three laws. Like, one, everybody knows the third law, which is any sufficiently advanced technology is indistinguishable from magic. But the point that he was making in that book is that it's very hard with high precision for anybody, no matter how smart they are, how good they are as a futurist or a scientist or an engineer to predict exactly what the future is going to be like in 20 or 30 years. And he had some practice with this because

he was a science fiction writer in addition to being like a really quite a good engineer. His framework for looking at the future is like, you know, don't look, don't try to make very specific predictions about the future. Like, make predictions about what the shape of any possible future must be. Like, where are the, you know, progress trends going? Like, what is going to have to be, you know, like a set of things that will be true in the future? And then, you know, sort of direct your effort, like, to your point, along those vectors. You know, you could not have predicted when we were uh even when you and i were in college that we were all going to be carrying a smartphone around in our pockets like it's like it it may have i think our smartphones right or even like a little bit crazier than the like tricorders from Star Trek.

Well, yeah. And like in that era, I remember the Cray-1 was like the hot computer. And I think that's about as powerful as like an iPhone 4 or something like that. Yeah. So, you know, you knew Moore's Law intimately and you liked math and did the math. Like you still don't have a visceral feel for how much that's going to change the world. But you just named the trend. So the trend is you have these exponential progress curves on the price of compute, the density of compute, the energy density of batteries, the transmission and receiving bandwidth of wireless communication.

And so like all of these things are sort of inflecting up. And so if you look at all of that, like you may not predict Android or, you know, the iPhone, but, you know, something's going to happen there. One of my favorite things right now is the learning rate thing, right? We understand there's about an 18% generational learning rate with solar panels and about a 6% generational learning rate with batteries. So every time you double the deployment, you get that percentage increase in efficiency. And it seems kind of funny, like where'd that come from?

But if you look back, like literally like 40 years, you can see it. I was just talking to my wife yesterday about EVs and we're looking at like buying an EV. And I remember when the Tesla came out, that was like six generations of batteries ago. Originally, like the 300 mile range that everybody wants was really hard to achieve. And now all of a sudden, like every EV that's out there has a 300 mile range. And you're like, wow, like did they all get smarter or whatever? Like, well, no, kind of they did. But also, batteries just got way better because there's a baked-in learning rate.

And you look at a lot of alternative energy and climate change. I still think people don't fully factor this in, that solar is going to just keep getting better. Wind, I think, is 12%. You can see the learning rates in a bunch of these technologies. That trend is durable for a while. Yeah, and you get to the point where at a certain level of scale, you just get much broader participation in an ecosystem than in the beginning. So it was sort of an elite thing you know like elite elite engineering

schools and like you know just a very small handful of companies could have even conceived of building something like the tesla uh roadster when uh you know like when when tesla really started getting going today like you know i was hanging out with someone the other day where like they've got two classic cars they're converting into evs right now and like you can buy all the components uh offline the motors the inverters the power management systems the like all the components in the car talk over this car area network that's standardized. Yeah, like not everything is great, but it's almost like the early days of the PC ecosystem where you can just like sort of buy a bunch of stuff,

build your own PC. You can buy a bunch of stuff, build your own electric car. Yeah, you can totally see this coming. I also like to talk about one of our favorite subjects too. I think this is going to happen right now with AI. I think this is like, I have the same sense, you know, all kind of pull all these threads together. Like I have exactly the same sense of imminent disruption

and that kind of uncomfortable pit in your stomach when you know something is going to happen and you have to do the heavy lifting engineering for it about AI. And I think, you know, we're, I think we're at the cusp of a fourth transformation, right? So like PC, the internet, you know, and mobile were these three transformations where a lot of activity became suddenly accessible to computer programming, right? And I think that accessibility or legibility, however you want to describe it,

is like the core thing, right? Once like an area of human activity becomes accessible to software in some way, you get these explosions of businesses and value. You know, we got all the stuff that the internet gave us and then the mobile let us take it out into the real world. And like every stage, like the businesses got bigger, the value got bigger, got more ubiquitous. And I think AI is going to take this into the cognitive realm now.

All this stuff that's cognitive, like natural language and object recognition and video and just reasoning about the world is now becoming very accessible, not to programming in like the large scale, like, oh, we've got these large models to do that stuff, but programming in the sense of like, I can give it to like a guy who's like the guy building the,

or gal who's building the car, you know, and like they can snap pieces together and build interesting things out of it. And that like accessibility to folks of these cognitive, like, I think it's going to be hugely transformative. And I suspect people listening to this, half of them are rejecting that out of hand. Like, you know, and here's

the why not. Like, it's happened before. It's really hard. I don't like that. I don't want people to be out of jobs. Like, all those reasons. It's too expensive. It's dangerous in these ways. But, like, we'll solve them. Like, all those reasons. It's too expensive. It's dangerous in these ways. But, like, we'll solve them. Like, we'll solve those.

And the what if is amazing. Yeah. Yeah, and that is the thing. I wrote an essay that got published in the March issue of Daedalus, which is the journal of the American Academy of Arts and Sciences, where the framing that I was proposing for thinking about this stuff is just sort of cognitive work. And like thinking about AI as a set of tools that we're developing that will help us with our cognitive work. Like the same way – and like we had a whole – in the Industrial Revolution, like you had this big set of changes that were happening with physical work. And so, the practice arrived a little bit before the theory. But the theory came that you had thermodynamics, like you had very precise definitions of what physical work was. Then you had whole engineering practices that developed around it.

And it is just ubiquitous in the industrialized world. Like, I think that's the kind of transformation that we're about to go through. And I think, like, similarly to that, like, you have, there's this process we need to start of, like, unlearning some stuff, right? Like, we had to unlearn what work meant and kind of relearn it for that process. And I think we've spent 40 years convincing ourselves as computer technologists that we have to do a bunch of work for the computers. Like we have to build a schema and we have to put things in a form and we have to like structure data in certain ways. And like, it was true. We didn't have the tools for that. Like, but now we do. And now we understand how to deal with messier stuff. And we don't like,

we don't have to have forms. We have classifiers instead. And we have things that have judgment. And we have things that can synthesize. Well, and there are even some more subtle things. So, yeah, I've begun to think about part of – so, you as a trained mathematician and like both of us as practicing computer scientists and engineers for a handful of decades, part of our job is we wield these big stacks of abstractions to go solve problems. And part of the reason that we need these stacks of abstractions is you can only do so much with the brain that you're born with and it's almost like the abstractions are a compression algorithm to allow you to like go approach problems uh and like you know just sort of get all of these tools inside of your head so i you know a lot of people i think think that

mathematics itself is sort of this uh like universal thing that it's like a necessity of the universe. And, you know, Ted Chang, who wrote the short story that became the movie arrival, I think sort of pokes at this in a bunch of different ways. Like,

you know, mathematics may be more about humans than it is about the universe. And when you, when you have a tool like AI that's a different way to do a whole bunch of cognitive abstraction, you can sort of get after a bunch of problems that we used to solve with mathematics that you're now going to be able to solve in a different way. And, like, that will be uncomfortable. It's really interesting. Like one of the most fascinating things about AI is, is less that it thinks like us and more that it doesn't. Correct. That's the interesting, like it thinks as well as we do, but in a different way.

That's fascinating, right? Like a lot of that, like the, you know, the deep, deep dream stuff, you know, that just like blows your mind with some of these weird ways they'll produce images or just dolly things. Like, it's just really interesting, like, what comes out of these, man. Yeah, and even, I mean, like, even what you said, like, we've got to get new vocabulary for talking about it because, like, in a sense, it doesn't really think at all. Like, you know, when you say think, you are thinking about what, you know, we are doing right now. Right. right now. You have this cognition that's running on your brain. And it is a different qualitative thing than what a model does. Well, it's interesting. This is going to get too abstract and too deep for us to dive into. But to some degree, our brain is a bunch of

these networks that are strung together in a certain way. What we're doing is making new ones external and also folding them into the mix. Just defining what to think means is challenging all by itself. We have this, to your point, very particular sense of what that combination of networks. Yeah. Look, there's just a bunch of things that you're not going to be able to get a machine to do. I wrote about this in my book, and it started as a dinner conversation I was having with a bunch of people.

It was one of those weird dinners. It was Yuval Noah Harari, who wrote Sapiens, was there, and a bunch of other people. And we were talking about some of this stuff and i i made this assertion that i believed at the time that uh that this group of people made me rethink which is like oh i i don't think that um a machine will ever be able to compose and perform music in a way that gives me the same emotional response that Murray Pariah gives me when he is performing Chopin's G minor ballad. And they were like, yeah, why do you believe, they just sort of pushed back hard on the assertion. And they probably

were right. You could build a system i'm guessing that would produce music in a feedback loop like you could sort of put a bunch of you know ekg probes on you uh and like you could probably have the system in a closed feedback loop give you the same uh like autonomic response like the the you know the shivers up your spine or goosebumps. But that to me still is qualitatively different than what I get when I listen to this performance because it's like this human composer from a different time who is trying to communicate something to everyone else. And then this performer from another different generation and a set of experiences who takes this work and interprets it in a way that no other performer interprets it that produces this emotional response in me.

And like, part of the beauty of that is that, you know, it's this stacking of human connection that has nothing to do with... Which, of course, like, of course, like, the definition of that is, I want to stack a bunch of human connections up. Then you can't do that if there's no humans in the stack, right? Of course not, by that definition. One of the things I really like, though, somebody, I forget who gave this example, it was someone who was talking to a bunch of students and they said, do you think it's possible to build a computer that's roughly the size and shape, roughly the size of a human

brain with all the capabilities of the brain. And the students mostly said no. And then he said, well, what did you use to come up with that definition? Well, you used a computer that's roughly the size of the human brain with the capabilities of the human brain, right? So it's like, there's a little bit of tautology in there

of like, you know, to your point though, it depends on the definition. Like, I think you're right. You can't replace humans by something that's not human definitionally, right? To your point, though, it depends on the definition. I think you're right. You can't replace humans by something that's not human definitionally, right? If part of the meaning and the emotion is about the human. Well, and look, I think the single most important thing for people to get through their heads is it is not necessarily a goal of the AI work. Certainly not the goal of the AI work that we're doing or that I'm doing is to replace humans. So you even look at something like GetUp Copilot, the reason that you build an AI that helps people

turn natural language expressions of a thing that they want code to do into actual code is to help you be able to get to the more important parts of your job more quickly. We did get a very nasty letter early on in writing, doing rightly from the medieval scribes guild saying that we were replacing scribes by doing this word, newfangled word processor thing. I'm kidding. But like, you know, of course, like nobody cared, right? Like we had long since replaced them and we, we understand that people need to write their own stuff and it's awesome.

It makes you better. Like the tool makes you better as a person, makes, makes there lots of more value in the the world which is what's going to happen with these yeah tools too right it unlocks creativity it it does a whole it it does the thing that all of these tools uh eventually do and like the trick is just sort of managing the transition i think right so um we are at time i could go on for i knew we would not get through any of our all of our questions man like yeah but but so i do have one last question so i ask uh i ask everyone uh on the podcast at the end um outside of their work and most of the people we talk to their work is a thing that they find fun but like outside of the things that you get paid to do, what do you enjoy doing in your free time?

I do lots of different stuff. I'm kind of a maker, as you are. So I got really into fountain pens for a while and making fountain pens. I got really into old-time music. I played bluegrass mandolin and Celtic mandolin and old-time mandolin. So I go hang out with people. That's really been fun. I took a mandolin with me all over the world. It's a little travel mandolin. So I played bluegrass on a stage in Tokyo and played bluegrass with some folks in Sydney and played actual Irish music in Dublin. So that's super fun. I mean, it's all kind of connected to me. I like cooking. I like making

stuff. I kind of like getting into the world and doing things. So I have too many hobbies, actually. Like I could answer that. I could spend an hour on that question. Yeah. I reject that assertion. I don't think that there's anything like, yeah, there's no such thing as too many hobbies. But that's, yeah. You know, you got me going on leather working recently. So that's kind of, yeah. I look at guiltily every once in a while i think i should do something with that oh you should totally do something with that cool well thank you so much for taking the time to chat with us today this has been a great conversation and we'll have you on the show yeah super fun but have been looking forward to it. So it's great. All right. That was a super fascinating conversation with Sam

Scalace. So one of the really interesting things you two were talking about, you know, you're getting into AI and one of the things that was said was like that AI, what's so interesting about it is that it's not that it thinks like us, but that it doesn't. Could you expand a little bit more about that? One of the things with AI is we're just seeing it more and more over the last handful of years as being an assistive tool, like a thing that helps us with the things that we're just not great at. And, you know, this has been sort of true with most of computing technology. You know, CPUs are better than human beings at arithmetic. Like, their computer algorithms are better at dealing with certain types of complexity

than human brains are. And, like, this is certainly true with AI. There are things about human creativity and intuition and all of these things that we actually take for granted that are very hard for AI systems to do. Whereas there are all of these things to do with complexity that AI is getting increasingly good at doing. And so the thing that I think we ought to be doing over the coming years is like just sort of leaning more into those differences than trying to, you know, build things that are exactly like a human brain. Because like, basically what we want to do is build these tools that can help us offload all of the repetitive, irritating, cognitive things that we're doing. So,

we can go spend more time on the things that are uniquely human and special and rewarding and fulfilling. Yeah. No, and you kind of, and you're mentioning you wrote a journal article kind of to the same point, right? Yeah. Yeah, I mean, so the journal article was in the American Academy of Arts and Sciences, did a special issue on AI and society. And my article was on a framework for thinking about AI as a tool for assisting us with cognitive work for picking heavy things up, for moving heavy things from one place to another, for doing this sort of rigorous, heavy mechanical work over and over and over and over again. It is essentially AI as a new type of steam engine. And we don't quite understand everything

that it's doing yet, the same way that we were inventing these machines before we had the physics of thermodynamics right um like we're slowly figuring out exactly what the science of cognitive work is and like how you measure it and how you sort of see whether you're making progress towards making it better and better and like you're building the right machines to go do it um but, you know, that is, I think, the right framing for thinking about it. It's like they're just tools, like GitHub Copilot, for instance. Yeah. It's just a tool. Like you, you know, instead of having to go look up a whole bunch of stuff, like you just

sort of tell the machine like what you're trying to accomplish. Right. Like it writes some code for you. Right. Right. It's, it's just a tool. I mean,

that's why it's called a co-pilot. It's, it's not, you know, the pilot, it's not doing it for you. It's, it's,

it's assisting you. You know, you're exactly right. One, one other one kind of final thing I wanted to touch on, you know, Sam kind of mentioned, like,

he feels like we're on like the cusp of like, kind of like this, this big fourth kind of like wave of, of advancement. Sounds like we're on the cusp of this big fourth wave of advancement. Sounds like he can just taste where we're going to be. Are you on the same wavelength with that? Oh, yeah. Very, very much. and where we're likely to be over the next handful of years than at any point in my career as a programmer and computer scientist and like i'm 50 years old so i've been doing this for

almost that sound like my grandpa now i've been doing this for almost 40 years uh now like i wrote my first program when I was 12. And yeah, so like I've lived the three big revolutions that we've had. So you had this transition from computers being scarce to like everybody has one. And I went from information being scarce to you have the internet and information is no longer scarce. And you went from computing being sort of physically tethered to being physically ubiquitous, your smartphone. I think what we're about to have happen now with AI is another transformation where the thing that's hard right now is technology is brittle. It is super challenging to use still.

Things crash. User interfaces are complicated. You have to have five programs come together or five apps to do the thing that you want to do. And I think very soon now we will have co-pilots for everything. Yeah. You will have AI that can just sort of listen to you and hear what it is that you're trying to accomplish, and then can do all of that management of complexity for you to get your thing done. No, yeah. I mean, I think it's exciting.

And I mean, I think, you know, Sam had mentioned this too. Like when you think about just how quickly all of this has happened and how much quickly, you know, how much shorter distance there is kind of between these different waves. Like I think that that's really exciting too. So I'm glad that we have people like you and Sam who are thinking about these things and are leading teams and doing research and all this stuff because that gets me excited, right? And I'm sure that gets a lot of members of our audience excited as well. Yeah, it's a good time to be in tech.

All right. Well, that is all the time that we have for today. Huge thank you to Sam Scalace for being with us today. If you have anything that you'd like to share with us, please email us anytime at behindthetech at microsoft.com. And you can follow Behind the Tech on your favorite podcast platform, or you can check out our full video episodes over on YouTube. Thanks for tuning in. See you next time.