In The Arena by TechArena - Exploring AI Innovation with Hitachi Ventures' Gayathri Radhakrishnan

Episode Date: September 9, 2024

Guest Gayathri “G” Radharkrishnan, Partner at Hitachi Ventures​, joins host Allyson Klein on the eve of the AIHW and Edge Summit to discuss innovation in the AI space, future adoption of AI, and more.

Welcome to the Tech Arena, featuring authentic discussions between tech's leading innovators and our host, Alison Klein. Now, let's step into the arena. Welcome to the Tech Arena. My name is Alison Klein, and we are coming to you from AI Hardware Summit in the Bay Area, and I'm so excited to be with Gayathri Radhakrishnan. She also goes by G. She is a partner at Hitachi Ventures. Welcome to the program, G. How are you doing? Thank you so much, Alison. Doing great. Great to talk to you. Now, G, you and I have known each other for a long time, and I know that you're an expert player in the venture arena. Have you ever seen a time like this in terms of investment in AI? I have to say no.

I mean, the pace now, Alison, when you and I were in Micron, we had an AI fund. And that felt like we were maybe now in retrospect, I feel like maybe we were early. Exactly. And there were a lot of AI investments happening then, but now AI has become more ubiquitous. It's becoming more the norm as well. And with generative AI really taking center stage, and we've seen how 2024 has evolved. One thing that I wanted to talk to you about is as you look at this AI market, how do you see the adoption of AI moving forward in the next couple of years? And what do you think it's going to take to get enterprise to start adopting this

technology at scale? So it's a great question. So AI by itself, I think is already being adopted. But AI, as we all now know, it comes in various flavors. There's the computer vision, machine learning. There's transformer models, which have now evolved to generative AI. I think there's some question marks around generative AI, not from an adoption standpoint. Everyone wants to adopt it. It's around, is there going to be more innovation coming? Is this the right time?

How much infrastructure spend do I need to do? And people are still worried about hallucination, the quality of the output. Can I trust what comes out of the system? How do I have the system be more explainable? What does the regulatory impact mean? What does the data like ownership mean? So there are a lot of these side questions that are

cropping up. And I think that's where the hurdles are. It's not hurdles or opportunities, the way you look at it, right? So I don't think it's a matter of has AI been adopted? I almost feel like every company these days has some element of AI in it already. But what extent and how cutting edge is the AI is probably a better way to think about it, at least from my perspective. That's how I'm looking at it. Now, there are tons of companies who are adding AI to their marketing to attract broader attention. And, you know, to be honest, Tech Arena has joined that bandwagon by changing our URL to.ai this week. Thank you so much. It's a big move for us. You're in the business of

evaluating technology to ferret out unique innovation from maybe a lot of great PowerPoint that you probably see in front of your desk. What is the approach that you use to determine potential value? No, that's a great question. So every time when you see like a hype cycle, right, I saw this during the peak of cloud investing. Every company claimed to be a cloud company and they used to call it cloud wash. Only thing they didn't do was like sell CDs, but they would put it on a web server and create a login. And then they would say, oh, we are that time, we'd say, okay, are you truly multi-tenant? We would go into the principles of what does it mean to be a cloud company? That's what drives the advantages of being a cloud company, right? So we drill into the economics, we'll drill into the technology architecture of what it is. And then in the first wave, and when I say first wave of AI, I'm talking about in 2019, 2018, when we were doing sort of AI investments.

There were a lot of companies claiming to be AI, but they were using basically advanced statistical modeling and so on. So it was not really machine learning per se. So nowadays, the lines are getting much more blurry, statistics, everything becomes a part of it. So we as investors, I think, unless you're investing in the stack, when I say stack, I'm talking infrastructure, middleware kind of layer, where understanding what they're doing is important. If especially you're in the application layer,

let's take Tech Arena as an example, right? You just use your name to dot AI. So I'm not going to care much about the AI. I'm going to care about, Alison, what is Tech Arena's value prop? Maybe how is AI enabling you to do your job better, maybe help you drive better margins or expand your customer base, right?

So that's the kind of feeling that we do right now. If it is a hardcore middleware tech infrastructure company, then we go into the tech piece and that's where we drill into more. But if it's more of an application layer, then we look at, okay, what is the real pain point you're solving? And then try to understand how the technology there is more around how fast are you going to be able to adopt to changes, any future changes? Is your stack ready? If you claim you're AI, but you're not really AI, we don't ding you because you said you're AI, but we may ding you to say you're not going to be able to catch up to other companies that are truly leveraging AI to get to that scale

because they will get there faster. Their tech stack allows them to get there faster. So suddenly every company is a generative AI company. And then you look at them and you go, okay, you were founded in 2015 and Gen AI was still in the labs back then. So now if you pivoted to Gen AI, how has your technology architecture changed? Are you just a wrapper on top of Open AI's chat GPT, or have you done something fundamentally different to adapt your technology architecture to? Those are the kinds of things that we try to dig your technology architecture to. Those are the kind

of things that we try to dig into a little more. And I just want for full transparency, this podcast was not an invitation for Hitachi Ventures to be seeking venture funding for Tech Arena, but you will receive my proposal next week, I'm sure. But all joking aside, we're at the AI Hardware Summit, which is one of my favorite events of the year. I think it's so cool to see the innovation from the hardware industry when it comes to the new challenges for AI platforms. And silicon innovation, in particular for AI, is such a focus. Why do you think that this is a focus in the industry? And what have you seen at the conference that has opened your eyes? Many things, basically, from a silicon standpoint, right?

In fact, there are a couple of blog posts that one was posted by Sequoia that I would urge people to take a look at. It's called AI $600 Billion. And then there's another one that's from Altimeter, which talks about where the spend is going. They basically compared the spend versus revenue in cloud versus what's happening in the AI space. And I thought those were very enlightening, interesting articles, which show that there's a lot of spend and investment going into the semiconductor space. Companies like Nvidia, that's why you see them reaping the results and their value in the

market cap and so on. But as much as I have the opinion that a lot of that is needed because AI is going to become the norm. And think of what we have as dirt roads right now. And for us to move forward in the AI journey, infrastructure plays a very critical role and sort of hardware investments play a very important role to make that journey to this AI world more feasible. When 4T launched their cars, there was no tar roads, you know, with dirt roads, but the performance of the vehicle improves based on sort of what's underneath. And I think of the same, the only big difference is those roads that were made in the Great Depression still hold,

but infrastructure innovation happens at a much faster pace these days, right? Every year, NVIDIA, AMD, all of these guys are announcing newer chips with better performance and so on. But I think the race is still on only because they're becoming more and more power hungry. The algorithms are... So from an efficiency standpoint, energy consumption standpoint,

it's still pretty intense. So there is still a lot of innovation to be had because our brain doesn't use that much energy to do some of the tasks that a chat GPT. Maybe we are not as fast or we may not be as fast in certain things, but we don't consume as much power and energy as well. For this to be much more predominant, I think I'm interested in looking at, are there novel architectures from an infrastructure standpoint that can reduce power consumption? Are there better software design or things like small language models, which are more fine-tuned, are more sufficient?

Not everybody's trying to solve AGI. So if you're just trying to solve for a specific business problem, you don't need something that powerful. Something more simpler is fine. So trying to understand, is there a newer version or what comes next after transformer models? Is there a different sort of paradigm

in terms of approaching these? Those are things that sort of are intriguing for me. And that's why I'm here. I want to see what I can learn from all the speakers in terms of what are they looking at. The startups in the show tell you what's possible, where are they thinking, what problems they're solving. And then the corporates give you a sense of what are their hurdles that they're seeing in terms of deploying AI in scale. I think you get this nice yin and yang of startups pulling or stretching the boundaries of innovation and moving the goalpost. And corporates talking about, great, I want to get there, but I have these foundational issues that I need to fix before I go there. I've heard some companies say, AI is great, but my data is a mess.

It's in 10 different databases and they're not in a format where I can serve them up for training, let alone get any of it, right? So it's interesting to see that push and pull on both sides. I mean, I think that what some of the storage innovators are doing with unified namespaces

for that is really interesting. And I think that we underestimate sometimes what exactly it takes for enterprises to take on some of these challenges with integrating AI into applications. And I think that's where I wanted to go with my next question is you're looking at the startup arena, you see all of the innovation that's happening, really playing to what are larger players going to need from an acquisition perspective, as well as are they going to deliver a home run and deliver the next breakthrough technology that scales their company to be a monster? But how does that interact with the enterprise trust and the challenges with hallucination and some of the things that you discussed earlier? Do you think the industry is well aligned with where the enterprise really

needs solutions today. I think they're well aligned. I think this push and pull is important because if that didn't happen, we won't move forward. And at some point you see this mass conversation, the hype. The hype serves a purpose. When there's hype, then everyone has FOMO, they want to jump on that bandwagon. And so I feel like it forces them to have those conversations. But if that hype didn't exist, everyone's comfortable with status quo. So as VCs, then I wouldn't have a job. Everyone will just go there. Oh, I just want something that works. I don't care if it's innovative or not. Right. So in that sense, it's good because it

creates an awareness. It starts at the boardroom or the C-level, and then it trickles down to, okay, here are the challenges. And then there is a concerted effort to addressing them. Or that those hurdles are being highlighted to startups. I like your solution, but this is where I'm having issues, right? You know, TrustWise is a company that's focused on how can you drive AI performance, but with AI safety, alignment, and sustainability as core foundational elements. We knew that those will be questions. Not that we have a magic ball, but when you see the writing on the wall, the question is still going to be timing.

Did we nail the timing? But those are conversations that are happening. For instance, at Hitachi, one of our executives recently put out a blog post on how Hitachi as a conglomerate is thinking about embracing AI. So it's not that we do or we don't want to embrace AI, but it's the question is around what are the foundational elements that we need to consider before embracing AI? And one of the things they talk about is responsible AI. Hitachi is a large industrial conglomerate and we run critical infrastructure like rail and Hitachi energy you know that's high power transformer grids and so on you cannot afford to say okay I'm gonna just

put something full out there because it's AI and have service go down it's just not on those kind of settings but if you're a consumer app that's targeting real recommendations maybe your AI recommendations can come with a warning that says stick with your doctor before you do something, right? So it's not as mission critical. Or if it's like a makeup app, really the use case drives the adoption as well, where what is the boundaries in a healthcare setting, in a financial set, in the fintech kind of areas, people are much more sensitive to AI, how it's being deployed, and they're much more thoughtful about how it's being deployed. Now, you talked about moving beyond gen AI models. You talked about delivering new levels of compute efficiency, which I think is on everybody's mind as we look at the rising scale of data center consumption of energy along with other things. What are the other things that you want to see next AI hardware summit in terms of the industry's focus on innovation?

Is there any other topic that comes to mind? So we've invested in an agent company as well recently. And we've invested in a company that does cognitive control and cognitive robotics. And the reason I'm highlighting these companies are because we think on one end, AI, while it cannot solve a lot of our problems, I think it can make a difference. So we think AI agents are going to play a very important role in the future. In fact, the founder basically said something to the effect of software as a service does not make any sense anymore. It's going to be service as software.

Anything that humans used to do before will now be done by software, which I thought was a very profound statement. And then a few days later, I saw Klarna announce that they're cutting off use of Salesforce. And then they're also going to be shutting down their use of Workday because they have deployed AI to do the job of what Salesforce and Workday was doing in their company. So it's a very interesting trend that we're monitoring. So if we then extend agents from a software, can agents also become a physical embodiment of AI from a robotic standpoint in an industrial use case?

There is a lot of labor shortage in some of these critical industries, right? Because some of them can be dangerous work if it's in the mining or if it's in oil and gas or areas like that. And when you think about those kind of areas, if we can leverage AI and embodied AI agent to actually do the work and minimize human casualties, it becomes quite fascinating. So those are all, you know, again, fodder for thought. And then from an infrastructure standpoint, you know, we look at what should be on our radar to drive some of this. Security is going to be critical.

We saw recently with the CrowdStrike, Microsoft, blue screen of death everywhere, how things came to a halt. It tells us more than ever, software security and vulnerability and general security is going to be important, especially if you're going to have AI systems talking to each other. How do we know what's happening? How do we know when it's right, wrong, or malicious? Those are all things we think will be interesting to watch out for. Awesome. Ji, it's been a pleasure talking to you, as it always has been. And I'm so excited to see the progress that you and the Hitachi Ventures team are doing in taking a real leadership role in this space.

Where can folks connect with you to continue the dialogue that we started today? So check us out at hitachi-ventures.com. From a website standpoint, we are also very active on LinkedIn. Again, Hitachi Ventures' LinkedIn page. And there we have started putting out our thought pieces, blog posts on what we are seeing in the agent landscape, for instance. Or we think the data pipeline is evolving and changing thanks to generative AI. What does the next data stack mean, right? So we are putting a lot of our deep dives.

We are now starting to put them publicly as a blog post. So please check it out and reach out to us on LinkedIn or on our website, either works. You're the best. Thank you so much for being here today. It was a real pleasure. Likewise. Thanks for having me. Thanks for joining the Tech Arena. Subscribe and engage at our website, thetecharena.net. all content is copyright by the tech arena