On this episode of the Six Five On the Road, hosts Patrick Moorhead and Daniel Newman are joined by Microsoft’s Eric Boyd, CVP, Azure AI Platform, and Asha Sharma, CVP, Product, Azure AI Platform for a conversation on the latest innovations and insights from Microsoft Build 2024.
Their discussion covers:
- The latest developments in Azure AI technology and how they’re driving digital transformation
- Strategies for implementing AI solutions within various industries to optimize operations
- Insights into future AI trends and how businesses can prepare for what’s next
- The importance of ethical AI and Microsoft’s approach to responsible AI development
- Collaboration opportunities within the Azure ecosystem for developers and partners
Start building with Azure AI Studio today and join us at Build via your home for Microsoft Build.
Watch the video below, and be sure to subscribe to our YouTube channel, so you never miss an episode.
Or listen to the audio here:
Disclaimer: The Six Five Webcast is for information and entertainment purposes only. Over the course of this webcast, we may talk about companies that are publicly traded and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors and we ask that you do not treat us as such.
Transcript:
Patrick Moorhead: We are back, and The Six Five is on the virtual road at Microsoft Build. It has been an amazing, amazing event. I mean, there’s stuff happening all over Microsoft, Daniel, I mean, on the left, right, we have Windows, we have devices. On the right, we have Azure and all things cloud. Pretty awesome.
Daniel Newman: Yeah, it’s been a great day and a half or so, so far, Pat. We knew this would be a big inflection point. You’ve been on stage more than a few times, talking about the future of the AI PC, and we’re having a moment here, but it’s also a massive moment in the data center, in the cloud. Enterprises are driving their strategies forward around AI, and trying to build on top of what they have, trying to keep things private, keep things safe, keep things connected, but also taking advantage of all of these exciting generative tools. The ones that we’re seeing in our everyday consumer lives are also, Pat, going to be driving into our business lives. So it’s a great event. I couldn’t be more excited. And of course, Pat, it’s always great when we’re on the road and we have a chance to talk to some of the executives as well.
Patrick Moorhead: Well, how about this virtual on the road? No, it’s great. So let’s bring in Eric and Asha from Azure and focused in on the AI platform. Welcome to The Six Five, Asha, Eric.
Asha Sharma: Thanks for having us.
Eric Boyd: Yeah, glad to be here.
Patrick Moorhead: Yeah, it’s amazing we’ve gone this long and haven’t talked to y’all yet. It’s like the last 18 months for us analysts have been pretty much AI nonstop, but again, thanks for coming on the show.
Eric Boyd: It’s been pretty slow on our end.
Patrick Moorhead: Yeah, like feature every 90 days, it’s great.
Daniel Newman: Well, Pat and I are pretty much like professional event attenders, so we can see it on the faces of the vendors when they’ve kind of got… There’s a different stages, the beginning of the event, the encouragement, the excitement, the fear, “How is everything going to be responded to? Are we going to hit the note?” And then you can see by the third day, as you’re winding down, booths are starting to come down, and you actually could see people’s shoulders relaxing a little bit.
And it’s not because happy the event’s over. You’re just, for the first time, you’re decompressing, the balloon is… Just the air is like, “Oh,” and you’re going to finally sleep a little bit tonight. But it’s been a wonderful couple of days, very exciting. Eric, I’d love to kick off the Q&A here with you and talk a little bit about what are some of the ways here at Build that you guys have discussed helping customers build with AI? And talk a little bit, with these new things you’re offering, what kind of momentum are you seeing?
Eric Boyd: Yeah, I mean, there’s been so much going on with our customers, and maybe it’s easiest to let them tell our story. I mean, we worked with Mercedes-Benz, right? And so Mercedes-Benz is integrating GPT-4 directly into their car, the MBUX voice assistant. And so now, you can just talk naturally to your car, as opposed to everyone’s used those voice assistants in the car that just don’t actually do anything. By bringing that power with it, you can get a much richer experience, but they’re not stopping there, right? They’re using GPT-4 Turbo with Vision on their dash cam, so they can understand what’s the environment, and what are the things that are going on for the drivers as they’re driving their cars, and provide additional context for the speech assistant, things like understanding the signs that they’re seeing, finding parking spots, or things like that. So really interesting use cases.
And so that’s sort of a car consumer end, all the way to the other end, where we’re working with developers, like we work with Unity, who it’s a leading platform for game developers to build and develop games. And they’re building a bot directly integrating… They call it Muse Chat, where now you can see their documentation, and help users and developers really create the perfect video game, sort of exactly in that environment. And so those types of stories we’re seeing over and over, across all the 53,000 organizations that we have using Azure AI. And so with customers really just coming in and driving all of these things, we’re having to move really quickly as well, and just continually bring out new functionality for our customers. And so most of the things that our customers come to us to try and do is retrieval of meta generation.
How can they get their data into these models? And the way that you do that most effectively is you need some sort of search engine, a vector search engine. And so Azure AI search, we just announced its mega storage enhancement, which in this update, we really decreased the cost of it, while increasing the storage capacity and vector index size. And so that makes it really cost-effective for customers. And so really interesting to see how all these things come together. GPT-4 Turbo with Vision, and so you can sort of see all the things you can do with that. And then of course, you know, saw GPT-4o, the new flagship model from OpenAI, which is now in preview on Azure. And so super fast, super efficient, and fully multimodal, all these things coming together. It’s really exciting to see just the momentum we’re building with our customers.
Patrick Moorhead: Yeah, and I love those two examples. And not that they’re the bookends, but on one hand, on the one side, you have users who can make their own games, along with developers. And my gosh, how long have we been talking about backup capability? Probably for about a decade. And we’re getting closer and closer here. Super exciting. And then, on the automotive side, and by the way, you also hit a big keyword, highlights here, RAG. RAG is everywhere. Dan and I can’t go to a conference where we’re not talking about that capability and the added precision it gets to, to get to the result that a lot of companies, and consumers, by the way, are looking for. So Asha, I have to ask you, what are the building blocks that you brought out here at the show that helped them build with AI? Maybe that helped Unity, maybe that helped companies like Mercedes?
Asha Sharma: Yeah, you probably heard the Copilot stack mentioned several times already throughout the last day and a half. And one thing is Eric and I get to build that with Azure AI, and we made several announcements across almost every single layer of the stack. So for example, on the model side, Eric was talking about GPT-4V, but also GPT-4 Omni. In addition to that, we made several announcements around expanding the 5 Family and introducing new models. So big time gen one from Nexla, Cohere, Arirang, Core 42 JACE. And so we’ve been doing a lot to think about how do we help people find the right model, for the right job, at the right price and quality curve? And so that was a big set of announcements we made.
We also continue to lean into safety and responsibility, especially with technology, as it continues to advance, so does our responsible AI. We’ve been the leader in the space for the last eight years, and that’s not going to slow down. And so we now have 90 plus features, 20 tools available, and we announced several new features. One of my favorite is the introduction of custom categories. So users can now create custom filters for their content moderation needs. We talked a lot about groundedness detection as well, prompt shields, and new partnerships. And of course, all of this is available in Azure AI Studio, which has been a labor of love for us, as we are building now GA, a tool chain to really help developers build, manage, and deploy these applications in one place.
Patrick Moorhead: I love it.
Daniel Newman: Yeah, you’ve seen a lot of momentum. Of course, you hear a lot about the cool new gizmo, which is Copilot, and everybody’s talking about, obviously, GitHub, Copilot, and how it’s kind of all the rage, it’s gotten a huge wins, the use has exploded, but you guys are really building that stack for developers from top to bottom. That’s like, “Yes, we of course have the low, the no code, and then the Copilots, but then also the tools for those very sophisticated developers to continue to build the most robust and comprehensive applications.”
And I think that’s a big thing that’s important coming out of this, because that community really is, it’s the underpinnings. And while we are making it easier, and there’s some debate on how developers will evolve, let’s just call it that, they are going to be critical this next wave. And I think Microsoft has been really, really consistent in their desire to make sure that audience doesn’t feel ignored, and they actually feel very embraced for the importance of the role they’re going to play. Eric, another thing that’s been making a huge splash in the Microsoft community is that PHI… It’s PHI… I don’t know how many times I’m going to have to say that it’s PHI-3.
Eric Boyd: That’s right.
Daniel Newman: Can you talk a little bit about what PHI-3 is and the significance in the AI landscape? And I’m going to try to say that six times quickly. PHI-3-
Eric Boyd: PHI-3, PHI-3, PHI-3, exactly. It’s been really exciting in the last month to release PHI-3 and get the customer feedback on it. Really, it represents new innovation in the way that you can train a model to get really high performance at a really small size. We thought through very carefully, almost giving it a curriculum, like you would give to a child, to sort of teach all of the things that it needs to know. And so when taking that approach, we’re really getting tremendous performance. Really, the PHI-3 models tend to punch a full weight class above their size. And so what we’ve done is we’ve released PHI-3-Mini, Small, and Medium. And so now we have all three of those available, in addition to PHI-3 for vision. And so you get the high performance. They’re also directly integrated into Azure AI, and so you can just directly call them, and they’re open source.
And so PHI-3-Mini is open source, so you can use it exactly the way that you need to. But they also, they were built really responsibly as well. We went through the same safety measures and the same evaluation techniques that we go through all of our models. And so you can know that when you’re using something from Microsoft, you’re really getting secure… It’s going to be something that will be safe and on your brand’s message. And so that should give customers a lot of confidence that they can work at it. And so to bring all these capabilities, and I think PHI-3 for vision is going to be really interesting, bringing vision into a much smaller form factor. It’s going to be exciting to see the types of applications customers are going to go and build with that.
Patrick Moorhead: Yeah, it is funny how everybody was just, “Large language models,” right? “That’s all there is.” And then kind of looking at the pros and cons of that and a lot of discussion, I call them “slims”, some people call them SLMs, whatever. And it’s cool that the smaller the model, the smaller the resources to take advantage of that. And that’s a really good thing when it comes to inference and infrastructure cost as well. And a lot of the clients that I’ve talked to, particularly when they’re working on very narrow types of answers that they want, in the healthcare community, in legal community, being able to narrow in on certain set, they don’t need a 70-billion parameter model.
The other thing that seems to have taken off a couple weeks ago, it’s not new, is this idea of multi-modality, which is large language models. We have videos, we have images, we have a voice in addition to the text. Can you talk a little bit about how this has evolved over the past years, and what might the impact be to different industries as we move forward, right? We’ve seen a lot of consumer goodness, which is great, and I know Azure serves companies that serve consumers, but what does it mean for certain industries that you’re piecing together here?
Eric Boyd: Yeah, it’s been really interesting to see just how quickly this space has moved. I mean, certainly, everything in AI has moved quickly, but as we thought about multimodal models and bringing vision and text together, it was just a few months ago, I was saying, “I think that’s going to really be big.” And now, you see what you can actually do with GPT-4o, and how seamless improving the performance and improving the accuracy, it starts to improve just the fluency and the types of things that you can do, as you’re interacting with and talking to this model. And so just thinking about, “All right, well, what are the types of applications or things you could build with it?” The obvious answer is like customer service and things like that. We’ve already seen customer service get supported a lot by bringing these text models and bringing them in. But now, you can bring in different data, diverse data inputs, and lots of different ways of engaging with the customer that really feels natural, in a way that it hasn’t in the past.
I think other areas like analytics is going to be super interesting. Now, having models that can really use vision to understand trends, and charts, and images, and things that they haven’t in a totally new and different way. The content use cases, these models have always been great at generating text, and sort of helping me write my document, and things like that, but now being able to create images and understand the images. You know, “Hey, here’s my PowerPoint deck. Go and look at what I’ve done, and now give me some ideas of how I can make it better.” We do that today with PowerPoint Designer, but just think of how much richer the capabilities that we’re bringing are going to be as we pull all this together. And so it’s really pretty exciting that just how quickly all of that new multimodal capabilities have come together, and the impact that they’re going to have on so many industries.
Patrick Moorhead: Yeah, and it’s not even just the getting that one mode right, it’s the mixture of these modes that I think we’re finding out is not easy.
Eric Boyd: Yep.
Patrick Moorhead: It’s not easy to do, from a latency standpoint, from an accuracy standpoint, which modality gets the higher weighting. But that’s why we have people like you and Asha to figure this stuff out. It’s hard.
Eric Boyd: We do our best. Yes, it is moving quickly and it’s really pretty exciting to see what customers will do with it.
Daniel Newman: Yeah, well, if these weren’t hard problems though, Pat, we wouldn’t have a lot of work. So we need hard problems, and we need people to solve hard problems, and it’s been a really fun conversation to cover the gambit of what’s going on. Of course, AI leads the way. I don’t even think right now it’s possible to have a technology conversation and not zero in on quite a bit of AI, generative AI.
Patrick Moorhead: Is anything in tech sold that’s not AI now, Dan? I mean…
Daniel Newman: I’m sure there is, but I just haven’t heard of it anymore.
Patrick Moorhead: Yeah, okay.
Daniel Newman: I’ll dig deep, I’ll dig deep and think of something. But it’s either AI itself, it’s driven by AI, it’s foundational to, it’s sensing technologies, it’s developers for AI, on, and on, and on we go. But I love the conversation, multimodal, of course, what we’ve seen in the last several weeks, the launch of 4 0 and some of the things that are going on there. And just seeing the ability for us to have much more human interactions with many different modes at the same time is super exciting. It’s also, I always say, it’s a little scary, but I think we’re all getting over that, and we’re learning it and we’re seeing it. And we’re just blown away by how it’s going to be able to change our lives.
But I’m actually going to ask Asha a question that’s not specifically going to say AI in the question. It’s not going to actually say. I want to talk about cloud native for a minute. This is something we spend a lot of time on, cloud native development is hot. What are the updates you’re going to bring from Build or what were the updates that you had come from Build that you think are driving to make cloud native development faster and easier?
Asha Sharma: Yeah, I mean, it does go to the point that you talked about, though, with AI feeling like it’s part of every application at this point. It feels like, increasingly, applications are broken or feel broken without it. And it also feels like every developer is starting to work with AI, which is different than what it was even a year ago, where it was more experimentation. And so now, as we’re starting to think about how do we get this to scale, and we look around, Microsoft has the best IDEs, the most loved IDEs. So we’re spending a lot of time thinking about, “Okay, how do we make our tools accessible inside of those?” And so we’ve been working on rolling out a number of announcements.
So we announced the preview of GitHub Copilot for Azure, so that’s the ability to just interact with your Azure services from wherever you’re comfortable, VS Studio, VS Code, even GitHub. And so that’s going to be a huge step forward. We also announced AI toolkit for Visual Studio Code. So think about this as just an extension that provides tools and models for developers to start to acquire and run models, fine tune them locally, deploy them to Azure Studio, all from VS Code. And so, as much as possible, we want to meet developers where they are. We think that’s tapping into how they’re already building their application, providing more tools to make that accessible. And so we’re doing it a lot more. We’re going to see a lot of advancements in how we’re even bringing safety and security to that as well, but we’re pretty excited about it.
Daniel Newman: Well, listen, you went a little quicker than I expected, Asha. I was waiting for the rest. But you know what? There’s so much to digest. I thought that Build was an absolute momentum builder for you guys. And while Pat and I tend to look at panned Microsoft often, I think the work you guys are doing to bring the developer community forward is going to be so critical. And of course, building on Azure, we’ve given credit over and over again, Patrick and I, that Microsoft saw the opportunity, the AI inflection, and has completely reinvigorated and really become the company to watch and the company that is leading in the AI space, and congratulations on that. We all know that there’s an inflection so many times, and the rates between inflections gets shorter and shorter, and at each one, there’s an opportunity to rise. And so Microsoft came out quickly, has been very definitive and very decisive, very interesting what you’re doing with PHI-3.
We’ve talked a lot about the big models and everybody’s super focused, but we all know we can get really high fidelity with these small models, and that creates so much excitement for what we can do on device, and even just power. Even just lowering the amount of power that needs to be used, because we all know what a massive problem that is going to be, that unfortunately no one on this call is going to solve, no one on this video is going to solve, but that does need to be solved. So I want to say thank you both. It’s been wonderful to have you. Congratulations on all the announcements at Build. I’d love to have you back on The Six Five soon, if you’re willing.
Eric Boyd: Yeah, thanks for having us. It’s always great to talk to you guys, and you’re totally right. So many exciting things happening across all the things that we’re doing, and we just love to share with our customers and people who want to figure out. It’s just so exciting to see the things that our customers go build, right? Things that you could kind of dream might be possible, but we’re just putting the tools out there and people do incredible things with them.
Daniel Newman: Well, you didn’t call it Build for nothing. So Eric, Asha, wonderful having you. We’ll do it again soon. Thanks for joining, and thank you everybody out there for tuning into The Six Five. We’re virtually, as Pat likes to say, on the road, and we appreciate you tuning in, subscribing, and staying with The Six Five for all of our great episodes, this one being the best one you’ve watched today. Take care, we’ll see you all soon. Bye-bye now.
Author Information
Daniel is the CEO of The Futurum Group. Living his life at the intersection of people and technology, Daniel works with the world’s largest technology brands exploring Digital Transformation and how it is influencing the enterprise.
From the leading edge of AI to global technology policy, Daniel makes the connections between business, people and tech that are required for companies to benefit most from their technology investments. Daniel is a top 5 globally ranked industry analyst and his ideas are regularly cited or shared in television appearances by CNBC, Bloomberg, Wall Street Journal and hundreds of other sites around the world.
A 7x Best-Selling Author including his most recent book “Human/Machine.” Daniel is also a Forbes and MarketWatch (Dow Jones) contributor.
An MBA and Former Graduate Adjunct Faculty, Daniel is an Austin Texas transplant after 40 years in Chicago. His speaking takes him around the world each year as he shares his vision of the role technology will play in our future.
