Bill Gates sat down with Michael Miller to discuss his role in the tech world over the past 30 years, and what the future holds for him—and for us.
by Michael J. Miller
Buzz up!on Yahoo!
At the end of the month, Bill Gates is stepping down as chief software architect of Microsoft, and retiring from his day-to-day role at Microsoft, the company he co-founded and led for most of the past 30 years. Michael J. Miller, former PC Magazine Editor-in-Chief, sat down with him to look back at how the computer industry has evolved over that time, and where it is headed.
Miller’s own thoughts on Gates and the interview can be found here.
Part I: Hits and Misses
Michael Miller: From a technology standpoint, what would you say is your biggest success? Was it making the PC popular or the graphical user interface?
Bill Gates: The most important thing is the creation of the software industry and the importance of having this platform that anybody can write to. That goes back to Microsoft Basic on the Altair, the Commodore PET, TRS-80, Atari 800, Apple II—building up a library of programs that people use to get something done.
There was no software industry before the PC came along. The whole magic was that computers became so cheap, and you need a lot of software, so people can sell software in volume and price it quite reasonably. That magic of high volume, low price just wasn’t possible in an era with a very modest number of very expensive computers. Most companies that did software did it as a sideline before we came along.
Basically we take that idea of a whole software industry, and we ourselves do some of the big things. We evolved the platform of the software industry from Basic to DOS; from DOS to Windows; from Windows to .Net Internet; to modeling, cloud computing, natural user interface…
The platform is changing because of the hardware improvements and the kind of scenarios that are possible. It’s just so different. We are about building a software platform and a software company.
M: Over the years, you have also talked about a number of technologies that you thought were going to be successful but haven’t reached mainstream appeal as much as I think you thought they would: SPOT, tablet PCs, speech recognition, stuff like that. What do you think it is: a software or hardware problem, or just society? Why did some things become incredibly popular and some things stay more or less as niches?
B: Look what was written down from when Paul and I started Microsoft. Half the things we dreamed of as scenarios for software to solve are still in front of us.
Natural interface including speech, and the kind of inking that comes out on the tablet. My prognosis, you can call it stubbornness, is simply that it is not ready for the mainstream yet. We have to keep improving the software and hardware. But I have no doubt that the current way we interact—which is overwhelmingly a keyboard and mouse way—in the next decade will be changed deeply. Not that it will go away, but it will be supplemented by speech, vision, ink-type things. And this is the kind of issue where Microsoft gets to put billions of dollars behind those beliefs.
The tablet: it’s taken off in some niches. There are millions sold a year, but not tens of millions. My belief is that we will get to hundreds of millions. So we are a factor of 100 away from what I wanted to happen and I believe will happen, where every student instead of having paper textbooks has this great device connected to the Internet that allows them to edit, create, record voice, browse, in this very deep way.
M: So why did it take so much longer than a lot of people thought it would?
B: Well, everything [took] longer. Why did graphics interface take so long? Why did connecting computers in a network take so long—seven or eight years? Then you get to the hockey stick point.
Take graphics interface: Steve Ballmer and I would fly around the country and do these seminars about graphics interfaces, people would say “No, it’s too slow, it’s too hard to write the software,” and it was very weird that we were pushing it. Then in a period of about year, it became so obvious that nobody discussed it anymore. So it is one of those weird things where you battle for so long and then there is no period where people say, “Gosh, you were right.” All that happened was that immediately they say “Why are you still talking about that, of course that was right. Now we want to discuss your next thing.”
The early 128K Macintosh almost failed. The only reason that thing survived at all was because some craftsmen at Apple and Microsoft were able to squeeze a few programs into the thing and kept it going enough. Then the 512K Mac came along and finally they got that thing to reach its critical mass. In the meantime, 95 percent of the companies that showed up to write software for the Macintosh went bankrupt, just as happens with all these things.
So the things that did happen: the move to PC computing as the mainstream office device; the move to graphics user interface; the move to PC-based servers taking over the data center. I remember there was a Fortune magazine article saying that some workstations had a higher growth rate than PCs did, of course off of some tiny little volume. What happened is we in our high-volume/low-price way created PC-based systems that did everything those workstations did. Not only did we take over the specialized functions, but we became the standard device for many of the jobs in the economy.
All these amazing things happened for the information economy, based on this software platform. And those things took time. There is nothing that was overnight.
The ones that are still in progress but are partly there are Tablet PCs, Internet TV, natural user interface. Those are three of the things you can say that they have failed to date. If you take my optimistic projection of how quick those would take off, we are well behind.
But the nice thing about Microsoft and the situation that we are in is that we can afford to be early. We can have our work sitting there ready to just keep being improved and improved.
Take Internet TV: until someone like AT&T; comes along and says “OK, we want to do a video platform that is better than cable, that is personalized, targeted,” it will only be another two or three years before it will be a common sense that when you watch a new show you can skip over the things that aren’t interesting and see more about the stuff you care about. When that finally happens, will anybody say “Oh, what a brilliant idea”? No, they will say “Of course! Why was it any other way?” Because it is insane that it’s any other way. Broadcast was this weird hack that was created to deal with very finite ability to get bits into households. And now the rules, at least for urban households in rich countries, are in the process of changing.
Things take a long, long time, but a lot of those things actually will happen. Many of them in terms of shopping patterns and behaviors, you almost take a generation in terms of how people are used to doing things before you get the full impact. No matter how good the technology gets. Mobile phone interface and natural interface will accelerate it, but all these things have the technical issues of the hardware and software, the price, and the behavioral thing around them.
M: When you go back and look at the whole Internet revolution, what did Microsoft do right and wrong?
B: The key thing we did right is we got more than 100 million PCs out there ready to be connected.
Then it was set that at some point the cost of the connectivity would get to the right point and some protocol would be picked. The fact that it was the ARPA protocols that came out of the university environment, could you have known that? I used those protocols a lot before I started Microsoft. In one place, one niche became where those protocols exploded, and they were particularly good protocols, but any of them would have worked. So getting all the machines out there and creating a volume environment meant the connectivity was guaranteed to happen.
Knowing what period it would hit the hockey stick, we didn’t know. For many years, “the year of the network” was declared by many people, and we were supportive. It was when we were going up and college interviewing that we saw within a particular environment—universities like Cornell—that a threshold was crossed in terms of expecting to put course outlines online, ordering pizzas online.
Now if you take the population as a whole, you have different groups with different adoption patterns.
M: As this was happening, you guys spent a lot of time developing sites, Internet tools. What did you do right and what did you do wrong with that?
B: So many things have to be learned as you go along. We bought some things that you have to look back now… As I say, in that party room, we were the sober guys over in the corner who had half a punch. You had valuations of $2 billion for software companies that had 10 guys. There was an ad-rotation company that had a valuation of $500 million, and I had guys working on ad rotation, who were saying “Can’t we have a fourth head?” If you had a certain kind of story, infinite capital went to you. That was very strange because we had been the people with the capital to back cool software, internally within Microsoft.
This whole concept that if you can get a piece of software so popular that it creates an ecosystem around it—that was our insight back in 1975. As that became recognized, people took it too far and didn’t think there could be many people going for that all at the same time.
We were caught in some of that craziness in the late ’90s. We bought a few things. I still believe in the Sidewalk vision: the sites for a city where you could see all of the events and merchants and plan things. It’s really going to happen more on the mobile phone. That concept was good. We created this company that we spun out for over $1 billion that became Expedia. It became about booking airlines, not about software. We created Expedia, Slate, Sidewalk, MSN. We bought Hotmail, we bought Link Exchange. I bet I can’t remember all of those crazy things we bought.
The thing we did well was helping businesses think about Internet and what they should do. That we did really well. Some of the more consumer-ish things and what would or would not catch on, we didn’t have the magic answer.
The search thing is the one that jumps out where you ask “Why didn’t Microsoft do a better job in search earlier?” Well, hey, we can’t do everything—we don’t expect to do everything. We do a lot and we have a longer time horizon than anyone else.
Say there is an ocean out there and some big wave starts up, there is going to be somebody who is right where that wave is and they are going to be surfing before you know it. [The Google founders] are bright guys and were there when it started. They weren’t even the first, but they executed very well and they took the Adwords stuff and they created a good market with it. Boy, did they execute well. They’ve stayed ahead and did a good execution on that.
One thing that never comes out is that the software business is bigger selling to businesses than it is to consumers. Microsoft is really in touch with what are the practicalities, how do you make workers more productive, what are the pains in an IT department, what does corporate site development involve. We have built up over the decades the real ability to have great ongoing dialog with businesses about how they do their software and what they do. We have a very strong position.
When you think about why information workers will be far more productive 10 years from now than they are today, I would say the ideas in Microsoft Research Labs. There is more there about interactive whiteboard and the Surface-type desk and the way communications will work and how modeling will let you express things. There are more of the key ideas about why workers are going to be more productive there than anywhere else. In fact, it’s hard to think about what is the other place you’d name, because most people don’t have that broad view of software in common for that information worker. The economic value in that is very high.
This is a strength that Microsoft intentionally developed and it took a long time. We were first about bottoms-up guys who ignored IT, then we had to have the balance. Now we have that tension to meet that constructive needs from both top and bottom more often, even from the beginning of the product. What we did with SharePoint, what we did with the .NET development platform—that took a long time.
This is the whole time that the Windows platform really scaled up in terms of performance and security. Take Windows Server; it is unbelievable what has happened. Today the things that are still on mainframe have nothing to do with performance. It has to do with the code we wrote. We don’t want to rewrite it. In terms of absolute price/performance, nobody does anything on mainframe, but there is still a lot because the code is still up there. The Windows-based server and the Unix-based servers using the same hardware, they totally have taken over everything except the code museum stuff. —next: The Power and Problems of the Cloud >
M: Everybody now is talking about software as a service (SaaS), cloud computing, and those sorts of things. How does the move towards those kinds of models impact desktop computing, which is clearly Microsoft’s legacy, the thing Microsoft is best known for?
B: There’s always been this question of “Where is computing being done, right next to you or far away?” And the more bandwidth and lower latency you have, the more flexibility you have about how you split that computer task. Time sharing had terminals where almost nothing was happening locally. Whether it was a character-based display, a 3270 or X protocol, everything but presentation was happening centrally. Then the PC swung it, before the Internet shows up, to where you’re doing everything on that local device and only the file store and in some cases, the database store, are done remotely, but you have most of the business logic as well as presentation, editing, and interactions done on that device. The beauty of that is you can work offline, you get great responsiveness, you don’t have to worry about the latency. Those of us who grew up with time sharing understand going back to timesharing, even with great capacity, is not that great.
Now you have more of a balance. HTML is back to the terminal model. When you browse a Web site, although HTML is way more complicated than most presentation protocols, it is a presentation protocol. Now you mix that in when you put active controls in or local script. All that AJAX stuff lets you now do some code execution. So it’s ironic that the good websites are the ones that aren’t using HTML, they are using local execution.
Now we are in a world where you can get the best of both worlds, when you call a subroutine, that subroutine can exist on another computer across the Internet. We now have tools for developers so they can call a service right across the Internet and they think they are calling a local subroutine.
Everything in computer science is to just write less code. What is the technique for writing less code, and its called subroutines. Everything that has ever been done—object-oriented programs, software as a service—it’s about taking this idea of subroutines and being able to use them broadly. When you want to draw a map, you say “That’s hard, a lot of data; I just want to call a subroutine.” Well now you can call Virtual Earth or Google Earth and get back the presentation in this great form. You don’t have to think about the data, the format. So we are taking subroutines to this next level and making that simple. Actually debugging the stuff, performance, making it work offline—there is still work being done on this.
In the extreme case, we can take somebody’s data center and run it for them on the cloud. All the issues about administrative, capacity, who owns the data, what happens when things go wrong, when people are getting error messages, that’s cloud computing and there is a lot of deep invention and work. I would say we are investing more in letting businesses use cloud computers than anyone is, and we have some brilliant projects that Ray Ozzie will be talking about more over the next year.
None of this means you don’t want local intelligence that is very responsive to you. You don’t want everything to just be a terminal-type approach.
People get confused. There is storage in the cloud, which is clear that your file should be up there and geo-distributed and backed up, and there is computation in the cloud. Those are both great, appropriate things, but the one that is without any tradeoff is to have the logical storage master up in the cloud. The one that you have to be careful of is what about computation, because computation is not free. And you have big problems with latency, offline, and scheduling that resource, which is a finite resource. But we are actually taking some pilot customers and moving huge parts of their data centers into our cloud where we manage it for them. Over the next couple of years, a portion of the data centers will start to move. Some people say data centers will move to the cloud very quickly, but I tend to think it will vary a lot
So you have two things moving to the cloud to be clear. You have stuff that could be done on the client, like storage where the master moves up and you just do caching. Then you also have server-based computing that could move into the cloud. Well that’s just a different data center, but it may be one that has the scale and pooling. With some of the early efforts, like the Amazon S3 stuff, it still forces you to write the program that understands there are different computers and how things work on that.
The thing we’re doing that Ray Ozzie will talk about later this year at the PDC is how you make it easy to write those programs that are high-scale running in cloud data centers in a way that you really understand what is going on.
M: When you look at operating systems, like Windows, what has gone right and wrong? What do you see that you need to add over the next decade or so?
B: Well, there is a famous quest of mine called integrated storage, where you have not just a file system but more of a very flexible object-type database: things like your contacts, calendars, favorites, your photos, your music—how you rate those things are stored in a much more structured environment. And so they can be navigated easily and move between applications very easily. And that hasn’t happened. It will happen as part of the move to cloud storage. We will get this extra storage structure. Say you want to move data between multiple phones, multiple PCs, TV, car. You don’t just want to move files, you want to move things that have more structure. So there’s integrated storage or unified storage that hasn’t happened yet, and that’s too bad. You see a little bit of it where Apple and Microsoft are doing string indexes in the background, but it’s only a partial step. It doesn’t give you the full structure.
Now operating systems have a huge role to play in natural user interface. Now we have taken Windows and put it into our Surface device, but how do you add the programming model, how do you interact with those types of things? There’s a lot that still has to be done there.
There’s a lot of work to be done in security. Right now the tradeoff is between the user having to see a lot of things that they don’t really understand versus choosing how promiscuous things want to be and what they want to run. We haven’t made the breakthrough that makes it easy for people to understand what type of risk they are taking for which actions, so they are just getting a lot of stuff that they don’t know how to answer. Even with all these great mechanisms, they can do things that are quite dangerous. It is not an easy problem to solve, but there is a lot to be done with that.
You know the whole thing with the operating system existing across devices, where I updated the operating system on this machine, I update it on this machine…. If you have a houseful of machines or multiple machines, you should just say, “Hey, I want this Adobe application on all them, I want this file on all of them.” That way you can just look at your machines and do things holistically. We are in the process of solving that, but that’s really a mess today.
With some of these things that aren’t done, the cloud stuff gives you a slightly better way of doing them. We can store your music rights, all of your preferences, software rights, we can store them in the cloud. Then when you buy a new device, if you’ve got connectivity you can pre-configure the thing. When you get a phone, it’s a lot of trouble today. Why should it be? You should say “Hey, I am Michael Miller. Make this thing like that other phone I have.” Even if it’s from a different manufacturer or has different software—for things like contacts and schedules, there is a mapping. You shouldn’t start like some newborn