Hello, I'm John Furrier with the Cube. Welcome to the Cube podcast. This is the NYC Wired program, a Cube original, part of the Cube family and also the NYC Wired community headed up by Brian Bowman. It's an open network of leaders coming together. We have Cube alumni and kind of a contributor for the Cube. He's almost like our own reporter, Howie Xu, but he's also the Chief AI Officer at Gen Digital. Pedigree goes way back to the early days of Cisco, invented many things in technology. If you know the Cube, you know Howie Xu. Howie, great to see you here in our studio.
Hi, John. You know, so so glad to be back to the Cube and the New York Stock Exchange.
I mean, we were just talking before we came on camera. I wish we were rolling. We were just having some good chats about some of the historical perspectives. And you've seen many ways. We've talked about this many times on the Cube. Big go just went IPO yesterday. Former Netscape CTO.
Which is the building I work in day in and day out now in Mountain View.
I mean you got former Netscape and Netscape was a cultural revolution. AI and crypto revolution, cultural revolution in crypto. AI is going through a cultural revolution as physical AI is right around the corner. But it's so hyped up. It's so changing. I mean we had the wave how we paper. You write stuff down on paper. You mail stuff and then the internet came along. Then you got mobile and cloud. So that generation was really a great movement. Everything was invented. The computer industry was thriving and scaled up beautifully.
Yeah. And then now we started to see that transition. We saw crypto early. We saw some AI stuff happening. But it really go back five years, you know, as the AI stuff came out of Google and other academic areas, ChatGPT came online and the rest just has been a barn burner of activity. This wave is you almost started okay transformer AI ChatGPT to autonomous everything.
Yeah. Okay. Not quite yet but we are getting. But this is the similar wave print to mobile and cloud with internet in between that was a cycle. This super cycle is AI to move along and then at the endgame fully autonomous stuff, physical AI, digital all converging together. That's kind of well that's talked about in the industry in that context but it's happening. So I want to get your thoughts. You were on a few months ago on our AI leader series. You're very active in your job obviously as Chief AI Officer but you're also out and about talking to entrepreneurs in Silicon Valley. What are you seeing? What's the big trend line observation sentiment around where AI is now?
Yeah, I think you know you see conversations on different sides of the spectrum right? On one side it's the AGI right? You know building huge data center right? You know that's the one end. The other end is is that an AI bubble going on right? You know it's like there's a conversation about OpenAI sure $20 billion revenue is big right? But still they are committing like what $1.5 trillion maybe by the end of the conversation $2 trillion in for a build out. Is that too much? Is that a bubble going on? And then everywhere in between right? Personally from my point of view you know there is a you know AI is real you know we see it. I mean from the ChatGPT moment we realize that AI we've been talking about for ages finally it's the you know reality now. It's the question is how to bring this you know technology to marry with your in the enterprise world marry with your data. In the consumer side how do I change the way we live right? So it's about you just marry the technology to reality at this point. I think we are three years into ChatGPT moment and I would say you know it's already big but give another three years upside down.
I absolutely agree. I like how you said that bringing AI to the reality and that's physical that's our life. So on consumer side AI into our experiences. I mean I was interviewing Frontier Digital Health they are saving lives. I mean that's as real as it can get. We're seeing surgery. We're seeing retail experiences. So there's no doubt that the physical nature of what we do as humans, we go to stores, we go, we walk around, we drive, we drive around, we have families, we do things. So that's clear. Enterprise, they're in business to serve customers who are users in reality. So they're moving around, right? So let's break those out. So let's start with the enterprise. We'll come to the consumer. I think they're more fun, more active. But the enterprise has been struggling because not struggling in the sense of momentum. The strategy is clear. Infuse AI in your business. That's a strategy. No debate. How to do that? The execution risk is coming.
I know where you're coming from, right? You know there is this MIT report that said 95% of the budget or the experiments didn't go anywhere. Gartner you know reports none of those reports show the super optimistic sort of the result out.
I didn't like I didn't like that MIT report. I'm against that because experiments generally do fail one number two we have enough evidence we've just on the Cube interviews we've done where the benefits of people who are doing agentic and agents and AI actually have done it right they executed knocked down some wins. And whether it's a saving life today or increasing profits for a retailer like Zebra Technologies and others, it's working. It's not a fail. It's not a category fail. It's just where it is. So, you know, you speak my words. I'm not going to repeat. So, I'll just give you my own insights into why that's the case, right? There are two things going on. You know, by the way, it's not nothing new. This has been going on for the last three years since ChatGPT moment. One is technology still need to mature, right? You know when it first came out the reasoning capability wasn't there but until you know September the o1 model out right? Now we finally the model has the reasoning capability but it will take a few more iterations to get to a level that the technology is mature enough to do a lot of the not just less hallucination but also can take actions with higher confidence right? Enterprise think about it right? You know people want higher accuracy unlike consumers when something is wrong hallucination haha Right. Yeah. Try again.
Yeah. Enterprise fraud detection. We missed it. Oops. So that's the technology side, right? But the more important part is the data. So, you know, the AI is not going to make a huge inroads into enterprise unless you get married the technology with the data. But what does the data mean? Right? Data is in different departments has all sorts of the tribe knowledge. You know even with and without AI we have a hard time to understand what data do we have. Do we have the comprehensive data? Do we even have the right data right sitting there? So without the context without the data AI is not going to make a huge difference. So now we are at a point that technology is getting mature to a level that you know I wouldn't say it's not a bottleneck it's a lesser bottleneck. The bigger bottleneck is getting the data. How to marry the AI versus the data is I wouldn't say it's a science. It's not a playbook. People know left and right. Every enterprise is different. So we are in that stage. But I have no doubt.
But that's just the momentum of where it is. It's not so much a vote on viability.
Exactly. So that's why when MIT report came out, I was like I mean why do you even report that? Right. We're in the middle of doing that.
Yeah. It's all failing. Like this is expected if Edison failed 10,000 times before he invented electricity.
Exactly. You know by the way electricity didn't make a real dent in humans life for 20 30 years after it was invented. So I think you know this time it's not going to be 20 30 years.
So it's not a bubble in the sense of what we know as bubbles. And the most famous bubble that we live through was the internet.com bubble. And that popped and then was carnage everywhere. People lost their jobs, families were impacted. It was just a terrible 101 was empty. It was terrible. However, there was no the demand curve wasn't as strong because people were on boarding to the internet. Mary Meeker used to have her famous, she was the analyst at Morgan Stanley at the time. She had a famous chart that she would chronicle all the online population, internet population, user population. And it would go up and because the internet was great, right? Who didn't like the internet, right? So the web was growing and so that was fact and all the fiber laid down. I don't think AI has a demand problem. Okay. I think AI doesn't have a demand problem partially because a lot of things we wanted to be autonomous is what we are otherwise doing today anyway.
Oh here's my point. So I want to get your reaction because if the internet built faster everything that might have accelerated a little bit of the online population but today my premise is if you have great AI it will be consumed. You can see that with ChatGPT. So your point about the consumer side which I'll want to pivot to is that we all love it. It's changed our lives. People are getting better at it and adopting it. They're infusing in how they do business. I was talking to some folks in Silicon Valley recently and they're building a platform for ad insertion. Not like a display ad but like data. So why wouldn't a brand want to be available because people are using ChatGPT in their daily life? So there it is. So okay, why if that's where the eyeballs are are the people?
Yeah, I think we're still bottlenecked by the compute resource and as we all know computer resource is also bottlenecked by the build out of data center electricity that kind of things and every you know whether OpenAI Anthropic they all went at least order of magnitude more in terms of the compute capacity in the last two years or alone right? So I think you know my premise is AI technology is still at least order of magnitude if not two order of magnitude or too expensive. Arguably too slow, arguably, you know, it needs to be a lot more accurate, but I see it's coming. All right. It's not like there's no there's no.
I would point out too and you saw this that the internet was a consumer first, enterprise second, even though people had internets. I mean, I was working at HP in 1995 and they think they had the first intranet. I mean, it's basically DNS. They had all the internet technologies essentially an intranet in internet. Yeah great they had an internet big deal they click on a document but the enterprise adoption on the web was much slower than the consumer so the question for you is as you talk to developers and young the young guns and seasoned pros who are coming back what are some of those consumer things you're seeing because in the internet we had breakout brands Yahoo eBay the list goes on and on and with the cloud mobile bubble or growth we had Airbnb, Dropbox. So SaaS created new brands that were basically kids in the dorm room, kids coming together, entrepreneurs. Will there be a breakout in the consumer? Do you seeing anything? I mean, is that the pursuit of the holy grail by the young developers?
We're not seeing we're seeing people use AI, consumers using AI tremendously more today than two years ago, than three years ago, right? But whether you like it or not, ChatGPT or maybe Gemini a little bit more these days.
Anthropic too. They've got some revenue.
No, Anthropic is more enterprise for consumers. It's predominantly ChatGPT and then Gemini. If you count the consumer include if you include the developers then you know the Claude of the world you know they have a lot more adoption but consumers and then consumers. So we don't see too many very big variety of the consumer AI native applications. Part of the reason is ChatGPT is pretty good. Gemini is pretty good.
It's very good. They're very good.
They're very good. Right. Think about Gemini. They already connect well they as long as I'm allowed to you know it can connect to my Gmail connect to you know so it's a very good chat. It's not just a Q&A. It can do a lot more things. So these two applications are pretty good. A lot of time people have a new things people say well isn't that just a ChatGPT wrapper right? So that slow down the innovation a little bit but I don't think that's going to be the norm for too long. My prediction is in three years there will be you know just like Instagram for mobile world YouTube for internet there will be new applications that's not tied to the old guards. I think I'm pretty sure.
And you think you think there'll be room for people to come and sit next to OpenAI on the app side.
Yeah, absolutely. I'll give you an example which is dear to my heart. Browser, right? You know, we discussed the Netscape. Netscape is generation one. Of course, IE sort of eventually took over because.
Internet Explorer people don't know IE is Internet Explorer from Microsoft for people do not know. The monopoly who beat Netscape. That's a historian. That's a we'll leave that to historians. But yeah, Gen one. So gen one, Netscape to Microsoft Internet Explorer. Gen two, it's the Google's Chrome, right? Which started taking over the browser 2008, 2010. By 2012, 2013, it became the dominant browser. That world didn't change for almost 13 years.
Yeah, it's time for that change. Now, you may argue, well, Chrome may change, but let's leave.
No, no, I think I think you're on the right. I think I like your path because let's just kind of chronicleize what happened. Put this is good. For the folks who don't know with a little history lesson here. So when Netscape came out with the first browser, everyone went oh wow. And by the way, their stock popped. They became the most valuable company at that time. It was very OpenAI like which sitting the building went IPO.
I know that building very well. Been there many times talking to the Netscape folks. But let me explain. So everyone was so excited about Netscape. Mark Andre my friends Greg Sand he wrote the business plan for Netscape the founder of BitGo he's from Netscape as I mentioned so Netscape was the winner they were the OpenAI they were at the time internet the gold rush blah blah blah blah blah okay in comes Microsoft they were OEMing Spyglass browser and they were building their own Internet Explorer IE3 Joe Belsky and all the teams there was knew those guys too what they did was they tied the browser to the Windows the system software. Windows is the operating system as you know for the folks out there. So we saw the beginning and that was considered bad behavior and the government the DOJ broke up. I have this debate with Dave Vellante all the time. I think that was about other collateral damage. Microsoft was told that they were a monopoly. Netscape complained. Jim Barksdale hired the lawyer. I forget his name and they tried to break up Microsoft. So that was the beginning of what now is a feature. So tying the browser to the system software, the OS, Windows, was an advantage. Actually, genius by Microsoft, competitive strategy 101, you know, job's not done till Netscape doesn't run. That's what they did and they did that and they were broken up. Fast forward to SaaS, tied to the cloud, not necessarily the same company, but kind of an operating environment, large scale, horizontally scalable, don't need a data center. We saw that kind of nestedness. So, okay, now comes the AI era. So, I like the thesis. So, Chrome, I was in the room when Sergey and Larry launched it and Sergey I asked Sergey directly, is this going to be tied to Android and Google as system software and he and the PR people were coaching him, don't answer the question. And he smirked and Kushi was right next to me in that meeting. What did you ask him? and he smirked and I knew it was and what Chrome ended up doing was syncing up with all your services Gmail. So again scratching the surface a little bit gunshy with all the regulatory they didn't flaunt it they got away with it but I think.
And they flourished for two reason one is the you know technology advantage right they were much faster than IE when it comes to JavaScript kind of the thing but the other thing is they still play the distribution game because eventually they won because of the distribution.
They had distribution it was faster so better product they had the distribution and the nestedness with Google services and at that time Safari and Firefox were the dominant browsers if you remember everybody remember Firefox is still around but okay so Chrome I so I like where you're going so now let's push forward to the AI era OpenAI I like that browsers have this native systems hooks call it hooks data hooks OpenAI hooks look at Perplexity people who use the Perplexity browser love it so I think the browser wars are going to be back.
Oh the so take me through your thinking this is third browser war, right? The IE versus Netscape, Chrome versus IE. Now, it's kind of the everyone else versus the Chrome that the the the browser the third browser war is going on. I don't know what's your definition of AI browser. Unfortunately, a lot of people tie AI browser with agentic browser. I think it's wrong. Agentic browser is okay. Hey, you know, doing some funny stuff like, you know, type a command, it will open the tab, you know, log on to my amazon.com, do things, right? I think it looks fun, but I don't know it's really useful.
But that's agentic browser. I don't know it's useful enough today, but there is another kind of AI browser definition, which is my definition for foreseeable future, which is a companion, meaning that you do all the browser stuff. There's AI be your companion and then make a recommendation for you remember things for you be your brain be your not be your brain be your partial brain be your partial memory so that you are augmented.
I think you're right and I would let's unpack this is kind of in the frontier of what's happening the word browser came about because the web had a browser you browse sites and that was the user experience of the utility yeah you go to a website you browse and you click and you do okay we're not in a browser utility. We are in a user utility of getting things. So browsers would be discovery and navigation. Discover a site and you navigate to a destination. That's the search engines. You click destination URL navigation. These are words that are used in that era.
By the way, I would have broadened that definition even. I want to say browser is my entry point to my digital world. Whatever the entry point means. It could mean navigation. It could be my memory. It could be autonomously doing things for me when I.
Well, I'm saying browser is an irrelevant term for what you're saying because what you're saying is we do things differently now because what we do with the internet is not just browse. Yeah. It's native. I have now a lot of applications. I have a lot of applications that I use. They have different identity systems, different utility, but I also do a lot of things that are kind of horizontal. I could be at work. I could be at play. I might want to buy movie tickets. I might want to go to a play. I might want to go to an event or talk to a friend. These are utilities. This is like internet usage. So I think the word browser. So okay if you.
I view browser as the AI era consumer is operating an interface to the world. Interface to the digital world that let's just leave it at generic call it Howie. Okay. That application. So if you believe that then you say okay what's the underlying system software. So if we have if we're tying in this the advantage of native experience I think what Perplexity is doing is you get down to the root level of data you want to get at the data. So the question is what is the system software of the internet? Now in the AI era it's a multitude of language models.
You got cameras on everything. You got autonomous vehicles. You're situation. You're contextually situational, right? Connect to your personal data or maybe if it's a browser for the enterprise, your enterprise data, right? You know, tool use the connection to the data. Not just answering question for you but also sorry loading a page for you answering question for you be your chat but also taking actions for you.
Yeah so look I like the companion angle I think that's a great way to start thinking about it but there's more okay so if that's the case lay out the playing field if we were sitting here advising OpenAI or starting a company what would we do what would be the land grab to me it would be like okay reinvent a new app that consolidates everything it could have agentic. But it's got to be personalized. I mean, all the health care benefits I'm covering in AI is personalization, precision medicine. So, precision, personalization, accuracy. Where's my data? Do I store it somewhere? So, there's a lot of things missing in the picture of what could actually work. I mean, that's my I guess that's my question. What's required to make it work in your mind.
I think it's similar to the enterprise world. I said that there are two things that's sort of I wouldn't say bottleneck that's sort of slowing down the adoption. One is technology maturity. You want the gen AI to be having better reasoning capability, less hallucination, right? That's technology set. But you also want the technology to have access to the right data at the right time. Same thing with consumers, right? So think about it. I do so many things every day. Can AI help me to remember things, right? Can AI help me to distill the insights, right? That's sort of the data access or whatever my eye sees hopefully AI see the same and then can amplify the insights then human would do so that's that angle but the other thing is the technology side you know the multimodality right? You know the accuracy I think we need the both advancement.
I want to talk about OpenAI we were talking and you and I were talking before we came on here this pod episode around OpenAI and we're talking about job loss what people fear for their jobs you've been hearing some rumblings or observing that more people.
Yeah. Asking me what did I hear during the holiday right before this show. You know my biggest aha moment I want to say aha moment the biggest surprise to me is I met so many engineers researchers from Meta OpenAI Anthropic Gemini of the world right? You know dozens of them in the last two months and of course I met so many other people and the surprised thing is people from OpenAI Anthropic they worry whether I still have a job three years from now far more than the rest of people far more.
What are they worried about?
Because I think they see that the technology can do so many wonderful things right? You know it can even do research it can you know build models right? It can do coding sure you know coding is still not mature but guess what you know at least Anthropic the CEO says Dario says 90% of code is done by AI so I think within Anthropic they see that being more real than people outside of Anthropic. I actually invited even one Anthropic engineer to my team you know and I say hey you worked at you know Microsoft one of the division not long ago and you were using Cursor of the world now you kind of at Anthropic why you are telling me that you have a very different view and he told me the following he said before I see oh AI can do this AI cannot do this okay but today I don't think about that I said how do you what do you think he said when AI cannot do X. I think about how can that make AI do X? What context does it have? What data does it have? What improvement do I need? So by the end of the exercise, AI can do X. So what he said, what I see is AI can do everything. It's a matter of how I'm going to make this happen. But before I'm just thinking, oh, AI can do this. Okay, then great. But otherwise, I do whatever AI cannot do. My mentality shifted a bit.
So they're enabling AI. They're marrying the data to AI. That's just the one big part.
And I would say in my Nvidia coverage the one thing Nvidia is doing right now.
