#4: The AI Enterprise

When you're sitting down with a technologist at a company, how do you get started? Because there's a lot of businesses that haven't even gotten started with their AI effort. A lot of companies are budgeting for next year, so they're planning on hiring maybe somebody to run their AI effort or something like that. What would your advice be to that early effort company?

Yeah, I think the challenge is this. It's sometimes as technologists we get wrapped up in just trying to use technology for technology's sake. Before this whole massive interest around AI. If we rewind five years ago it was blockchain, right? It's like, oh well, I need to use a blockchain for this and I have to use blockchain for that. And it's like, actually you don't. You can use a database for that and it's fine. And you have to resist that with AI too. It might be academically interesting to apply AI to every use case. You probably don't want to. So the way I like to approach this and working with different VMware customers and partners is look at the problem statement. What is the business problem that you're looking to solve that you think you might be able to benefit from AI? Now let's work backwards on what is the most cost effective architecture to support that. I'll give you a practical example. I've been working with a retailer and they do a really interesting, they have a really interesting AI use case. The common one you hear about often in retail for computer vision is I want to prevent theft at my self checkout. But this retailer in particular, what they're doing is they're looking at observing customer behavior in each of their different aisles. And what they figured out is if a customer is spending too much time in front of a shelf, they're probably just need some help. Like for me, if I'm going into a home improvement store, right, it's like if it's plumbing, I am like just staring like I don't want to have to go back, right? So I want to try to get it right the first time. But if an associate would come over and just be like, hey, what are you working on? Okay, this is probably what you need then. Awesome. And this particular retailer, what they've seen is a, like a tangible increase in sales since they put this technology in, which, you know, I think is a great use case. And then architecturally they're just running a couple of very small servers in each store and that's how they're rolling this out. So they don't need a whole lot of infrastructure or capacity for a use case. That's driving a lot of really strong business value.

It's an honor having Chris Wolf from VMware here. VMware is a company that a lot of people don't know about that runs underneath a lot of big companies virtualizing their data centers and helping them manage the data centers and doing a whole lot of other things. And so let's start with my usual question. Who are you?

Yeah, so I'm Chris. I lead AI Labs for VMware. I've. Let's see, I've been working at VMware coming up on 10 years and I've been working with VMware products for coming up on 25 years.

Now VMware and Nvidia just announced a deal. We should talk about that. But VMware has a unique place in the world of AI. VMware is being used by all these big companies and I was honored to be able to speak to your top customers. And these are people running huge businesses all over the world. What are you seeing? Give us a state of the art of AI because you're, you have a viewpoint that is very different than most people that I'll have on my show, let's put it that way.

Yeah. So I think a couple things. First, it's definitely. We've seen in this calendar year especially AI has really taken off in terms of just interest in exploration of use cases, things like that. You know, at VMware, we've been working on AI technologies for the better part of the last five years. I would say five plus years is really where we've been involved. And in my particular organization, we started to really hone in and focus on AI just exclusively in about the last year. So other areas where we were investing in, such as new products for Edge and things like that, we've moved them just into other businesses to focus on. And to answer your question, what are we seeing? We're seeing a lot of interest in use cases from customers looking to get started. But at the same time, the concerns about, I don't want to place the wrong bet. There's all of these technologies out there and I don't want to just go and buy one platform where, where there's so many things happening so fast. I don't want to have buyers remorse in say three months or something like that.

Yeah, my, my thesis is that a lot of the big businesses have trials using large language models. Large language models have changed the game for sure. I have a list of 400 companies who are making developer tools and almost all of them are using large language models in some way to help businesses. Right. What are you seeing? What's the pain point? Because I keep hearing from business owners that I'm scared of my intellectual property. I'm scared of the hallucination problem where it makes up stuff. And there's a few other problems. If you're trying to build a AI that you're talking to, the latency is a problem. What are you hearing from these big customers and what are they trying to do? And how do you think that's going to change over the next year?

Yeah, yeah. So definitely the problems that you've described. One of the things about AI and generative AI in particular is that it doesn't just get things wrong, sometimes it's confidently wrong. Right. So it's like I just say something with such authority like we are when we just laid out the llama2.70 billion parameter model as a service internally. First question I ask is, who's VMware CTO? And it literally just made up a name. Like I don't even think that name has ever been in a VMware employee before. So, you know, pretty, pretty interesting that that happens. But hallucinations, as you mentioned, that's a concern. Privacy and control of data and intellectual property, that's also a real concern. You know, organizations don't want the data that they're using to train a model to also benefit, say their competitors, which are things that has happened and is a certainly a concern for organizations. And I think the other one is, to my point on buyer's remorse is the space is moving so fast that there might be this amazing model that comes out tomorrow that I had never even heard about. So thinking about building in like an AI services platform that gives you optionality going forward is really important for a lot of folks.

Yeah, tell me about it. You know, just last week I got multimodal so you can aim a chat GPT at anything in your world. In fact, I did the spice rack last night and said, hey, here's a picture of my spice rack. Can you make me an inventory? And it did right. Of everything it saw on in the photo. This shows that there's new capabilities coming for both enterprise and humans, right?

Yeah, for sure. And that's the key. I think there's all of this fear around AI, but it's a tool. It's funny if you look back, like back in the 80s, it was the personal computer was going to displace the office worker. I think office workers are doing just fine. Right. It's like learning how to use a tool more than anything else. It's the same thing with AI. AI is a tool, and that's something that, like even VMware, when we were looking at the AI services and our own products, we branded them as intelligent assist because we see AI as something that's assisting the human, not ready to take over for the human. Which is, I think the fear that a lot of folks have.

The deal that you guys announced with Nvidia, and I heard Jensen talking about why he did the deal with VMware is really interesting because if you're going to build a model of any size, you're going to need a bunch of Nvidia cards in a data center to build that model.

Right.

A lot of these companies like OpenAI or Tesla or other AI companies, they have 10,000, 30,000 Nvidia H100 cards sitting in a data center. And I believe it's VMware's virtualization software that joins those all into one big Nvidia card that we can use as a model builder. Right?

Yeah. So I think there's a few things there. One of the things I think that's important is when folks think about AI, the first concern is, oh, wait a minute, I can't do AI because I need hundreds of thousands of GPUs. And it's something that we found not to be the case in a lot of use cases. For example, when we took a open source model for software code development, which, which was Hugging Face's Starcoder model, and it's a 15 billion for parameter model just tuned on permissively licensed software code. We went and then tuned that against some of our own internal C code that we've been using to write our core platform for decades now. And we were able to tune that model using two GPUs in half a day. So that was a, that was probably the first aha moment for us, was like, wait a minute, we don't actually need to have all of this capacity to take advantage of AI. So that was pretty exciting. But the other thing to your point that we've seen in working with customers is GPUs are scarce and when you are using them, you really want to use them efficiently. And with our software, you're able to virtualize the GPU or multiple GPUs, pull them together and then just share them among different applications, different data scientists. So you're getting the most effective use of the capacity you have and that's been huge for organizations to lower their total cost of ownership and then just to make AI more accessible to their different applications.

Yeah, it's a new world. I've been talking to a lot of people who are building systems for big companies and one of the trends I'm hearing is they're building orchestras of AIs. In other words, they might have multiple large language models arguing out an answer. Right. Or taking a different point of view, something like that. And then they're building smaller, much more fine tuned or much more specifically trained models on their business so that their employees can ask very specific questions and get right answers. With your CTO answer, if you trained it, a smaller model that has that answer, maybe the bigger one doesn't need to hallucinate so much. Right. And orchestrating those together with all, all sorts of software to make them all work together. Are you seeing the same thing at vpn?

Yeah, that's something that really excites us is that there's all of this attention on these mega large language models that yet take a massive amount of GPU capacity and often months to train. But when you get into the smaller models and smaller models that are tuned to particular business functions or processes, you can train them and over a weekend or in the evenings when you have some idle capacity, things like that. So the benefit is, I think there's multiple benefits. One is you are able to train more frequently so you can get really accurate results. And then second, also equally important is you're doing so in a way that's very less impactful to carbon. A mega LLM takes like massive amounts of energy to train and these smaller models, they don't, they're much more carbon efficient. They're, they're power efficient, so they're more quickly trained. There's a lot, there's a lot to really look forward to here. So this is a very quickly emerging space and something that we believe in. We've seen great results from our peers such as Microsoft Research. So there's a ton of momentum here and I think that's going to be pretty exciting for the future.

Yeah, it sure is. What are you seeing as the early pattern that's starting to happen? I'm sure some of your customers are getting fairly far along in their, in their AI trials that they're trying to do. What are you finding is working? What is your advice to other businesses that are just starting their path of having to try to figure out how to use AI in their business?

Yeah, yeah. So first what folks are looking at is where are some use cases where I can get some quick wins. A lot of CEOs even have board mandates where they're looking for some usage of AI this year to help fuel the business. And in the quick win category that we're seeing a lot around is like marketing and content creation. It's a great use case for AI and generative AI. Maybe they're giving me 10 options that I can choose from and that's just starting the ideation process where you can land on something. At the same time though, an interesting observation that I've seen in that use case is I had met with the CIO of a retailer just in the last month or two and he was using this AI tool or AI service to create customer emails. And then the weekend before I met with him, he had gotten an email from his competitor and cause he was a customer of theirs too. And what he saw was the language structure, the sentence structure, the tone of the email was exactly the same as theirs. So he's I gotta rethink this because now we're just vanilla. We sound just like our competitor does to our customer. And that's because we're all using the same AI tool. And that got him thinking about how do I start to take a, like an open source model and just tune it against my own customer data so that my voice is preserved. Right. So, so that's a pretty interesting use case that I think is going to. Beyond that early adopters, you're going to start seeing more customization and more personality between different companies in terms of how they're approaching that. Other use cases we've seen that have really taken off the bin. Code development is starting to gain a lot of traction. Enterprise knowledge base and enterprise search or customer service and support, where I can put a chatbot in front of all of my documentation and all of my knowledge base articles and Goals. Even for VMware, we're working on our own internal large language model for VMware support. And we're hoping if we can even reduce support calls by 3%. That's a huge savings to us. That's a huge savings to our customers as well. To hit these types of numbers.

It's getting really good. I just saw an app, a company called Hippocratic AI and they're building an app for the healthcare system. And I talked to their system, right? And it was like talking to somebody answering the phone. A human. I couldn't tell it wasn't a human. It was crazy and it Knew all sorts of things about. In this case, I was setting up a colonoscopy appointment. Right. It knew all sorts of things about where to park, what the procedure was, what the fears were. I was asking all sorts of random questions and it was answering them real fast. But they had done a lot of training with 10,000 nurses answering questions. Right. And they used that data set to train their mo. A very specific, smaller model to be very accurate and very specific to that use case. Right?

Yeah, definitely. Key. And then also to to that point, serious concerns around HIPAA compliance. Right. And patient privacy and really has to get these organizations to think about how are they even architecting AI. AI to assure those levels of privacy.

Yeah. My, My psychiatrist uses ChatGPT to listen to her therapy sessions in a HIPAA compliant way. So the first component that her software runs is a system that takes out any personally identifiable information before it gets sent to the large language model up in the cloud. So that HIPAA compliance is there a lot of new change, a lot of new thinking that developer teams have to do. Let's talk about cultural change because VMware has a huge sales team, works with businesses all over the world. The fears, there's a lot of fears here. Right. Beyond the technology fears of it's making up mistakes or, or might steal your intellectual property. There's employees are really freaked out by this. Might steal my job. Right. So that's a huge resistance inside corporations. How are you seeing the corporate, the cultural change efforts going on inside these companies to introduce these new technologies to normal people and normal work?

Yeah. And it's. The fears are, are definitely there and, and they, they should be. I think even if you go back to a more recent example, think about like how the Internet was disruptive and how companies like Google came along and really disrupted Internet search and people that very quickly were able to figure out how to use Google as a tool became more effective at their jobs. And it's the same thing with AI. There's. When you mentioned hallucinations, folks that still have expertise are going to be needed to understand this is what was incorrect. Folks that lack expertise in the discipline are just not going to be as valuable with those types of tools. Now in cases such as sales or legal contract generation, maybe I do need fewer people in these types of fields because there's more automation that's happening, but I still need to have expertise that's able to review the content that's created, make sure it's in line with the company compliance, things of that nature. And in Any kind of negotiation, you're still going to need a human involved as the arbiter, certainly for the time being. So yes, there's going to be some disruption and some jobs will change, but that's part of, part of life. I think, at least in like my experience, the first job I had in technology was doing component level circuit board repair, like figuring out which transistor was bad on a printed circuit board. And that's a specialty that's obsolete. Right. So if I had held on to my early tech expertise, I would have challenges. But it's just the world evolves and we just have to evolve with it. And I think it's the same thing regarding AI.

Yeah, What? When you're sitting down with a technologist at a company, how do you get started? Because there's a lot of businesses that haven't even gotten started with their AI effort. They've been hearing about chat GPT and they haven't really geared up or budgeted yet. I see end of the year a lot of companies are budgeting for next year, so they're planning on hiring maybe somebody to run their AI effort or something like that. What would your advice be to that early effort come from?

Yeah, I think the challenge is this. It's sometimes as technologists we get wrapped up in just trying to use technology for technology's sake. Before this whole massive interest around AI. If we rewind five years ago it was blockchain, right? It's like, oh well, I need to use a blockchain for this and I have to use blockchain for that. And it's like, actually you don't. You can use a database for that and it's fine and it's the same. You know, some of that you have to resist that with AI too. It's not like while it might be academically interesting to apply AI to every use case, you probably don't want to. So the way I like to approach this and working with different VMware customers and partners is look at the problem statement. What is the business problem that you're looking to solve that you think you might be able to benefit from AI? Now let's work backwards on what is the most cost effective architecture to support that. What is going to give you the most flexibility going forward so you can pivot as a business. Right. And start to work backwards from there. Like I'll give you a practical example. I've been working with a retailer and they do a really interest, they have a really interesting AI use case, the common one you hear about often in retail for computer vision is I want to prevent theft at my self checkout. Like everybody is trying to do that. And the results can be mixed depending on the retailer because of the cost of the services versus how much inventory is actually missed out of self checkout. But this retailer in particular, what they're doing is they're looking at observing customer behavior in each of their different aisles. And what they figured out is if a customer is spending too much time in front of a shelf, they're probably just need some help. Like for me, if I'm going into a home improvement store, right, it's like if it's plumbing, I am like just staring like I don't want to have to go back, right. So I want to try to get it right the first time. But if an associate would come over and just be like, hey, what are you working on? Okay, this is probably what you need then awesome. And this particular retailer, what they've seen is a, like a tangible increase in sales since they put this technology in, which, you know, I think is a great use case. And then architecturally they're just running a couple of very small servers in each store and that's how they're rolling this out. So they don't need a whole lot of infrastructure or capacity for a use case. That's driving a lot of really strong business value.

Yeah, that's a good point. The retailer perspective, it's interesting. You just remind me, I, I used ChatGPT to find a screw in a Lowe's and it knew what aisle it was on and how far down the aisle it was. Whoa. You have a lot of knowledge of the world that people really haven't figured yet. The problem is that use case has a lot of holes. It couldn't find me a coffee at the San Jose airport. Right. But it could find me a screw that Lowe's, but not a coffee at airport. So it has a lot of incompleteness. It's not quite perfect.

Right, right.

Are you seeing businesses start to think about using large language models to do all sorts of things like help their customer find things in their store or help employees help customers?

Oh yeah. Customer service is definitely like the top use case that we see. Everybody has that use case. And so for you in VMware, what we've been doing ourselves internally, we've been working to produce more documentation so that customers can take from what we've already learned and deploy their own types of solutions. So that one is just a no brainer. Because if you think about it like the human experience is like you're calling into a service agent and then what are they doing? They're sitting at their computer, they're running queries, right. They're trying to get you the answer you need. So that's something that AI could do and really start to help customers to be, you know, more efficient and then really have your customer service agents working on the hard problems, not the things that the AI could just take care.

Of in this new world of programming. In the old world, I used to work at Visual Basic Programmers journal and ran a magazine about Visual Basic. In that old world, when you send an API call to a piece of software, you always got the same answer back. With AI you can ask it the same question over and over again and you'll get a slightly different answer back. It's a very loose system that way. And your software team has to build systems that can deal with that kind of every answer is a little different because doing QA on that is a different thing. Right. In the old world, if you didn't get the same answer every time, you knew the system was broken and you knew where to look in this system, you don't know really what's going on. Tell me about the new architecture and the new software team and the new skills that a software team needs to deal with these AI.

Yeah, yeah, yeah. And it's not just like looking at output, it's how am I adjusting model weights. I'm looking at bias and models. Is there parts of the where the model's being overlooked? I other areas are explainability, like I want to be able to auto generate some explanation as to how the AI arrived at a particular answer so that the humans can feel more confident in what they're realizing. Understanding correctness of the model and how accurate are the models predictions turning out to be. These are all new skills that we're going to need. And still AI like a large language model. I still need an application in front of it, right. So to be able to get the data out of the model. So that's important prompt engineering. Understanding how the model is, can interact with the user and to be able to get the right information out of the model itself. So there is definitely this shift that's occurring. But at the same time again these are all new jobs in tech that are going to be quite interesting. Even in the legal space there's all of these unanswered questions when it comes to AI. So that's the other area where there's just a lot of. There's a huge void right now that that is going to get filled over time. But it's creating opportunity for people to step up and assert themselves in these fields as the next wave of experts.

This new AI system brings new security concerns too. Right now you have new attack vectors inside a corporation. If you have a large language model and you have to have a security team that understands how to deal with these and how to protect your systems. What's VMware's thinking about security? Because I know you've made some acquisitions in the past around security, that. Where is security going?

Yeah, yeah. So for us this is really key. We see our value in AI as being around security and privacy. We think it's important that an organization maintains both privacy and control of the data that they're generating that might be used to influence a model and then providing protections against that. So one area that we've seen both on the customer side and even on the AI vendor side is like customers want to have confidentiality assurances of their data, but then the AI vendors as well, they want to have confidentially confidentiality assurances of their models because that's their intellectual property that they're trying to protect. And this is where we see emerging use cases for confidential computing coming into the fold. Where I could work with say a company like VMware and vendors like intel or AMD and use a trusted platform module or a hardware chip that can ensure the confidentiality of the data as it's being processed or even being stored in physical memory, which we see as really important on the technology front. So that's one area that VMware is investing in. We've been doing a lot of investment in security to protect the APIs that are being used for AI, understand any types of zero day exploits before they're happening, having full context around how AI applications should work so we're more intelligently able to find malware and things like that. So it's quite interesting. You're seeing AI being used to generate malware and now we're seeing AI being used to defend against it. Definitely interesting times. As these areas start to evolve.

What else is happening from VMWares and you guys are going through an acquisition, Broadcom is buying you. So there's a lot of change inside your own company that you're probably having to deal with and maybe thinking about how to use AI to join together systems in the future, stuff like that. What else is happening from VMware's point of view?

Yeah, yeah. So we're going through this merger with Broadcom, and hopefully that's going to be wrapped up fairly soon. We think we're in the, really the home stretch here, which is great. And on the innovation front, I think what's exciting is we just haven't slowed down. We went fairly big around this private AI concept in just the last couple of months. And private AI is not something that is a term that VMware coined. It's been in our industry for, I don't know, at least six, six, seven years at this point. I'd say what we've done is we've really helped to contextualize it and explain that this is. Private AI is about an approach to architecture. It's preserving data confidentiality and data security while still providing businesses with access to what they need. And we see that a lot. Customers say, you know what, for my most sensitive data, I want to keep it within my firewalls. I want to maintain control of it. So a big part of what VMware has been focused on is building a private AI infrastructure platform that allows an organization to be able to gain the benefits of AI, but then have massive amounts of choice around it for either commercial or open source AI models and AI services, which we think is just how you do it. Because, like, we're our own consumer of AI services and we want to make sure that we have the flexibility built in. So we've taken a platform that we've built internally and what we've now done is we've made it available to our customers and partners as well to take advantage of. And we're continuing to share everything we're learning through our own internal use cases, through technology partnerships. And then there's just more technology investments we're making around those next generation of services for security, privacy, explainability and things like that.

One problem for development teams or executives is there's so many new companies that are selling product. I have a list on x.com of 400 companies that are just building AI development tools. And I have 400 a bigger list of 4,000 companies that are building all sorts of things from like writing tools, marketing tools, text to speech tools, stuff like that. How do you help these companies evaluate technology that they might be using? And separately, how do you evaluate the models? Because there's, like you said, there's a new model every week or every day sometimes in some cases. Right. So there's a lot of change that the corporation's being faced with. And how do they decide which partners are the appropriate partners, which technologies are the appropriate technologies to Use do you guys help with?

Yeah, we do. It's a good point. If I look at just ML Ops by itself. So ML operations and associated management, that's that ecosystem today is probably two to 300 different vendors. So I have that challenge. I'm getting approached on a weekly basis, probably three to five new software companies that want to talk with us and say hey, we want to partner with VMware. And the reality is for us less is more. So we have an ecosystem now of roughly 20 to 30 partners. And I'm being very selective in terms of who is new that we bring on because I'd rather do things do AI with a smaller group than just collect a bunch of logos. Right. So that's been our approach. So we will guide customers on. Here's the operations vendors that we've had success with, here's the software vendors and the open source models, et cetera, that we've had obsessed with. We're creating what we call starter packs that can help our customers and partners deploy these AI use cases on our platform using a variety of these other partner solutions. And we're also being just very transparent that we're also in learn mode. Like some of our sizing guidance is based on what we've success, what we've successfully deployed internally. But we know out in the wild different customers are going to run into different scenarios and we're going to adjust our guidance going forward and we're being transparent about that. But yeah, that's basically our approach. We can guide the customer on everything from the hardware and software footprint you need to what we've seen to be successful in the ecosystem. And again, everything for us is around just having a core platform and then optionality above it because we want to just win business based on customer trust and not by having to do anything unnatural to make our platform sticky. So it's upstream kubernetes, it's upstream Ray open source. It's just a lot of open source hooks into our platform that again gives customers the amount of choice that they would like, especially because this is new space and they want to be careful.

What a new world it is. Personally, what are you seeing that's got you excited? You know, I'm excited about this new chat GPT features of being able to aim the camera at anything and being able to talk to it about it. Saw a Ferrari on the freeway the other day. I took a picture of it and it started telling me, oh, all sorts of facts about the Testarossa that was in front of me.

Right.

It's Whoa. That's pretty cool. New technology, right? That might be useful in my life somewhere. What got you excited?

I think so far it's still, it's probably more than a year running. Like, my favorite AI application is plant identification, just especially for landscaping in our yard. Like, we live on the side of a mountain, we grow just a lot of wildflowers and we just let them do their thing. And so you're always still wondering if something's coming up, is this a friendly or is this a weed? Right. Is this something invasive that I have to worry about or not? So I think gardening has been, for me, pretty exciting. On the tech front, just the emergence of the community and ecosystem around Meta's Llama 2 model has been super exciting. And that's my earlier point on how we're going to see more of this evolution of these smaller and finely tuned models for particular business functions. That's going to really start to take off thanks to the work of Meta and others. Companies like Hugging Face and what they've been able to do in terms of open source model serving is also just super exciting. I love the fact that we're seeing so much choice, which is great. That brings out competition and that's an area where both us as consumers and also businesses are really going to benefit.

And to close this out, what should we be watching for in next year, particularly the enterprise customer. Right. What should they be planning on, working on, thinking about watching for?

Yeah, I think, I think you're going to start to see this emergence of steady use cases. To us, what we've seen very clearly is that like our launch around private AI is something that's really taken off and is sticking. One of our, one of the CEOs of one of our largest partners that said, the private AI, because your legal team requires it. So I think there's going to be a lot of, a lot of work around maturity of AI solutions and a lot better understanding from organizations as to how they can apply these technologies and how they can also make sure that they're safeguarding themselves. Even at VMware, we can't just, we'd love to just say, throw a ChatGPT interface in front of every product. We can't because we don't own the data that our products even sit in, our customers do, and we just can't arbitrarily make decisions on what they do with their data. To me, it's gonna be a growth year for AI. You're gonna see a lot of uptake in content creation and Customer service use cases, you'll start to see co development start to take off a bit more as well. And then I think that's where you're going to see organizations start to step back and say, okay, I made these investments. What did the business value look like? Where can I expand? And I think that's going to be the key. People have to resist the temptation to do AI for AI's sake and look at a particular surgical use case that's going to benefit their business and then learn from there and then start to expand after that.

You don't. The customers I talked to, when I Talked to the VMware Customer Council, they weren't the really sexy stuff like a Tesla robotics team or autonomous car team or something like that, or augmented reality glasses team at Apple. They're more people who are trying to run a factory, building something or running a healthcare system, something like that. Do, do you have any insight on the sexy stuff though? You know what do you have an autonomous car team like Nvidia's driving cars around Silicon Valley?

It's funny, right? We're Broadcom's acquiring us for around $70 billion. And it's funny because I get out in public and a lot of people have no idea who we are and are like, wow, how big is VMware? We've always liked to be the enabling technology that allows a lot of flexibility and choice. In that way we want to work with customers to do these exciting things. Like we, we're, we've helped like Dish Networks like really redefine telecommunications. And Dish started as a, as a pure play, more like TV provider. Right now they're a massive telco. We've had a huge part in that. Like they've revolutionized the way that telecommunications are delivered from all of this amount of hardware equipment at the base of a cell tower to a tiny little footprint that's highly virtualized, super efficient. We're working with the largest global manufacturing companies on factories of the future which are just purely defined in software and will even allow manufacturers to sublet out their factories so that I can maybe I have a partner that just needs to use some capacity for some time. So I have these efficiencies that are coming and just these entirely new ways of doing things. That's exciting. We've put in the last one I'll mention to you, this is in police cars. We've been able to put in technology that provides full redundancy of different telecommunications carriers. So if I'm thinking about like high definition video streaming from a car, say they're like in pursuit of somebody or whatever it might be for officer safety, or they're coming up on a scene and maybe dispatch has to automatically disperse another officer or something like that. We're able to connect these cars over multiple carriers so such as like at&t and Verizon so that even if you had a brownout area from one carrier, the strength of the other carrier signal is going to pick you up. So this is just helping officer safety. This is also taking a first responder car in the past that's had a lot of very old and legacy technology in it. So we're totally modernizing the technology footprint. And you know what, I'm sorry, I'm on a roll. I'm going to give you one more, but then I'm done, I promise. So the last one is think about the energy grid. We are, our technology now is going into the electrical power substations that power the global's energy grid. And again, an industry that was really very outdated in terms of the technology that would sit that was holding grids up globally. Our software is now coming in and creating this foundation by which you can now run all of these new modern applications, have better security, have a more resilient posture for how the energy grid gets operated. Again, you may not read about VMware in these use cases, but it's all of our underlying technology that's enabling these things.

That's awesome. Where should people look for information on VMware's AI efforts? Do you have a page or do you just want them to go to the homepage?

Yeah, yeah, our homepage is. I want to say, it's just if you go to via.comai that should take you right to our AI landing page and then you're going to see all of the things that's happening with the company and you know, our projects. My team is VMware AI Labs. So even if you Google VMware AI Labs, you're going to go here and that's going to be all the new stuff. So all the next gen stuff that we're working on is what you're going to see there too. So hopefully folks will check it out and then reach out and provide some feedback.

Very cool. Thank you so much for spending some time, valuable time with me today. And thanks for the efforts that you do for America, for the world's business.

Really appreciate it. Thanks, Robert. Thanks.