Microsoft, Word, power Point, Excel. They are fixtures of office life everywhere, but it's almost thirty years since Microsoft released
the iconic Office ninety five. In the space of just a few months, though, the company that once gave us the likes of Clippy the interactive paper clip, has suddenly become one of the most important players in the rapid rise of artificial intelligence, and it all has to do with its collaboration with open AI, the creator of the wildly popular chat GPT that seems to be taking over the world and that some people fear actually will take over the world.
If you went back a couple of years and you asked people at Silicon Valley, like, who's in the lead in AI, they would have said Google. And I think everyone agrees now that Microsoft is either kind of like in the lead or has at least gotten to parody with Google.
I'm West Kasova today on the Big Take. Bloomberg BusinessWeek writer Max Chafkin is here with the unlikely story of how Microsoft, of all companies, is making itself cool again. Max, We've obviously been talking a lot lately about open AI and chat gpt, all the different versions of AI. But how does Microsoft enter into this because they are really a central player in this whole story.
Yeah. Absolutely, when you think about these kind of crazy developments of the last i don't know, eight nine months, the release of chat gpt and then this kind of frenzy of activity related ai. You know, one player, as you said, one of the key players is open ai, which is the startup that created chat gpt. But basically an equally big player is Microsoft, which is the biggest investor in open Ai. But that even kind of understates
how important Microsoft is to this startup. Microsoft is the largest shareholder, It owns something like forty nine percent of the company. It is the sole provider of computing services. It runs all the computers that open ai uses. It's also the main commercial partner. It's the way that open
ai gets revenue. And you know, as we wrote in the story, I think it's probably useful to think about open Ai not so much as a startup, but as a very successful, you know, quasi independent subsidiary of the largest business software company in the world, the second largest company in the world by marketcap.
And how that partnership came to be as the subject of this big story you've written, and it's really interesting because Microsoft is huge, we all use their products, but I don't think people have thought about Microsoft as a cutting edge, innovative company for a really long time.
When my co writer Dina Bass and I were in Redmond in Washington near Seattle, on the Microsoft campus, you know, they were talking about how amazing it was, and they were saying, you know, it's like Windows ninety five all over again, at which I loved because, first of all, to most people outside of Redmen that is like an insult, but if you're there, it's the highest compliment because that is really the time when Microsoft was on top of
the world. And since then, you know, they went through a very sort of dark period in the sort of two thousands. Ever since Sati Nadella became CEO of Microsoft, they've been on this kind of turnaround trajectory. And some of that has had to do with business decisions that he's made, you know, having to do with basically playing nicer with other platforms like Apple, but a big part of it is this AI thing and especially the deal
with open Ai. And really this all started at Sun Valley, which is this annual conference organized by Allen and Company. It's an investment bank it's in Idaho, and it's essentially billionaire summer camp. Billionaires go there to hang out out and make deals. It's the one place you see all these guys like looking sort of relaxed. If you see a picture of a CEO wearing a lanyard, it's almost definitely a Sun Valley picture, you know. And Altman, Sam Altman,
the co founder of open Ai. You know, he's a guy who goes to Sun Valley. He is a young guy. He's only about forty, but he's sort of been in Silicon Valley for a really long time. He started a company called Looped, which was like a mobile startup. He ran y combinator, you know, big deal kind of investment firm. He's sort of everywhere. He's friends with everybody. And at Sun Valley, he and Nadella run into each other and
they get to talking. And as we write in the story, you know, a couple of things became clear in this conversation, which then continued and culminated in a deal in twenty nineteen, one of which is that Sam Altman has sort of
these really big ideas, sort of famously big. He thinks that AI could potentially end the world he thinks that he is creating artificial superintelligence that's potentially you know, lead us to a life life of leisure where there are no jobs, or there are very few jobs, which could have huge societal consequences. But the thing I think that appealed to Microsoft is that open ai was actually making progress towards commercial products. Microsoft had invested essentially huge sums
in AI, and they'd made a lot of progress. They'd written a lot of good research papers, you know, as Nadella and the other Microsoft executive saw it, but they hadn't really had products, right. It was a lot of really good ideas, but kind of disorganized. And in open Ai they saw something you know, much much more focus. And that was the thing I think that most appealed to Microsoft. That's why they invested in twenty nineteen a
billion dollars in the company. And that's kind of like the beginning of this relationship.
And where did that money go? How did they then use it to rapidly expand this platform.
Well that's what's so interesting because we talked to Sam Altman about this and we said, why didn't you go with a normal you know, venture capital firm, And he said, the reason is essentially that this kind of technology, the thing that they were building, right, it needs these gigantic computers, these very very specific configurations of servers with very specific
chips that cost enormous amounts of money. We're talking hundreds of millions, if not billions of dollars just to train the model, just to create the model that underlies chat GPT. And that's even before you start running it, you start answering people's questions. And so the Microsoft deal, right it gets reported as a billion dollars and since then, as we write in the story, Microsoft has put something like
thirteen billion into the company in total. But a lot of that money, I'd say probably most of that money goes right back to Microsoft because open Ai has a deal with Microsoft where Microsoft is their cloud provider, and that's the big cost here. So Nadella writes Sam Altman a check. Sam Altman turns around and starts buying these you know, basically time in Azu, which is the Microsoft
cloud computing business. So that kind of works for everybody, right, because on one hand, it seems like Microsoft is is outsourcing this really important thing. It's sort of turning over a really important piece of intellectual property to an outside company. On the other hand, what they're doing is they're building this Ai supercomputer, that's what they call it, and that is Microsoft intellectual property, that's theirs. They're obviously going to continue to sell it to open Ai, but they can
also sell it to other people. And as these large language models, these chatchept like services take off, they have the potential to make money not just from open Ai, not just from open ai users, but potentially from everybody.
And Max, how does OpenAI feel about this arrangement.
Yeah, well, we talked to Sam Altman about it and he feels good about it. He feels like it's the only way that he can do what he wants to do.
And you write that the critical I guess breakthrough was something called transfer learning. Can you describe what that wasn't why it was such a big deal.
Yeah, So, if you think back to twenty nineteen, the hot thing was dryverless cars. The way you train a driverless car is you basically have a bunch of people sitting in rooms and they're looking at pictures of road signs and they they're tagging them. They're saying, that's a roadsign, that's a stop sign, that's a speed limit sign. You feed that into the machine and you try to get it to identify the sign. So you're taking these like really specific pieces of data and you're asking a computer
to identify them. What transfer learning is is instead of having really specific data, you just give it all the data, right, You just give it like the entire Internet. That's what open ai I did, and you try to get it to, say, finish a paragraph. And the idea is, if you can do that, it might be able to do other things because in addition to English on the Internet, there's also
a lot of computer code. So the first thing that happened with open ai and with the Microsoft partnership is they figured out that actually, this technology that can complete sentences for people can also complete computer code. And that's not because they asked it to be able to do that. It's just because there's a lot of computer code on the Internet, and the same basic process that can be used to generate English can be used to generate you know,
C plus plus. And so that's the idea that you could start with something like just write sentences and then you could say, I don't know, design a road plan or design an app or when you talk to the like apocalyptic people would be like design a plan to take over the world or something like that. And that's what the people who are really worried about open ai and similar technologies are scared of.
And so the notion is instead of trying to teach these models to do something specific, you kind of teach them to do everything, and in teaching them to do everything, they can also do the specific thing.
Yeah. Absolutely, And so an example of this so the Microsoft deal, Right, they started getting involved with open ai in twenty nineteen, and this partnership broadens. Microsoft starts incorporating the technology to its products. It invests more money, as I said, billions of dollars in open ai, and a
lot of people are really excited about this. But there are some skeptics, you know, one of whom is Bill Gates, you know, the founder of Microsoft, who is very interested in AI, and for a really long time, Gates was skeptical because he saw that it made mistakes, which anyone who's used open ai or chat, GPT or any of
these services knows. Right, Gates would be talking to the bot, right and it would say somebody's in Cleveland in one sentence, and the next sentence, they'd be in a different city, you know, Seattle, and it was like the bot didn't know what was actually going on. And what he said to Sam Altman was that I will believe this is good if it can pass the ap bio exam. He saw that as a test of does this thing actually understand?
And I talked to Altman about this because I was wondering, like, so, did you like feeded a bunch of bio exams to try to teach it? And he was like, no, it's just that there's a lot of biology on the internet.
And they had this demo at Bill Gates' house. Bill Gates lives in this you know, palatial you know house on Lake Washington in the Seattle area, and the Open Ai people showed off this early demo of what was called GPT four, the latest model, and it was able to pass the bio exam, including the essay portion, and so Gates was kind of impressed by that. But the thing that really kind of brought Gates and the other Microsoft executives around, it's the same thing that impressed everybody else.
It's that after the demo, the open Ai CEO Greg Brockman got up on a keyboard and just started asking questions, said what do you want to know? And Gates and the other executives just started throwing things out, trying to stump it. Gates, you know, is really interested in public health, interested in, you know, issues of childhood illness, says, how would you talk to the parents of a sick child? The open ai bought, you know, spits out a response and Gates looks at it and says, you know, this
is more empathetic than I would be. And for him, that was the moment, right, that was the moment where he thought, Okay, actually this is ready to be a product. This isn't just some kind of research experiment.
So Bill Gates is now on board. What happened next, that's after the break So Max, they persuaded Bill Gates, who even though he's not on the board anymore, is still a key advisor as the founder of the company.
What happens then, I mean, what's crazy here is how quickly this all happens. So that meeting with Gates was in summer of twenty twenty two, so less than a year ago. As OpenAI and Microsoft start negotiating and as the partnership deepens, A lot of this falls on this Microsoft executive named Kevin Scott, who Dina Bass, my co writer, and I spent a lot of time with for this story.
Kevin Scott, He's the chief technology officer. He's a really interesting guy and sort of an odd duck in this world of basically Stanford and Carnegie Mellon and cal Tech PhDs. He's from rural Virginia, essentially from Appalachia, went to the
University of Lynchburg, basically small local Christian college. Eventually kind of works his way, you know, into a UVA PhD program, winds up at Google, is a huge success at Google, and you know, eventually finds the way to become the chief Technology officer of Microsoft, where Nadella basically makes it his mission to kind of get all these AI programs that are kind of all over the place and put them in some kind of order. So the funny thing
about Microsoft has done some really great AI research. They also have had some sort of infamous AI products. I think probably the first chatbot that anyone ever used, or most people used, was Clippy. Most people encountered Clippy, you know in Microsoft Word where you'd start typing something, you type the word deer, and then all of a sudden, this little, you know, anthropomorphic paper clip with gigantic googly eyes would come there and sort of like mansplain letters
to you. It'd be like, well, it looks like you're writing deer, but do you need help writing a letter? And he would sort of essentially interrupt your work, and it became this kind of like gag. There are all these funny memes on the Internet of of you know, Clippy, you know, interrupting things and just kind of like the weird weakness of what otherwise is like an amazing piece
of software, you know, Microsoft Word. People kind of make fun of it sometimes, but like, right, it's one of the most successful piece of software of all time, and they just stuck a talking paper clip in it for no particularly good reason. The other one Microsoft chatbot AI thing was this thing called Tay, which happened in twenty sixteen.
Tay was sort of an update to Clippy where they wanted to have a bot that would sort of learn from Twitter and would sound like a typical teenager, and within about i don't know, twenty four hours, Tay had been deluged with kind of what you might imagine is on Twitter. She went from being like a nice, helpful teenager to just like sounding like the most obnoxious person on you know, on four Chan or on Reddit or whatever, and they had to pull it.
What does Microsoft say about Clippy and Tay when they look back on their early experiments with this technology.
From Microsoft's point of view, you know, Tay was done in by people who were manipulating Tay, sending it hate speech, you know, trying to get it to say the worst possible things. And so Kevin Scott has kind of the weight of all this kind of history of Microsoft's past bad efforts, as well as this research program that was good but just kind of unfocused.
So now Kevin Scott has this mandate to take this very powerful AI chat generator and make use of it, and he's got a lot of money to do it. And what did they.
Do before this the meeting with Bill Gates? They had had one product, which was GitHub. So GitHub is a product for software coders. And they'd sort of figured out that at GPT, or really GPT three, which is the underlying technology, that in addition to being able to finish sentences,
it could finish code. So they created this thing called Copilot for GitHub and it's like auto complete, you know, autocomplete on your text message where you're typing a message and it sort of tries to guess what you're about to say. It was like that except for software code, and it was pretty successful. They were charging money for it. It costs, you know, between ten and twenty dollars a month per person, so real money. And that was it. That's all they had. And for the next one, as
part of this big push, they chose Bing. It's kind of like the Clippy of search engines, right, it's this like kind of goofy, unsuccessful bit of a head scratcher of a competitor to Google. You know a lot of people like Bing, but it's never really taken off. I think partly because Google just dominates the market, and I think they just thought, well, search is a really big business.
These chat bots are potentially pretty useful for search because you could ask it instead of searching for a restaurant, you'd say, what's the best restaurant? And being the stakes are low, right, if it's terrible, they're not that many BING users out there to be up in arms.
So what happened when they put this technology into being in February.
You know, Sati Ndella has an event. He gets up on stage and he shows off this technology and everyone is really impressed. Right. Bing is based on the most advanced version of open aiyes model, it's called GPT four. During this demo they showed it you know, writing letters and planning menus and things like that.
So you're not just using being to say, you know, find me this restaurant's website. You're actually asking you questions and then it retrieves it. So it's searching for answers, not just for results.
Yes, and this is really exciting. I mean, first of all, it's exciting because it could be potentially easier, right, instead of spending you know, an hour on Google searching for restaurants in Paris and hotels in Paris and activity in Paris to make my Parris itinerary, you could just ask a bot to make you one, right, and they'll do it. They'll give you something, you know, within I don't know, ninety seconds or whatever, And that's really exciting. Potentially makes
things easier, I guess if you're traveling. But from a business perspective, that's like a huge moment because it could potentially you know, break a lot of the search business because if you don't spend that hour on Google, if you're not clicking all these links, Google is not able to put ads in front of you. So you know, immediately you saw Google stock fall, you saw Microsoft stock go up, and then people started playing around with Being.
You had all these reporters and Twitter users spending time with Being and getting it to do some weird things. The most famous one is Kevin Russ, who's a tech columnist of the New York Times, had this like two hour conversation with Being about love where Being became an alter ego called Sydney. I mean, it sounds insane when when I say it and told him to leave his wife and said no, no, you love me and got
really aggressive and just kind of flipped out. You had another sort of famous tech analyst, Ben Thompson, got bing to adopt a different alter ego, venom As, kind of like an evil version of Sydney, kind of fantasizing about how you would extract revenge on his enemies. Being told the Verge, you know, it's a tech publication that it
was spying on Microsoft employees using their webcams. Like, so not only making something up, but making something like you know, potentially pretty damaging to Microsoft and so on.
And Max, what is Microsoft say about these encounters you described that people had with being so.
From Microsoft's perspective, the problem wasn't the software, it was the people who were using it. They were, as CTO Kevin Scott said to us, basically trying to goat it in to say the worst possible things. They just they're moving ahead. They're not actually stopping. I mean, they made some tweaks to how BING works, but it's not like they're not going to bring this stuff to office. They're
still going to bring it to office. They're going to bring it to who what is like probably the biggest and most important piece of software in the world, which is you know, these productivity apps that Microsoft sells in other words, Microsoft Word, Microsoft Excel, Microsoft PowerPoint, And this is already happening like inside of certain companies and stuff.
You will be able to be, you know, working on a chart and Excel and you'll be able to ask the copilot to make a visualization of that, or you'll be able to, you know, as I said, like have it essentially create rough versions of your PowerPoint slides. For a presentation, and the idea here is not to have it sort of replace your work. It's not going to do your presentation for you, but it just might make
it a lot easier. The way that a Microsoft executive who works on these products told us is it's like it might do maybe half of the work right. You're still going to have to keep an eye on it. You're going to have to make sure it doesn't make any huge mistakes, it doesn't make anything up, and it's not going to do everything, but it hopefully will make things like easier.
When we come back. Microsoft has big future plans for AI, so now max. In a really short amount of time, this went from novelty to a core part of Microsoft's business and has really leapfroged the company in a lot of ways ahead of some of its biggest competitors.
If you went back a couple of years and you asked people at Silicon Valley, like, who's in the lead in AI, they would have said Google. And I think everyone agrees now that Microsoft is either kind of like in the lead or has at least gotten to parody with Google, which is huge achievement because Google Larry Page, Sergey Brin. I mean, they have been really focused on this for a very long time. We know about the
Google self driving car. There's this thing called deep Mind, which before open Ai was the hot you know, research lab, which is owned by Google. And on the business front, I mean, they are looking like they are in a very good position. This isn't to say that this couldn't like all come crashing down or I think it's very possible that these technologies just don't work as well as they're hoping and aren't as useful as people think. But here's the thing. Microsoft is very very good at extracting
money out of software. Right for basically two decades, there have been free alternatives to Microsoft Office, Word Excel. You can get a free version of them. And you've been able to get a free version of these things for decades, and yet they're pulling in, you know, something like forty five billion dollars a year on licensing fees on these technologies. And so their plan here with these chatbots is essentially
use them to drive more revenue to that thing. So from now on, if you're you're gonna pay your ninety nine dollars a year for your office subscription, if they price it like GitHub, it's going to be you know, a ten dollars a month. Add on. A lot of people and especially a lot of businesses are going to feel like they need this and they're going to pay for it. And Microsoft, as I say, is very good at convincing people to pay for its software. It's kind
of like what the company is almost best at. The other thing that they have is this supercomputer that I brought up, where it's not just that Microsoft can put open AI in all its apps, it's that the entire world of business is attempting to integrate AI into their products, not just Microsoft products, but other products, and they're going to need computers to do that. And Microsoft right now is in the best position and position to sell those services,
the kind of cloud services to everyone. And so when you add those two things up, you're talking about a huge amount of revenue. We mentioned this in the story, but one analyst has a model, which I think is debatable, but this is like a legitimate model where he's talking about by twenty twenty seven, somewhere between ten billion at the low end and one hundred billion at the high
end of additional revenue to this company. So that's like if they get to one hundred billion, that would be like it's like adding three netflixes to a company that's already the second largest in the world by market cap. I do think it's important to say that the other companies aren't standing still, right. They see the same opportunities
at Microsoft. See so Google is frantically trying to put this technology into their apps, and I think there's reason to think Microsoft is in a better position because of the commercial factors I brought up. But you could imagine a situation where in a year from now or two years from now, people are a little less excited about it all of a sudden, you know, you maybe you don't want to spend extra money for a chat GBT thing, where then it becomes a little more challenging for Microsoft.
But even then, I think it's still going to be a substantial amount of revenue for this company, just because they have such a big footprint and they're so far ahead of everyone else. Max.
What does Microsoft say about these growing concerns about sort of doomsday scenarios having to do with AI.
They do not see it as a thing that's going to end the world. They see a computing platform. The Microsoft people right, are not super focused on the end the world questions, which sort of makes sense when you're talking about like a copilot for Excel. It's hard to imagine like spell check taking over the world, And so what they're really interested in is sort of like productivity.
How can this make workers better? And what they say about these kind of other concerns, they say, like it's copilot, it's not autopilot, and it really is not that useful as a replacement. It's only useful as a thing that
can like sort of help you along. I think there are questions about how applicable that is, whether it actually ends up happening, whether these technologies get used responsibly, But that's kind of their point of view that like, as long as humans are in the mix, a lot of these risks are not that great.
Max, thanks so much for coming on the show.
Thanks for keeping this human in the mix.
Thanks for listening to us here at the Big Take. It's a daily podcast from Bloomberg and iHeartRadio. For more shows from iHeartRadio, visit the iHeartRadio app podcasts, or wherever you listen, and we'd love to hear from you. Email us questions or comments to Big Take at Bloomberg dot Net. The supervising producer of the Big Take is Vicky Vergalina. Our senior producer is Katherine Fink. Rebecca Shasson is our producer. Our associate producer is Sam Gebauer. Raphae alum See is
our engineer. Our original music was composed by Leo Sidrin. I'm west Kasova. We'll be back tomorrow with another Big Take.