← Everyday AI Podcast – An AI and ChatGPT Podcast
Everyday AI Podcast – An AI and ChatGPT Podcast
Ep 791: Microsoft Build Recap: 4 New AI Features That Stood Out
00:41:16
Loading summary
Transcript3 lines
- [00:01]
A
This is the Everyday AI show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business and everyday life.
- [00:17]
B
Microsoft just wrapped up its annual Build Developer Conference in San Francisco. So Obviously there's like 197 new pieces of AI hardware, software, infra dev tools and more. And I'm not going to painstakingly read them all off to you on today's episode. You know I go long as we'll just link to those in the newsletter anyways. But I was lucky enough to get to attend the conference in San Francisco talking with Microsoft C Suite leaders, product managers, developers and more. And while the announcements were buzzy, I'm not personally walking away buzzing from any one source certain updates. And that may be because I'm personally a Mac user. So many of these AI advancements aren't going to touch my daily workflow. However, the totality of what Microsoft announced actually showed that it is beyond serious about competing at the highest level in the AI era, even though it may not win any single race. And the hundreds of new AI updates are obviously going to change how millions of enterprises work. That's why on today's show we're gonna highlight the four biggest takeaways that I had and that you need to know from Build and how they'll shape the future of AI in the workplace. All right, so here is the big picture. So even though this is a developer conference, there's plenty of new updates for non technical business leaders as well. And Microsoft showed all of its cards. It's going all in on Agentic Data Desktop, not just with its Open Claw integration, but with new local models, new hardware and software. And I think though that Microsoft is still very far behind in some areas of AI, yet it's setting the tone in other AI eras. But I don't even necessarily think it matters which category they may win or may not win. Because in the end, I think Microsoft is always just going to be a top two or three company in just about everything and that seems to be their strategy. So on today's show, you're going to know the four biggest AI moves that Microsoft made at Build and my thoughts on them. I'm going to tell you why I don't think it matters if Microsoft's new models that they announced. Yeah, seven of them total, not all of them new, but some of them new. I'm going to tell you why I don't think it matters they're successful or not, and you're going to Know why Microsoft's small hardware plays can help overcome their big model problems? Let's get into it. My name is Jordan Wilson. Welcome to Everyday AI. If you're new here, yeah, Everyday AI, you guessed it by the name. We do this every day. We bring you the latest and the greatest unbiased takes. In this daily live stream, podcast and free daily newsletter. I help you make sense of what's important, what's not. You use that information to grow your company and your career. So starts here with the unscripted, unedited daily podcast. But to take it to the next level, make sure you go to our website at your everyday AI.com we're going to be recapping today's show and giving you all the other AI news that you need to know to get ahead. So Microsoft Build Great conference, right? You know, I am lucky enough to get to go to a handful of conferences a year. I've done Microsoft's Ignite conference. That's maybe more for general business leaders and then Build is for developers. But I will tell you this, even what was covered in this build is much different than what's been covered in years past. Right. Some of the normal build tools, Right. Some of the more developer centric tools that normally are grabbing all the headlines barely got mentioned. And I think that's just indicative of where we're at at in today's day and age. Right. The tools are changing, the capabilities are shifting and even what it means to be a developer, that bar is being lowered. And I don't think that that's a bad thing necessarily. And I think before we even talk about, you know, here's what's new. Right. All those things. I think we have to hit rewind because it's important to put this into context. Right? Yes. The AI race is changing all the time and I'm going to get a little bit into why I think Microsoft strategy is actually really sm. But there was some reporting about Microsoft's own internal code red back in September. So that's essentially according to reports. There was an internal town hall where Microsoft CEO Satya Nadella said that he was haunted at the prospect of Microsoft not surviving the AI era. Did he actually mean that? I don't know. Right. Maybe that was just a, you know, a rallying call to, you know, get the troops, you know, working together on a more unified version. You know, know, we're going to talk about their, you know, super app, which they kind of talked about but kind of didn't. Right. But there was this narrative as recent as like eight or nine months ago that internally Microsoft felt that they were going to get squeezed out of the AI race. And I think some of these announcements show that, you know, their moves have paid off and just because, right. And I, I, I know what you're probably thinking if you're listening to this, right? You're like, oh my gosh, you just think of Microsoft in their AI play as Microsoft 365 Copilot. And there's no denying that it's not anyone's top three choice, although it could be if you get everything set up correctly. But it's an uphill climb to even understand where to use Copilot because there's literally hundreds and that is not an exaggeration. There's hundreds of hundreds of different offerings within Microsoft 365 that have that co pilot tag. Right. So I don't think when you're talking about Microsoft and just overall as a company, you can't judge them just by co pilot. And I think that, you know, maybe the response to the reported code red is paying dividends. And let's talk about it by the first big update. So this one is called Autopilots. All right? So essentially S. Nadella said one sentence that hardly no one talked about, right? There's obviously, you know, I'm getting to meet, you know, all the journalists from all over the world. So which is funny, sometimes I get thrown in as an analyst, sometimes I get thrown in as media, sometimes I get thrown in as creator. I think this is, Microsoft has labeled me as something else each time. Anyways, you know, I do get to talk to, you know, press and media people from all over the world and I literally didn't see anyone else mention this one thing that Sadie Nadella said. He in reference, did reference or sorry, in passing, he referenced the super app and he just said, the co pilot super app coming this summer. There's no on screen visual. So, you know, that was one of the things that I was personally keeping an eye on because we did see leaks, you know, so there was talk that maybe that they were going to officially announce it. So they kind of did, kid. So Satya just referenced it in passing and said, yeah, the co pilot super app is coming this summer. He literally said that one line. It was almost a rushed transitional line and no one talked about it, right? So he didn't mention it. So the super app is coming. But what completes the super app is this first thing and this is the autopilots. All right? So essentially, you know, you have these four different phases of co pilot in the newest phase, right, is the autopilots. All right, and the first autopilot is named Scout. So let's just quickly talk about this. So I would say this is the biggest announcement and this is really where, you know, Microsoft wants Copilot to start, well, doing work for you, right, where you know, they have their Microsoft cowork. And yes, you can run scheduled things on co work, but for the most part, you know, you are still going in there and pushing the buttons and building the workflows. It's not necessarily an autonomous agent. And that's what they're trying to get to with autopilots in their first autopilot named Scout. So autopilots are Microsoft's new category of AI agents that just keep working in the background. So they're not just waiting for a prompt. They have their own identity, their own permissions, and then the ability to act across different work systems. And then Scout is their first autopilot. All right, so Autopilot is the kind of the category name. All right. And presumably there's going to be different autopilots and Scout is the first one. So that is essentially the always on Microsoft 365 agent. And then it can work across Teams, Outlook, OneDrive, SharePoint, Calendar files, etc, right. Anywhere where your work context is or your work iq. You know, Scout can go there and take in all that data. So you know, essentially co pilot just answers questions. And autopilots are designed to do work. Right. So Scout, according to Microsoft can help with meeting prep, scheduling conflicts, follow ups, reminders, task handoffs and work coordination. So it can just work on its own, right. And kind of give you what you need or you can kind of, you know, communicate to it like you would an assistant and then it can just go off and work autonomously on a goal. So we've talked about this. If you haven't listened to our start here series, FYI, I think we might cap it at 30 episodes. Right? It's nice you can listen to one a day if you're new. But we have talked about the shift from reactive prompting AI chats like you would in Copilot to now proactive. So that's what Microsoft is trying to get here. So Scout is right now though limited to only select private preview customers and Frontier organizations also. Yeah, so you need to be enrolled in Frontier, need to have the intune setup and you have to opt in and you also need a GitHub copilot license. So at least right now there's a lot of criteria, a lot of checkboxes. If your organization wants to use Scout. I think I looked, you know, it's, it's been announced now for like 48 hours. I was like, has anyone actually used it? I think I saw like two videos on YouTube. Right, right. Which is not a lot. When, you know, obviously Microsoft Co Pilot has millions of customers. So to only have, you know, two people out there that have like, hey, I have access to this and I can share about it. You know, I'm guessing there's not a lot of people out there that do have access. But I mean, the upside's big, right? Obviously the rest of 2026 is going to be dominated by two things. They're kind of the same, right? But autonomous agents, number one. But the other half of that is ones that work on your desktop. Right. So that is big because then it can work with your local context. Right. And then it can also use your computer and use your browser. So it's also risky. Right. And that's why, you know, I do think Microsoft is obviously taking a very measured and rolled out approach to this because of the risk know, as great as it is, if you do have a, a background agent that is just proactively, you know, grabbing details that are important out of your calendar, something changed, you know, helping you respond to emails, you know, doing research on your competitors, tracking pricing changes, all those things without you having to tell it. It can also do those type of tasks, but going a little off the guard rails. Right. All you need is like a 1% drift from an autonomous agent and you don't check it enough all of a sudden. Yeah, that 1% is very far off base. So the real question you have to be asking yourself is, can this agent help? Right. Is it going to get you over a big hurdle that right now you can't do with any of the other hundred plus instances of Microsoft Copilot or within Cowork. Right. That's powered by anthropics technology where you can, you know, essentially have autonomous esque scheduled runs. So I mean, bottom line, Autopilot's Microsoft's move from AI assistant to, you know, proactive AI workers. And then Scout is the first example. But I think for most companies this is still just a preview of what Microsoft is trying to envision for the future. And I don't think. Right. In terms of a general availability, I don't think it's going to be for a while, right. Maybe, maybe late 2026, but I wouldn't be surprised if things get pushed back for security reasons. But they are regardless showing their cards there. Pretty big splash. All right, number two, we finally got the Microsoft large language models, well, technically medium language models. So Microsoft did a lot to position their new models, the MAI series. So essentially they had the MAI Thinking one and then the MAI Code one, Flash. All right, so they had already been releasing and announcing some of their other multimodal models over the last couple of weeks. So most of them are fairly new. But the two new ones that were announced were their versions of large language models. So Microsoft in total now has seven new in house MAI models across reasoning, coding, image transcription and voice. But the headline obviously is their new Mai thinking one, which is their first in house reasoning model. And it's a 35 billion active parameter model built for reasoning, long context work, math and code generation. So if you don't follow, you know, LLM stats like I used to follow basketball stats in the 90s, that's technically a, a medium sized model, right? We don't know the number of active parameters for the, you know, the GPT5.5 to the world, the Opus, you know, for Opus4.8 and when we presumably get a Gemini 35 Pro, but presumably those are at least 1 trillion, probably multiple trillions of parameters. So this is in order of magnitude smaller. Right? Which is in theory a good thing because in terms of benchmarks, right, we're not really going to go over this, but essentially this performs at about in Opus 46 level, give or take. Right? So what does that mean? That means they're about six months behind the frontier. So this is more on pace with some of the better open source models, right? Not the best. So a lot of people, it was actually confusing and kind of funny, you know, and this just shows me maybe I, I, I am way too terminally online and follow AI too much, right? They put something up on the screen, you know, saying that, you know, Ma Mai Thinking one, you know, beat Opus four. Six in a certain benchmark and like a third of the crowd went absolutely wild, right? And I kind of chuckled because I think those people probably don't follow benchmarks and they thought like, oh my gosh, you know, it's better than one of the best models in the world, you know, not knowing or already obviously on Opus 4.8. So I chuckled at that one. But still, that's telling, you know, if Microsoft's first in house, large model is already, and I say this already, right, if they're only, you know, like eight months behind the frontier. That's not entirely terrible. Right. Especially that it's only 35 billion active parameters. All right, so it is smaller. You know, those are active parameters, not total parameters. So, you know, it's smaller than, you know, the, presumably the competitors that are punching around the same weight. So not, not bad. But I do think that they've had a little bit more pressive, impressive results. These were not announced, you know, at build. These were announced about a week or two before. But their new image models did really, really, really well on benchmarks. You know, I think they, you know, it was like number two and then it kind of got pushed down a couple spots because there are some new image models that just were released. But you know, regardless, it looks like Microsoft on certain modalities is going from zero to a top three lab, which is pretty impressive. Right. I said the same thing. I think people were kind of, you know, poo pooing on Meta when they released their Meta Muse Spark. Right, right. Because rewind to two years ago. Right. The Llama models were serious contenders in the frontier and then they absolutely fell off and you know, essentially Meta started, they started over and they came out with their metam Spark. And even though it wasn't like a top five frontier model, you know, it wasn't that. Right. To go from zero to six to nine months behind. I know that sounds crazy for an organization, you know, a multiple trillion dollar market cap like Microsoft that has all the resources in the world, but it's not easy, it's not easy to go from zero to contender in six to nine months. So from that aspect, tip your cap to them. Is anyone going to be using these models? I don't know. Right. Maybe, you know, they are going to probably cost less than Frontier models if companies maybe have to stick within the Microsoft ecosystem. But at least for me, you know, I'll maybe demo it, try it, but there's no use cases, at least that I know of, where I would actually be using any of these in production. And I would assume most of our audience is probably going to be the same. Like I said, unless you're some unicorn of an enterprise where you can literally only use a Microsoft AI, which I don't think that's the case because Copilot obviously is built in support for the GPT and the CLAUDE models as well. Also, at least right now, Mai thinking one is not broadly available. Right. So some of these things you can go and you know, test them on arena with the different, you know, the Transcribe model and you Know the image models. So they're not generally available yet. So the upside, right, cost and control. Also there's the frontier tuning which is going to be limited and available to select early partners. Right. So to be able to tune a Microsoft model with your enterprise data, right, that's, that's a nice thing to have. But this does not mean that Microsoft is a top three lab. I did think, you know, heard references that they want to be considered a top four lab. Does this put them there? Not yet. Can they get there? Absolutely. Right. I would assume that we would see up in to the Ma Mai thinking one model right before Ignite. I would assume that they'll eventually get on the same, you know, 40ish day release schedule, 40 to 60 days. So we'll see. Number one, when are they going to bring it to the masses? Number two, when is the next update going to be and is that update going to be a small update or is that going to say, okay, they can be a top 5 model provider? All right, next big update, hardware. All right, so we did talk a little bit about this on our Monday news show because technically this hardware was announced by Nvidia CEO Jensen Wong at Computex. So we'll go through this a little bit quicker but obviously Satya Nadella did announce this and it's a pretty big deal. Essentially this one, they're going for Apple, right. For the first time I think you have highly capable Windows PCs that are really good, right? The specs, they look good. Personally, I like the marketing behind them. All right, but let's talk about them. So Microsoft and Nvidia partnered up and are pushing the new class of local AI PCs. So that includes the RTX Spark laptop. So that's the, you know, the RTX a series by Nvidia that was announced around Computex and then the new Surface RTX Spark dev box. All right, so the way that they're, you know, angling that is essentially saying it's a data center in a little, in a little box, right? Or at least what you, the, the power that you would have to have a data center for, you know, maybe five, ten years ago now you can get it in a little, little box that you can put on top of your desktop. But they're built for real, real local inference, right? So for those, whether individual developers or you know, businesses that are trying to bring things on prem certain tasks, right? This is what it's for. This isn't like for the most part you're like, oh, I need a new laptop for you Know, web browsing? No. Right. These are the type of things like for me as an example, I got a fairly expensive MacBook Pro because I'm running, I'm running models on it constantly. So same thing, I got a Mac Studio. So these are kind of the equivalents, right, of those higher, you know, like MacBook Pros and the higher end Mac Studios that have been selling a ton recently because of these ability to run local models. So this partnership between Microsoft and Nvidia, pretty big. And the fact, Sadia said it multiple times in his keynote that essentially 100% of the PC industry got on board, I think that's good because that sets a, essentially it sets the bar consistent. Right. So you know, if you're going to go spend a couple, couple dollars because these aren't cheap yet, Right. We don't have complete pricing, but presumably they're going to be multiple thousands of dollars for the laptops, I'm guessing the dev box is going to be way more than that. Right. So it's not like a random consumer hardware you're going to get going off to college. But if you are a serious AI developer, if you're someone that likes to tinker around like me, build projects and you are a PC person, this is the one to keep the eye on. So the RTX laptops and the Surface RTX Spark dev box, I think they matter less as individual products and more of just the signal, like I said, right. Local AI is becoming a part, a key part of the Microsoft AI stack. Right. So it's not just that they're, you know, building out their, you know, their cloud models, right. The Mai models, they're getting super serious about AI hardware to compete with Apple. But at the same time they also developed some small local models, the Ion models, Right. So they're also now offloading, you know, some simple local, some simple AI tasks locally. And that part's important, right? Because I think that's the future of where all of this is headed. Right. Doing transcription live on device, Right. So many of these things, being able to run on your local machine is a big step forward in terms of where the future of work is going. AI moves too fast to follow, but you're expected to keep up otherwise your career or company might lag behind while AI native competitors leap ahead. But you don't have 10 hours a day to understand it all. That's what I do for you. But after 700 plus episodes of everyday AI, the most common questions I get is where do I start? That's why we created the Start Here series, an ongoing podcast series of more than a dozen episodes you can listen to in order. It covers the AI basics for beginners and sharpens the skills of AI champions pushing their companies forward. In the ongoing series, we explain complex trends in simple language that you can turn into action. There's three ways to jump in with number one, go scroll back to the first one in episode 691. Number two, tap the link in your show notes at any time for the Start Here series. Or you can just go to start here series.com, which also gives you free access to our inner circle community where you can connect with other business leaders doing the same. The Start Here series will slow down the pace of AI so you can get ahead. All right. And then last but not least, and this one was a different one and I'd say, at least for me, this is maybe the one I was most impressed by. Just because we haven't really seen anything like this yet. Is it a gamble? Yes. But Project Solera, if it works, this, this could be one of the bridges, right? I, I like, I talk a lot about embodied AI, right. And, and what does that mean? Right now we're still in consumer enterprise AI, Right. Right now AI is happening in front of the computer and in the cloud. This is not truly embodied AI, but I think this is a bridge to embodied AI. So Project Celera, it's essentially a device based. Right, but it's a chip to cloud platform for agent first devices. So kind of the idea is that, you know, future devices may not be organized around apps and instead they're going to be presumably organized around agents. So the Solera is essentially a collection of hardware partnered with Qualcomm that essentially allows your work to be handed off right from, you know, agent to agent. You know, they had this cool like little badge one that they demoed, they demoed live, right? So it has like a camera, there's a microphone, you know, you can walk up. There's a version that sits on your desk. So if you walk up, it kind of has the sensors and it will unlock for you. So there's some security things in there as well. So I think that there's a lot of potential implications. This is one of those things. It's either, it's either going to hit or miss, right? This is not just going to be one of those things that's just like, oh, it's, it's warmly received and just kind of exists. There's, I think it's either going to be a Complete, you know, flop or it's going to be something pretty popular that I think could bridge the gap. Because there is a gap right now, right, between the future of AI, which is embodied AI. Right. So that's your, you know, robotics, your last mile deliveries, you know, like your, your waymos of the world. Right. But also all of these agents, well, they need a way to communicate as well. So this could be essentially a, a transistor, a relay device that helps kind of bridge, you know, our current AI, whether that's on our local machines. In this case, you know, you're probably going to have a Solera account that's probably going to use your Microsoft credentials. Right. But that's the way I see this. It's just kind of a potential baton until the technology gets better, which is why it may flop. But it looks cool and it could absolutely be a hit. So a little bit more. But like I said, it's really. They showed off kind of the two concepts for the most part. They showed a badge and then a desk device. So it's not a project, a product launch yet. It's an early platform and just kind of design direction, more or less. But Microsoft said that pilots are planned with companies including AccuWeather, Best Buy, CBS Health, Levi's, Target and others. But you know, kind of what they were pushing is that the business relevance is just frontline work. Right. So agents may need to eventually live in badges or desk devices or retail devices, healthcare tools, field service devices, hospitality workflows. Right. Industrial environments. Right. That's the other big one I think is, you know, just industrial warehousing, logistics, transportation. Right. In those instances where we are going to, and we're already starting to see, you know, humanoids and just more autonomous operations happening. You know, I think that whatever, you know, expert driven loops, they're going to probably need a device and an agent to go around with them to help communicate. And maybe it doesn't make as much sense to be carrying around a computer. So could this just be a phone? Right, that's. I think that's what you always have to ask yourself, right? Could this cool badge thing just be a phone? It's like, could that just be an app maybe? You know, I, I think that they were kind of saying it's just like how you would normally wear a badge in certain instances, right. Like if you work in a huge office, you know, for medical, you know, medical workers. So maybe it's just more to replace an actual badge. But for me, I'm like, yeah, I don't know if that's just going to be something that's like an unnecessary piece of technology or if it's, well, that actually makes sense. So, I mean, there's a lot of upside because the AI could help in places where laptops and chat windows are awkward and you maybe need agent to agent communication. But there's obviously a big concern there, right? Privacy concerns. You know, Badge with camera, microphone, transcription, identity and agent access. Right. Can quickly move from being a productivity tool to being a surveillance device. So there's obviously some concerns. But I, I personally see the optimistic side. You know, maybe that's because, you know, Microsoft was really pushing the healthcare angle to it, which I think is an obvious one. Right. I, I know especially here in the US healthcare workers are overloaded. They're working with antiquated legacy systems. So maybe this is something since it comes from Microsoft and it has that kind of enterprise backing. Maybe this is something that could really, even if it just flops in every other industry, but it works in something like healthcare. I think that's a net positive for humanity. So personally I'm like, this sounds cool. There's obviously some security concerns, not just the surveillance piece, but also just having more people having agentic access. Right. Like if these are truly agent devices, handing them out to, you know, I don't know, a couple thousand hospital workers. There's also, you know, the potential for agent drift or agent chaos. But I'll wait. So bottom line on that, I mean, Celera is just Microsoft trying to kind of define the enterprise hardware layer for agents because no one's playing there yet. And I think the opportunity it for the most part, you know, frontline productivity. But the risk is the worker surveillance, but just with better branding. All right, so those are the four big announcements. Again, just to recap. Number one, we have the super app is here with Autopilot and the first Autopilot is Scout. It is the autonomous worker. Number two is Microsoft's new Mai models, specifically their large language model, but medium language model technically with Mai thinking one and Mai code Flash, Code one flash. Then we had the partnership with Nvidia, the RTX Spark laptops and the Surface RTX Spark dev box, bringing some serious AI inference to your desk. And then last but not least, Project Solera, the agent first devices beyond the devices that we currently use. All right, so that's a wrap on those, but let me get back and answer some of those questions as we wrap up. So I talked about that story in September where, you know, Microsoft's internal Code Red and you know, feeling that they weren't, you know, or you know, that Satya Nadella felt haunted by potentially not surviving the AI area era. So I'm not saying that the Microsoft Build conference, you know, erased that, but I think we're well past that. And it's not just because of some of their recent moves, right? Things like adopting openclaw. Right. That's kind of the underlying infrastructure that runs scouts. But there is also official support for OpenClaw. Right. So the combination, you know, coming out with this ion small language model, you know, and I'm assuming in the future as the technology gets cheaper, Microsoft is going to move more and more processes on device. So I think Microsoft, because it is competing across the full stack. Even if when you think of Microsoft you're not like you know, googly eyed over Microsoft Copilot and loving it, I think by Microsoft being the only player right now, think of this. Apple has absolutely trash AI. They don't have any AI. Right. On the phone, on the computer. We'll see next week at their WWDC conference, you know, maybe that changes. I wouldn't count on it. Right. But Microsoft is the only one that has the hardware layer, the software layer, the inference, the cloud. Right. I think Microsoft can win by just not losing and that's fine. Right. And I think it's fine if a lot of people in the enterprise and if, if you're a Microsoft 365 copilot organization, maybe you use co pilot for 10% of your tasks and maybe you use chat, GPT or Claude or Gemini for the other 90% and that's fine. And that doesn't mean that Co Pilot is losing or Microsoft is losing the AI race. They're just running way more races than anyone else. Right. Obviously them and Google are running the most races. You know, then Apple is, I, I don't know, Apple is lost out back hiring other people to run the race for them and then OpenAI and Anthropic and others are just, you know, doing really well but not really in hardly any races comparatively. So I think that they'll just always be a medalist and they may not necessarily win anything or win, you know, best model, best, best hardware. Right. They may not win anything but I think as long as you can compete or medal in every single event, that's okay as an organization. So if anything I walked away from Build again not being overwhelmed but saying okay, yeah, Microsoft is absolutely serious about AI and you know, at least going back to that September story, you know, I Don't think Satya Nandela or anyone else at Microsoft needs to be haunted about not surviving the AI era. You just can't look at the AI era as just being a large language model, as just copilot. Because if I'm being honest, even if the super app that comes out is a hit, right? Even if these autopilots are great, I still think that they're going to be behind. I still think they're going to be behind specifically OpenAI and Anthropic and maybe Google as well. We'll see. Google has some work to do. Same thing. Bringing more their products in a unified way. You know, the harnessing needs to be improved. But I think they can win just by not losing. The other thing to keep in mind at least when you do think about the model race, what's Microsoft have to lose? Right. I have a little screen up here for our live stream audience. They have a $5 billion investment in anthropic, so we don't know what percentage of ownership that is, but Anthropic, you know, now, right now with a market cap of almost a trillion dollars, right? Same thing, open AI, Microsoft has a reported 27% as converted. As converted after the, you know, nonprofit conversion to a public benefits corporation. Microsoft is the technically the largest single entity shareholder of OpenAI, right. So presumably that's going to be worth hundreds of billions of dollars, right? Because their market cap, depending on what you're looking at, is between 850, $900 billion right now. Open AIs. So and then they also have an investment in Mistral, right? So not even if you, you're saying, oh, you know, the cloud this, right. Just on the model side, if Microsoft is losing Open AI, does it matter? If Microsoft is losing to Anthropic, does it matter? Because every time Anthropic and Open AI makes money, right, if you choose one of those other models, if you're a co, you know, user, Microsoft user, and you're like, oh, I'm not going to use their models, I'm going to use Open AI. Well, they make money either way. But when it comes to the bigger picture, you do have to acknowledge the fact that Microsoft is falling behind. Even though I don't, I, I don't think that they're going to fall out of the race. The, some of the other competitors are just getting faster, right? And I think, I don't think Microsoft is going to drop out of the top five companies when it comes to market cap, but they're actually in danger of Doing so. Right. So they recently got leapfrogged by both Google and Apple. So they're usually like a top three. But you know, now they're number four in terms of market cap behind Nvidia, Apple, Alphabet, Google and Microsoft respectively. Will they get passed by Amazon? I don't know, but it's a possibility. Or broadcom or when SpaceX iPodOS. Right. There's a chance that SpaceX within a year could be, you know, in Microsoft territory there's a chance that Anthropic or OpenAI could as well. So even though Microsoft I don't think is going to lose, they just may not hold on to that, you know, number two, number three, biggest company in the world by market cap. But like I said, that doesn't mean that they're losing. So here's my big takeaway. Microsoft doesn't need to win. They just can't be in last place. And they're not. Right. The MAI models are impressive, but I don't think they're ever going to catch up. Right. We'll see when the super app comes along. Maybe there is going to be a certain point and this probably not for like at least 18 months. Maybe there's a certain point when the model itself is not going to be as important. I think right now, if you're a power user, the models themselves are not as important as the harnessing for everyone else. I think the models are still important for your non technical, your everyday user. But I think there's going to be a certain threshold where the models at that point are going to be so good. Even if the MAI models are six to nine months behind, I think at that point it still is probably going to be good enough for 90% of your casual users. Right. They have the compute, they have everything. So now that they have the team off and running, I wouldn't be surprised if that six to nine month gap, you know, starts to just be three to four months. Right. And at that case then you do have to start looking at Copilot a little bit differently in the Microsoft MAI models. So I already said the small language models and the combination of the small language model ION and the new Nvidia partnerships, that's a big signal about what's to come in Microsoft's overall direction. So if, if you're an organization that is still looking, which is smart. Right. If you're still worried about moving some of your workflows over to be AI native, you're not going to have that excuse anymore because there's a good chance, maybe not in this generation, but probably in the next generation of these, you know, rtx, Nvidia partnerships, you're going to be able to run probably with proactive agents, everything locally on edge. Right. So at that point, you're no longer going to be able to say, oh, it's sensitive, sensitive data. Well, the sensitive data is not leaving. Right. Turn the Internet off. Right. Disconnect every service and it'll still run. All right. So that's what I think we have to look at from Microsoft. It's a lot of moving pieces. I think sometimes people over index Microsoft's success as a company or Microsoft's spot in the AI race. They over index just on Copilot. And even with the new models, even with the Copilot super app, I don't think Microsoft is going to be number one and number two, and that's okay. And I think when you start to look at Microsoft as a company, you have to understand they're running more races than anyone else. Right. Like if this is the Olympics. Right. Yeah. They don't have, you know, anyone grabbing the gold medals, but my gosh, they're competing in every single race. So they may be still one of the top, you know, top dogs just by having the most opportunities to bark. All right, I hope this was helpful. Recapping the big four things stood out to me at Microsoft Build. So thank you for tuning in. If you haven't already, please go to your everydayai.com Sign up for the free daily newsletter. See you back tomorrow and every day for more Everyday AI. Thanks, y'. All.
- [40:56]
A
And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going for a little more AI magic. Visit your everydayai.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.