WEBVTT - Satya Nadella Talks New AI Powered Windows, Fostering Competition

0:00:02.400 --> 0:00:06.800
<v Speaker 1>Bloomberg Audio Studios, podcasts, radio news.

0:00:07.240 --> 0:00:10.120
<v Speaker 2>Welcome to our Bloomberg television and radio audiences. I'm here

0:00:10.119 --> 0:00:13.600
<v Speaker 2>in Redmend, Washington with Microsoft CEO Sacha Nadala with some

0:00:13.680 --> 0:00:15.720
<v Speaker 2>big unveilings at Build today.

0:00:15.760 --> 0:00:17.440
<v Speaker 1>Sacha, thank you so much for joining us.

0:00:17.920 --> 0:00:21.080
<v Speaker 2>You've got a new version of Windows powered by AI

0:00:21.239 --> 0:00:24.400
<v Speaker 2>that basically means you have your own AI agent on

0:00:24.440 --> 0:00:28.520
<v Speaker 2>your own device. How will this AI powered hardware experience

0:00:28.600 --> 0:00:30.840
<v Speaker 2>be totally different from what we've seen before?

0:00:31.320 --> 0:00:33.080
<v Speaker 3>First, well, thank you so much for having me Emily

0:00:33.159 --> 0:00:33.519
<v Speaker 3>end up.

0:00:33.680 --> 0:00:38.159
<v Speaker 4>Yeah, I mean it's really a ground up, you know,

0:00:38.200 --> 0:00:44.559
<v Speaker 4>reimagination from Silicon So we have a new SoC that

0:00:44.720 --> 0:00:47.680
<v Speaker 4>has the NPU. So think about it as we've always

0:00:47.680 --> 0:00:51.200
<v Speaker 4>had CPUs and GPUs and now you have newler processing

0:00:51.280 --> 0:00:54.920
<v Speaker 4>units to in fact locally process these AI models. We

0:00:55.080 --> 0:00:58.200
<v Speaker 4>have forty plus AI models that are local.

0:00:59.480 --> 0:01:00.000
<v Speaker 3>On Window.

0:01:00.040 --> 0:01:02.960
<v Speaker 4>Those machines that then are being used by a variety

0:01:03.040 --> 0:01:06.000
<v Speaker 4>of experiences that we are built into Windows, starting with

0:01:06.120 --> 0:01:09.840
<v Speaker 4>this photographic memory feature called Recall, which is just phenomenal.

0:01:10.560 --> 0:01:13.760
<v Speaker 4>And of course Copilot is built in to Windows and

0:01:13.800 --> 0:01:15.800
<v Speaker 4>Copilot is just not an app, it's sort of a

0:01:15.840 --> 0:01:18.880
<v Speaker 4>shell affordance that's going to be there assisting you as

0:01:18.880 --> 0:01:22.200
<v Speaker 4>you be. Showed a demo of somebody playing Minecraft, and

0:01:22.319 --> 0:01:25.160
<v Speaker 4>you know me just sharing my Minecraft screen with Copilot,

0:01:25.160 --> 0:01:27.600
<v Speaker 4>and Copilot helps me finally me as competitive with my

0:01:27.680 --> 0:01:30.880
<v Speaker 4>daughter on Minecraft has always dreamed of, right and so,

0:01:31.440 --> 0:01:33.440
<v Speaker 4>and then of course the third parties, I mean, Adobe

0:01:33.480 --> 0:01:35.840
<v Speaker 4>showed how they're take bringing all of what they do

0:01:36.240 --> 0:01:40.760
<v Speaker 4>to Copilot plus PC's and many many more other developers.

0:01:40.760 --> 0:01:42.480
<v Speaker 4>So it's an exciting day for us to have a

0:01:42.520 --> 0:01:45.880
<v Speaker 4>complete rethink on the full stack of windows for the AI.

0:01:46.000 --> 0:01:48.600
<v Speaker 2>Ah, I bet you're pretty good on Minecraft just a hunch.

0:01:49.120 --> 0:01:52.040
<v Speaker 2>This new feature recall means your computer basically has a

0:01:52.080 --> 0:01:53.240
<v Speaker 2>photographic memory of.

0:01:53.240 --> 0:01:54.440
<v Speaker 1>Everything you do on it.

0:01:54.840 --> 0:01:57.320
<v Speaker 2>And the speed at which this is all happening is

0:01:57.440 --> 0:02:00.840
<v Speaker 2>just so fast. It seems so unpressed. How will this

0:02:00.920 --> 0:02:04.160
<v Speaker 2>really make things different for consumers and change our daily routines?

0:02:04.280 --> 0:02:06.760
<v Speaker 4>Yeah, I think if you think about the PC, it

0:02:06.840 --> 0:02:11.400
<v Speaker 4>has always been about you know, creating things right, which

0:02:11.440 --> 0:02:14.639
<v Speaker 4>is it's about being able to synthesize lots of information,

0:02:15.040 --> 0:02:20.520
<v Speaker 4>create new information, pass better judgment, gain insights, and act. Right,

0:02:20.560 --> 0:02:23.760
<v Speaker 4>that's what you sort of do on a PC. AI

0:02:24.080 --> 0:02:28.440
<v Speaker 4>is an assistant AI helps you, as a as an

0:02:28.520 --> 0:02:31.320
<v Speaker 4>end user, do anything right. One of the most beautiful

0:02:31.320 --> 0:02:33.720
<v Speaker 4>demos today was even this thing called co Create. Right.

0:02:33.760 --> 0:02:36.960
<v Speaker 4>I am in paint and I'm painting a nice spring

0:02:37.120 --> 0:02:41.080
<v Speaker 4>Seattle day and I want mountains with wildflowers, and I

0:02:41.120 --> 0:02:43.640
<v Speaker 4>can now have the AI assistant that painter and me

0:02:43.760 --> 0:02:46.280
<v Speaker 4>now can be inspired by AI. But ultimately I am

0:02:46.280 --> 0:02:49.160
<v Speaker 4>the one painting and I get that satisfactor like I feel.

0:02:49.440 --> 0:02:51.239
<v Speaker 4>One of the things that's been one of my coworkers

0:02:51.400 --> 0:02:53.520
<v Speaker 4>was telling me when her five year old is saying,

0:02:53.560 --> 0:02:56.840
<v Speaker 4>for the first time, give me designer time. That's what felt,

0:02:56.919 --> 0:02:58.880
<v Speaker 4>you know, it feels like in the late eighties and

0:02:58.919 --> 0:03:01.200
<v Speaker 4>the early nineties when we all got fell in love

0:03:01.240 --> 0:03:04.720
<v Speaker 4>with PCs, that was what really gods going, which is

0:03:04.760 --> 0:03:08.240
<v Speaker 4>the ability to use computers to create. And I think

0:03:08.280 --> 0:03:10.160
<v Speaker 4>this AI is just going to take it to the

0:03:10.200 --> 0:03:10.760
<v Speaker 4>next level.

0:03:11.000 --> 0:03:13.480
<v Speaker 2>Microsoft has had trouble keeping up with some of the

0:03:13.480 --> 0:03:15.359
<v Speaker 2>new cutting edge Max and you did do a lot

0:03:15.360 --> 0:03:19.080
<v Speaker 2>of side by side comparisons today. Have you narrowed the

0:03:19.080 --> 0:03:21.960
<v Speaker 2>gap today truly? And how much share do you think

0:03:22.000 --> 0:03:22.919
<v Speaker 2>you can take from Apple?

0:03:23.200 --> 0:03:25.320
<v Speaker 4>Well, I mean, I think today was a day where

0:03:25.320 --> 0:03:28.200
<v Speaker 4>we brought some of the efforts that we've had for

0:03:28.240 --> 0:03:32.840
<v Speaker 4>a long time all into focus and really brought cohesiveness

0:03:32.840 --> 0:03:35.080
<v Speaker 4>to them. For example, we've been on this journey to

0:03:35.120 --> 0:03:38.680
<v Speaker 4>bring I call it arm based PCs for a while.

0:03:39.160 --> 0:03:42.360
<v Speaker 4>Finally we now have the best ARM processors, which are

0:03:42.440 --> 0:03:46.800
<v Speaker 4>really the performance of the CPU itself is cutting edge.

0:03:47.440 --> 0:03:51.920
<v Speaker 4>Then you combine it with the NPU and these experiences.

0:03:52.080 --> 0:03:54.080
<v Speaker 4>I think we are now going to start just like

0:03:54.160 --> 0:03:55.600
<v Speaker 4>you know, when it comes to the cloud.

0:03:55.680 --> 0:03:58.040
<v Speaker 3>Let me take you another one, right, which is.

0:03:58.040 --> 0:04:01.680
<v Speaker 4>When we were We came from second and what became

0:04:01.800 --> 0:04:03.720
<v Speaker 4>a leader in the AI age on the cloud.

0:04:04.080 --> 0:04:06.440
<v Speaker 3>I hope on the PC we bring good competition. I mean,

0:04:06.480 --> 0:04:08.000
<v Speaker 3>Apple has done a fantastic job.

0:04:08.400 --> 0:04:11.920
<v Speaker 4>We now want to bring real competition back to the

0:04:11.960 --> 0:04:13.080
<v Speaker 4>Windows versus Mac.

0:04:13.640 --> 0:04:16.040
<v Speaker 3>When it comes to the aihe.

0:04:15.760 --> 0:04:18.880
<v Speaker 2>You've been doing a lot of unique deals to move quickly,

0:04:18.920 --> 0:04:21.679
<v Speaker 2>for example, this six hundred and fifty million dollar investment

0:04:21.680 --> 0:04:22.200
<v Speaker 2>in Inflection.

0:04:22.320 --> 0:04:24.560
<v Speaker 1>You hire the team, you hire the co founder, must

0:04:24.600 --> 0:04:27.440
<v Speaker 1>office Suliman. Are you at all.

0:04:27.279 --> 0:04:29.680
<v Speaker 2>Worried that regulators are going to crack down on investments

0:04:29.680 --> 0:04:31.599
<v Speaker 2>and aqua hires the way they have cracked down on

0:04:31.720 --> 0:04:32.120
<v Speaker 2>M and A like.

0:04:32.160 --> 0:04:34.600
<v Speaker 1>Do you wish you could be buying these companies outright.

0:04:34.680 --> 0:04:36.600
<v Speaker 4>I mean, look, I mean the regulators have to do

0:04:36.640 --> 0:04:39.320
<v Speaker 4>what they have to do to make sure that ultimately

0:04:39.360 --> 0:04:41.719
<v Speaker 4>there's competition, right, I mean, if that is the case,

0:04:41.880 --> 0:04:44.520
<v Speaker 4>then I think I think one of the fundamental things

0:04:44.520 --> 0:04:46.800
<v Speaker 4>that I think all regulators need to stay rounded on

0:04:47.839 --> 0:04:48.960
<v Speaker 4>is where is competition?

0:04:49.040 --> 0:04:51.400
<v Speaker 3>What are the sources of competition, and how.

0:04:51.240 --> 0:04:54.240
<v Speaker 4>To foster them? To me, that's what we're doing right today.

0:04:54.560 --> 0:04:57.960
<v Speaker 4>There is formidable amount of competition, right, whether it is

0:04:58.000 --> 0:05:00.800
<v Speaker 4>between the big players or the small pla and there's

0:05:00.880 --> 0:05:04.240
<v Speaker 4>partnerships are also a source of competition. So I don't

0:05:04.240 --> 0:05:08.000
<v Speaker 4>think that this is about any one company acquiring or

0:05:08.000 --> 0:05:10.640
<v Speaker 4>not acquiring, But it's about competition. And as long as

0:05:10.680 --> 0:05:13.800
<v Speaker 4>regulators focused on that and we focus on it, I

0:05:13.839 --> 0:05:14.679
<v Speaker 4>think we'll be fine.

0:05:14.800 --> 0:05:17.240
<v Speaker 2>Microsoft just took a big stake in the Abu Dhabi

0:05:17.279 --> 0:05:18.680
<v Speaker 2>AI company G forty two.

0:05:18.800 --> 0:05:20.640
<v Speaker 1>Brad Smith joined the board.

0:05:21.000 --> 0:05:22.680
<v Speaker 2>This is a company that at least at one point

0:05:22.720 --> 0:05:24.479
<v Speaker 2>had ties to China. So there do seem to be

0:05:24.520 --> 0:05:26.559
<v Speaker 2>some big geopolitical implications here.

0:05:26.960 --> 0:05:29.760
<v Speaker 1>Is this your play in a tech cold war?

0:05:30.560 --> 0:05:32.680
<v Speaker 4>You know, first of all, you know it is important

0:05:32.720 --> 0:05:35.279
<v Speaker 4>for us to be grounded on the fact that we're

0:05:35.880 --> 0:05:38.279
<v Speaker 4>a multinational company, but that also happens to be an

0:05:38.320 --> 0:05:41.559
<v Speaker 4>American company. So we did this partnership with G forty

0:05:41.560 --> 0:05:46.839
<v Speaker 4>two and the UAE, with sort of the real involvement

0:05:46.880 --> 0:05:49.120
<v Speaker 4>of the US government as well, where the US government

0:05:49.160 --> 0:05:52.520
<v Speaker 4>has now an assurance program with the UAE, which both

0:05:52.600 --> 0:05:54.719
<v Speaker 4>the government, you know, it's a state to state agreement,

0:05:55.120 --> 0:05:57.680
<v Speaker 4>and that allowed us in fact created the conditions for

0:05:57.760 --> 0:05:58.440
<v Speaker 4>G forty.

0:05:58.240 --> 0:05:59.599
<v Speaker 3>Two and Microsoft to come together.

0:05:59.640 --> 0:06:03.040
<v Speaker 4>But I think it's important for us as a company

0:06:03.120 --> 0:06:05.119
<v Speaker 4>to be grounded in the fact that we need great

0:06:05.160 --> 0:06:08.160
<v Speaker 4>partners who then allow us to expand what we are

0:06:08.160 --> 0:06:10.159
<v Speaker 4>doing globally, right because at the end of the day,

0:06:10.360 --> 0:06:12.719
<v Speaker 4>one of my dreams has always been Emily that can

0:06:13.080 --> 0:06:16.520
<v Speaker 4>the you know, the information revolution, perhaps after the Industrial

0:06:16.560 --> 0:06:20.320
<v Speaker 4>Revolution was magical in the sense that it diffused fast

0:06:20.400 --> 0:06:23.760
<v Speaker 4>around the world, but this intelligence revolution, to me, can

0:06:23.839 --> 0:06:26.200
<v Speaker 4>even take it even to the next level, right, which

0:06:26.240 --> 0:06:28.640
<v Speaker 4>is when people in the Middle East, or whether it's

0:06:28.640 --> 0:06:31.479
<v Speaker 4>in Asia or Africa or Latin America can enjoy the

0:06:31.520 --> 0:06:35.119
<v Speaker 4>benefits of this technological wave. That's not going to that's

0:06:35.200 --> 0:06:38.159
<v Speaker 4>going to require real partnerships, And I'm very, very excited

0:06:38.160 --> 0:06:38.800
<v Speaker 4>about this one.

0:06:38.920 --> 0:06:42.400
<v Speaker 2>Open Ai is your closest partner in Microsoft hit history.

0:06:42.920 --> 0:06:46.040
<v Speaker 2>They just released a chat GPT app for Mac before Windows,

0:06:46.080 --> 0:06:48.080
<v Speaker 2>and we're reporting the same as going to happen for

0:06:48.480 --> 0:06:51.480
<v Speaker 2>the iPhone. How do you balance partnering with them and

0:06:51.560 --> 0:06:53.279
<v Speaker 2>competing for their attention?

0:06:53.920 --> 0:06:54.080
<v Speaker 3>Yeah?

0:06:54.120 --> 0:06:56.480
<v Speaker 4>First of all, you know, like open Ai, you know,

0:06:56.520 --> 0:07:00.440
<v Speaker 4>will decide how to prioritize platforms and and how to

0:07:00.640 --> 0:07:02.520
<v Speaker 4>really go about managing their roadmap.

0:07:02.520 --> 0:07:03.479
<v Speaker 3>Would be very excited.

0:07:03.480 --> 0:07:05.599
<v Speaker 4>In fact, you know, I'm thrilled to you know today

0:07:05.640 --> 0:07:09.040
<v Speaker 4>when you saw the launch, we basically took gpto and

0:07:09.120 --> 0:07:11.280
<v Speaker 4>made it part of Core Pilot. So Windows is going

0:07:11.320 --> 0:07:14.120
<v Speaker 4>to have some of the latest innovation from open Ai

0:07:14.240 --> 0:07:17.800
<v Speaker 4>and the latest innovation from Microsoft built into Windows. So therefore,

0:07:18.560 --> 0:07:21.360
<v Speaker 4>I'm thrilled about their prioritization of the Mac.

0:07:21.520 --> 0:07:24.520
<v Speaker 1>Chat GPT four. Oh is you know, it's pretty big?

0:07:24.560 --> 0:07:24.800
<v Speaker 4>Wow.

0:07:24.920 --> 0:07:26.679
<v Speaker 1>Have you tried chat GPT five?

0:07:27.560 --> 0:07:28.840
<v Speaker 3>So that the latest? Oh?

0:07:28.880 --> 0:07:29.440
<v Speaker 4>Yeah, of course?

0:07:29.480 --> 0:07:32.400
<v Speaker 1>What about the next one? We hear reasons you mean,

0:07:32.720 --> 0:07:33.160
<v Speaker 1>so the.

0:07:33.120 --> 0:07:34.960
<v Speaker 4>Next version, which I believe is not been named, so

0:07:35.000 --> 0:07:37.960
<v Speaker 4>I don't want to talk for them, but you know, look,

0:07:38.080 --> 0:07:43.040
<v Speaker 4>I think open Ai as a company has really focused

0:07:43.080 --> 0:07:45.640
<v Speaker 4>a lot on not only moving the frontier forward, but

0:07:45.680 --> 0:07:48.720
<v Speaker 4>also doing it with really safety first, and so they're

0:07:49.160 --> 0:07:51.880
<v Speaker 4>very carefully thinking about what's the next big model. But

0:07:52.000 --> 0:07:54.680
<v Speaker 4>the thing that GPTO like, I don't want to understate

0:07:55.120 --> 0:07:57.680
<v Speaker 4>the innovation in what they just launched last week, right,

0:07:57.720 --> 0:07:59.680
<v Speaker 4>which is just the first time they've been able to

0:07:59.680 --> 0:08:03.080
<v Speaker 4>bring ultimodal capabilities natively into the model. And you're seeing

0:08:03.080 --> 0:08:06.280
<v Speaker 4>it like in the wild as you use copilot or

0:08:06.360 --> 0:08:08.440
<v Speaker 4>chat GPT you can see the power.

0:08:08.880 --> 0:08:09.040
<v Speaker 3>You know.

0:08:09.080 --> 0:08:11.600
<v Speaker 4>It's most people think, oh, GPT four came out what

0:08:11.720 --> 0:08:15.520
<v Speaker 4>fifteen sixteen months ago, but it's not been static. It

0:08:15.600 --> 0:08:18.440
<v Speaker 4>has been continuously getting better and we're excited about it.

0:08:18.520 --> 0:08:20.920
<v Speaker 2>So on that note, you stress trust a lot in

0:08:20.960 --> 0:08:23.440
<v Speaker 2>your keynote today, But there have been renewed concerns over

0:08:23.440 --> 0:08:25.840
<v Speaker 2>the week and about safety at open AI. More key

0:08:25.840 --> 0:08:29.600
<v Speaker 2>employees have left, one saying safety has taken a backseat to.

0:08:29.560 --> 0:08:30.640
<v Speaker 1>Shiny new products.

0:08:30.920 --> 0:08:33.199
<v Speaker 2>What kind of conversations have you had with Sam Molman

0:08:33.240 --> 0:08:34.320
<v Speaker 2>about this look.

0:08:34.320 --> 0:08:36.960
<v Speaker 4>I mean, at least one of the fundamental things that

0:08:37.040 --> 0:08:40.480
<v Speaker 4>broad Open Air and Microsoft together way back even in

0:08:40.520 --> 0:08:44.240
<v Speaker 4>twenty nineteen was that focus on how do we make

0:08:44.320 --> 0:08:46.560
<v Speaker 4>sure that we could make progress. At that time, it

0:08:46.600 --> 0:08:49.400
<v Speaker 4>is not even clearest whether things will even work the

0:08:49.440 --> 0:08:52.360
<v Speaker 4>way they work, right, But even there, that company was

0:08:52.440 --> 0:08:54.720
<v Speaker 4>very grounded on their mission around Hey, we want to

0:08:54.720 --> 0:08:56.679
<v Speaker 4>bring the benefits of this to the broadest set of

0:08:56.720 --> 0:08:59.720
<v Speaker 4>audience and do it safely. So and that requires, at

0:08:59.800 --> 0:09:01.520
<v Speaker 4>least in my mind, I fall into the camp of

0:09:01.640 --> 0:09:04.200
<v Speaker 4>iterative development of it. Right, you got you can't just

0:09:04.240 --> 0:09:06.160
<v Speaker 4>be in the lab. You have to get it into

0:09:06.200 --> 0:09:09.760
<v Speaker 4>the real world. See the adversarial attacks, make the safety

0:09:09.840 --> 0:09:13.040
<v Speaker 4>harnesses better. That doesn't mean you don't take safety first

0:09:13.080 --> 0:09:15.840
<v Speaker 4>as the most important feature. In fact, I say, even

0:09:15.840 --> 0:09:19.240
<v Speaker 4>at Microsoft, which is the most important feature of Azure

0:09:19.240 --> 0:09:22.240
<v Speaker 4>AI is our safety service, because that's what even gives

0:09:22.320 --> 0:09:24.080
<v Speaker 4>us permission to deploy anything.

0:09:24.320 --> 0:09:25.280
<v Speaker 3>You can't just deploy.

0:09:25.320 --> 0:09:27.640
<v Speaker 4>I mean, if I take the base model, like even

0:09:27.679 --> 0:09:31.400
<v Speaker 4>take today's announcement, it's just so important, right, Windows now

0:09:31.440 --> 0:09:34.680
<v Speaker 4>will have forty plus models that are going to be

0:09:34.760 --> 0:09:37.760
<v Speaker 4>onboard on Windows. But guess what, in order to keep

0:09:37.800 --> 0:09:41.640
<v Speaker 4>them safe, we will use the Azure AI service because

0:09:41.679 --> 0:09:44.880
<v Speaker 4>we will be learning every hour of every day what

0:09:45.040 --> 0:09:48.920
<v Speaker 4>is the new adversarial attack, and that knowledge, that learning

0:09:49.040 --> 0:09:51.559
<v Speaker 4>will keep all the AI models safe. Because the next

0:09:51.679 --> 0:09:53.959
<v Speaker 4>jail break that happens, we'll learn about it and make

0:09:53.960 --> 0:09:55.880
<v Speaker 4>sure that that doesn't happen on the local model. And

0:09:55.920 --> 0:09:57.600
<v Speaker 4>so that's I think the only way we're going to

0:09:57.679 --> 0:09:58.640
<v Speaker 4>keep AI.

0:09:58.440 --> 0:10:01.080
<v Speaker 2>Safe, all right, Microsoft CEA, oh saucha nidala, thank you

0:10:01.120 --> 0:10:02.719
<v Speaker 2>so much for joining us. I'm going to send it

0:10:02.760 --> 0:10:05.239
<v Speaker 2>back to you guys, and thanks to our Bloomberg television

0:10:05.240 --> 0:10:06.360
<v Speaker 2>and radio audiences,