Shaping the Future of AI Investments with Debarghya (Deedy) Das

In this Podcast episode, "Shaping the Future of AI Investments," join us for an insightful discussion with Debarghya (Deedy) Das from Menlo Ventures. We'll explore transformative technologies and sectors, including Generative AI, Enterprise SaaS, Cloud Infrastructure, and AI/ML investing.

Discover the opportunities, risks, and ethical considerations of large language models and generative AI systems that can create text, code, and images. Learn about the trends and innovations powering the modern enterprise tech stack, and what's next for SaaS, data infrastructure, and development tools.

Dive into the world of VC investment, with an inside look at theses, diligence, and working with AI/ML startups from idea to scale. Hear stories from the trenches about building stellar product and engineering teams, finding product-market fit, and scaling startups sustainably. Whether you're a founder, investor, developer, or just passionate about emerging tech, this is a discussion you can't miss!

🎙 AI50 Podcast

📹 Want to receive our videos faster? SUBSCRIBE to our channel!

👉 Visit our AI50 website

👉 Schedule a demo

📰 Receive our weekly newsletter

👉 Follow Hanh Brown on LinkedIn

🏛 Follow AI50 Business Page

Find Deedy on LinkedIn: https://www.linkedin.com/in/debarghyadas/

Transcript

Hanh: 00:00:00
Hello. Welcome to AI50, where your data meets innovation. At AI50, we skillfully deploy cutting edge AI technology. To create a large language model apps tailored specifically for the needs and preferences of the, uh, 50 plus demographic. So our mission is to harness the power of data and innovation to revolutionize the way this demographic interacts with technology. So today's topic is shaping the future of AI investments, and we have an

Hanh: 00:00:38
exceptional guest joining us today. Deedy Das, who is the principal at Menlo Ventures and a founding member of Gleam. His journey is nothing short of remarkable. As a leader in AI investments. DD has been at the forefront of shaping the future of AI and enterprise technology. So his deep technical expertise and strategic vision have undoubtedly left an indelible mark in the industry. So in a world where AI is rapidly transforming every aspect of our lives. Deedy's insights are so invaluable.

Hanh: 00:01:19
He's witnessed first hand the power of AI in revolutionizing knowledge management through his work at Glean. And now, as an investor, he's helping to fuel the next generation of AI powered startups. So Deedy's story is one of passion. Perseverance and unrelenting pursuit of innovation. It's a testament to the incredible potential of AI and the visionary minds like Deedy, who are driving its progress. So as we discuss and go into Deedy's journey and explore the frontiers of AI investments, We know that you, our

Hanh: 00:01:56
AI audience, will be captivated by his unique perspective and a wealth of knowledge he brings to the table. So get ready to be inspired, challenged, and enlightened by one of the sharpest minds in the industry. So, Deedy, welcome to the show.

Deedy: 00:02:18
Thank you for that lovely introduction. It's a pleasure to be here.

Hanh: 00:02:22
Yeah. Well, thank you so much. Um, yeah. So to start things off, could you share with us something about yourself, uh, that many people might not know, or, um, it could be a unique experience, a hidden talent or an interesting fact that help us understand what has shaped your journey and perspective.

Deedy: 00:02:43
I think there's a couple. I think the one I. I think people ask about most often is because I grew up in multiple countries, I spent You know, age three to 10, I was in America. I was in the Bay area cause my family moved here. And then from 10 to 18, I grew up in India. I often get asked, uh, how that experience was like and how that's shaped me. But one of the funniest quirks of that experience is there's a A bunch of people who do this as

Deedy: 00:03:13
well, I tend to code switch accents. So I have two accents. And then when I'm talking to people in America or Americans, I would speak in this accent, but when I'm speaking to someone from India, I would speak in my Indian accent, which is a nice quirk of having literally grown up in both places.

Hanh: 00:03:32
I can relate. I have an accent too, , except I speak Eng English with an accent for getting my S's and so forth. Well, that's exciting and, and thank you. And I'm excited to learn more about your journey. Now, your career path has taken you from engineering to product management and, and now to venture capital. So can you share with us what experiences or. Individuals have had the most significant impact on shaping your journey, let's say during your

Hanh: 00:03:58
time at Glean and Menlo Ventures.

Deedy: 00:04:05
Yeah, of course. I think I wanted to actually start with somebody who shaped my journey at Google that also, and then I can talk about one person at Glean and Menlo as well, even though it's only been several months for me at Menlo. The person at Google that was most instrumental in shaping my journey was maybe my first manager. His name was Eyal. He's an Israeli guy. He rose very quickly in the organization. He was extremely talented. Um, I still remember some anecdotes that

Deedy: 00:04:34
he used to say, which is I asked to ask him, Hey, how do you, how do you work so hard and how, how are you so effective? Um, and he would say, Man, Google is the easiest job I've ever had. You should have seen how hard my job was in the military when we were actually trying to save lives. Uh, I did that for six years. Search, that's kind of a low stakes problem. And with that anecdote aside, what he used to really say that really inspired me is at that time, I had a lot of doubts about whether.

Deedy: 00:05:08
I should be doing engineering long term, like what was this career going to look like for me, you know, 10, 20, 30 years from now? And I asked him, like, hey, you know, this product management thing is, is, is really hot. A lot of people seem to be doing it. And from a personality perspective, I think I'd be pretty good at it. Like, I enjoy talking to people. I, I, I like thinking about product and doing engineering. What should I do? And his answer was really simple.

Deedy: 00:05:35
He's like, most PMs can't do engineering. If you're an engineer, engineer who can do product, stay an engineer. Don't lose that skill. That's a hard skill. You never want to let that go away. And that's some advice that really stuck with me because I was seriously considering switching at that point. And I decided not to, which, you know, served me really well because when, and then I'll tie this together. When I joined Glean, there's no way I could have got a job at a, at a company with just five people at

Deedy: 00:06:00
the time when I was interviewing. Um, if I was in product, the only hiring need was engineer. So that opportunity wouldn't have even worked. So when I joined Glean, then brings me to the next person who was really, um, a big shaper of the way I think and, and, and my career, and that was the CEO and the founder, Arvind, Arvind is, you know, obviously a tenure Valley veteran, uh, started, uh, He was one of the first distinguished engineers at Google, where he reported to Larry, where his title was to lead speed at Google.

Deedy: 00:06:34
Very, very impressive guy. And then he started Glean. But if you speak to Arvind, he was just unlike any buddy of power that I would think. He's such a simple man. And all he would say is, Deedy, I don't do this for money. I do this because I love it. And, and I'd ask him questions like, Hey, what do you do on weekends, Arvind? Like, how do you relax? And And he would say, this is relaxing. I build product. That's what I like to do.

Deedy: 00:07:03
And I just love that passion. He wouldn't spend on much. He was a very, he lived within, very within his means. He just didn't care about the glitz and the glamour. He really cared about the job and was a very first principles thinker about execution. Anyway, lastly, I'll say at Menlo, even though it's been a very short period of time, One of the people who really inspires me at Menlo is somebody I work very closely with. He's, his name is Tim Tully.

Deedy: 00:07:30
He was used to be the ex CTO of a company called Splunk, which I think is a 25 billion public company right now. And what I love about Tim is, you know, Tim is about You know, pretty late in his career. He's he's in his late forties. I imagine. And for somebody of that age, I know I just grew up thinking that people of that age are a certain way from like that, but Tim just has this, this glitz in his eyes. He just loves technology. And I, I, I see him ask him, what do you do for fun?

Deedy: 00:08:02
And he's like, Man, this new thing came out and I just started coding on it. And it just has this like childlike curiosity, which I emulate, you know, I always say when I'm older, if I lose the childlike curiosity, it's game over. Um, that's sort of, I know they're really inspirational.

Hanh: 00:08:19
That is awesome. That it's just wonderful to learn how those crucial moments and mentors have influenced your direction. And. Inform your view in the industry. And now I like to know what your views are. People are approaching 60 because that's where I'm at.

Deedy: 00:08:36
You don't, you don't look like you're approaching 40 even.

Hanh: 00:08:40
Cause when you said, you know, he's in his forties late in his career. I'm like, wow, to me, 40 is not. Nearly late, almost 60. And I feel like, you know, thriving for learning. I mean, nowadays with the advancement of AI, you gotta, you gotta keep learning. You have to have an open mind, a life learning attitude and keeping up with the pace. It's challenging because every day is new and well, an example of. This, uh, Monday with the GPT 4. 0 release and how this transform

Hanh: 00:09:11
multimodal and so forth. And to be honest, I'm still kind of, I don't know, in candy land, just hearing all these advancement is so excited for it. So yeah. Okay. Well, as a leader, we often face challenging decisions that can have far reaching consequences. Can you share a story about a difficult choice you had to make while leading? The development of Glean or investing in AI, and what lessons you took away from that experience?

Deedy: 00:09:44
I think the most, maybe this is a cop out answer, but it's a recurring series of challenging decisions, which all are pretty similar in my opinion. What I've observed at Glean is What's happened in at least some part of the tech industry is if you look at these big tech companies, like, I mean, especially Google and Facebook, because I've worked there, I don't know as much about Microsoft and Amazon as deeply, but Google and Facebook for a very long period of time became a money printing machine. And because of that, a lot of the talent that grew their careers have

Deedy: 00:10:22
become quite senior have become. Quote unquote experts in their field. In my opinion, many of them haven't actually driven anything. They have sort of ridden the wave, which would have gone up anyway and did incremental work that isn't really that challenging. And in that culture, what was promoted was really an attitude of just being nice. And if you were to succeed at Google and Facebook, I think like the biggest way to succeed is just making everyone happy around you. Don't rock the boat.

Deedy: 00:10:56
This is what they would say a lot. I Google it. Don't rock the boat. Everything's going well. Metrics will move 1 to 2 percent every year. That's all you need to do. At a startup, that just doesn't work. No, different mindset. It's a very different mindset. What I realized is, you know, while we were building Glean Assistant or Glean in general, is you just need this idea of radical candor when you're making decisions and you

Deedy: 00:11:23
have to remove ego from the process. And that's not something that, you know, even initially I found that a little bit challenging, but there were a lot of hard decisions you had to make, even for people and colleagues that you work with that you really like, where you just have to be honest and sometimes say, you know what, like, I like you, but to make the correct decision, I cannot take the average of the five ideas of the colleagues I like. I need, we all need to figure out what is the best thing and what works. And you have to put your ego aside to make those decisions.

Deedy: 00:11:58
So there were multiple, multiple small decisions, you know, on a weekly basis where you really needed to dig deep. And it's really hard to do in practice where in a meeting you're sort of zoning out, like multiple people are talking, you're context switching, someone brings up something that they're doing, and you don't want, you don't want to go like, Hey, early morning, that idea, let's not do that. That's a bad idea. It's a waste of time. We're going to go the wrong direction. You have to be there always on always

Deedy: 00:12:23
being somewhat confrontational, honestly, in order to make that product succeed and to make the good decisions to make it succeed. And. A lot of that I also learned in, um, more, uh, it was sort of driven into me when I was working at Google Israel because I feel like Israelis just have that culture of, they always tell you they're very straightforward and candid. And, uh, without being personal about it. So I really learned how to do that there. And that sort of drove me to, those are the kinds of decisions you have to make

Deedy: 00:12:58
it in, in order to make the products.

Hanh: 00:13:01
And, you know, to add to that with a startup, the mindset is innovate, disrupt, incorporate a place like Google. It's conformance. Not to rock the boat. So they're completely in misalignment, but it's, it's wonderful to have folks like you that experienced both and a very understanding and adaptive. So that's great. And I appreciate hearing the stories, the journey, and it shows how dealing with complicated choices and, and, um, the role that you were in and the insights that you've gained.

Hanh: 00:13:35
Improves leadership and investment skills, which we'll get to next. So that's great. Now the AI and infrastructure sectors constantly evolving with new breakthroughs and trends emerging at rapid pace. So how do you stay on top of these developments and how do you leverage the knowledge in your role as principal at Menlo Ventures?

Deedy: 00:13:59
I think. Every investor has a different way of doing this. Most investors are, are not technical. And one of the reasons Menlo hired me is because I think they realized that actually having, especially in this era of technology, understanding the technology is pretty critical to making good investment. You know, we see all the time how. There's so many unicorn companies that sort of collapse and, you know, a lot of people are left wondering, okay, how did this company get

Deedy: 00:14:34
to a billion plus valuation if it can just collapse the next day? I thought a billion plus minute. Doing something. And when you dig into it, you realize there's multiple reasons. And that's a whole topic, but a big part of that reason is people are investing on credentials. They're not investing on technology because they oftentimes don't understand the technology. So if you look like you're a published person on Google scholar, you worked at deep mind, you went to Stanford

Deedy: 00:14:58
for your PhD, you're probably going to raise a lot of money, whether or not you're doing something meaningful. So anyway, long winded. Road to say, how do I stay on top of these things? It's pretty simple stuff. You know, every day I enjoy it. So I think that's really key to it. But every day I'll, I'll do at least an hour or two of just reading on hacker news, tech meme are some of the common ones I go to, I spend a lot of time on Twitter reading, but you need to create Twitter for

Deedy: 00:15:27
you to get real value out of it. Um, and then the most valuable thing for me, uh, the two other valuable things for me, I should say that one is you. I've noticed in startup world, there's such a dissonance today between what people are saying and what the actual thing does. And so actually playing with something often gives you, even running, even let's say GPT 4. 0, running five different queries on GPT 4. 0 will tell you way more than reading a hundred tweets about it.

Deedy: 00:15:57
Because all of those hundred tweets are going to say the same thing. They're going to take the announcement, they're going to be like, wow, this is the future, this is amazing. And you're not actually going to know what really changed. So that's one. And the second one is, I think, I think a lot of investors do this and we have to is you, you have to keep talking to other. Industry experts who have a different vantage point. Um, and I don't think of experts as necessarily, Hey, 20 years of

Deedy: 00:16:22
experience, senior director at now. I mean, you can talk to people with 25 years, we're 25 years old, and they will tell you so much more about a company close to the metal and just talking to them and understanding, Hey, how are you thinking about things? Why do you think this in probing also really helps, um, shape your, your vision. You need to hear multiple different views on things. For you to understand what is, what is actually closest to the truth.

Hanh: 00:16:46
Being close to the consumers, the ultimate users, right. And truly solving pain points and looking at it from their perspective, rather than just being product focused. I think I see that often. What's that?

Deedy: 00:17:00
I love that you mentioned that I have a story, the recent story that I can share with you about, about that. You know, like we're, we're in a little Silicon Valley bubble and everyone's always talking about the same things and AI and tech and all this stuff. It's great. I recently was with a friend of mine and I was telling him how, how, you know, crazy I was about all this stuff coming out. And it's sort of like, you know, half of the stuff you said, I don't even know what it is. Can you, can you show me?

Deedy: 00:17:25
And so we sat on a, on a couch and we watched a bunch of YouTube, a bunch of videos, And it was so interesting to me to see that I was drawn to the videos that were technically interesting to me. These are demo videos like OpenAI, Google, whatever. Um, UDO, just a bunch of AI tools. And the ones that he was drawn to were so different. Um, and that sort of made me realize. Like, yeah, consumers, as you said, users think very different from who are always, you know, in the middle and we're, we're, we're going to be

Deedy: 00:17:57
influenced by what people think about the technology more than your actual consumer. So it's, that's another thing that you have to be wary about.

Hanh: 00:18:06
I agree. I tell my team is that we're not in the business of producing products. We're in the business of solving, solving problems. For the demographic 50 plus, I don't care how old you are, but when you're working on this, you have to wear that hat 50 plus. That is, that is where you are, you know, and it's so, it's so important because, um, to understand and to come up with products that are usable in that demographic, it's not just product focus. You really have to understand

Hanh: 00:18:33
their behaviors, their, where they, they shine, where they decline. Right? There's cognition. There's visibility. There's ability to understand the instructions. There's so many, um, nuances. So it's very important to be consumer focused rather than product focused. And, you know, we can't explain enough, right? That's, that's a huge topic in itself.

Deedy: 00:18:58
Exactly.

Hanh: 00:18:59
Yeah. So building high performance team is very important for success of any organization, especially in a fast paced world of startups and venture capital. So what strategies have you found to be most effective in assembling and leading exceptional teams?

Deedy: 00:19:18
There's a couple that I want to talk about. One is, and this is something I, I learned from, from Glean and from, from Arvind, when you're building a team from scratch and you don't have a brand, uh, associated with it, there's an element of. You cannot apply the same hiring principles as a corporate would. Um, their needs are very different. When you're building a team, it, there's a snowball effect of reputation. So the first people that you hire, not only have to either be extremely exceptional at what they do, but

Deedy: 00:19:54
also they have to be almost, I would say an inspirational or, or a connector in their network. I'll give you one example. We could have hired one of 50, say new grads from Berkeley when we started hiring new grads, but Arvin had a very interesting insight. He's like, you always want to hire. The TAs from the, from the most popular classes, hire that TA next year, you get 20 more people who want to work at your company because they follow what he does because they're, they look up to him. And instead, if you hire the best person

Deedy: 00:20:26
in the class, who's just not well known, great, you got one good person, but the next year you might not get, you know, 20 more people banging on your door. So that's, that's one. The second thing is, is a bit more, uh, you know, I would say strategic about how you pitch what you're trying to hire for. And. When, uh, when we were hiring at Glean, there's some strategies you just don't want. One thing is you never start with comp. When you start talking about how much you get paid, you're already,

Deedy: 00:20:58
you're hiring the wrong kind of person, or you're hiring, even the right person gets distracted by that and starts making bad decisions. Then they start paying me, oh, but they're not paying me, oh, but I should get paid more. That's not what you, the mindset you want. That conversation comes after they develop conviction in what you're trying to do. So the first order of business is you need them to buy into what the vision and of what you're trying to build is. And you need to be, second thing is you need to be very honest with them.

Deedy: 00:21:25
At Glean I was always very honest about, Hey. You want to work, uh, you want to have a good work life balance at this job, you're just probably not going to get it. I don't know how to say it, you can make it work, I'm not, not everyone at Glean works like 80 hours a week or anything. There's people who work 40 50 hours a week, they're very intense for that 40 hours that they work. But, let's just say it's not going to be an easy job. And if you want an easy job, You know, frankly, I'll be up front,

Deedy: 00:21:52
like, you should go somewhere else. And that in itself is, it's funny because not only does it select for people who actually care about the other stuff that you're looking for, but even the people who might want work life balance are sort of, they feel threatened and they're like, no, I can do it. I feel like that's a challenge that I can take. And so they, they lean in even more. They actually appreciate that, that candor instead of you trying to dance around it and say, Hey, you're going to get work life balance and

Deedy: 00:22:19
you're going to build awesome stuff. And you're going to. You know, you're not going to get all of those. So those are some tactics that I found that are useful.

Hanh: 00:22:27
Yeah. And you know, I think, I think to start out with, I think it's wonderful if you can have that honest conversation with full conviction of your values and your goals so that they are aligned. It's easier to know a front when they're not aligned as opposed to finding out six months into it. Right. I feel that that's some fundamental. In hiring and keeping a team, and I want to add back to the TA example that you cited. That's exactly what I told my son.

Hanh: 00:22:58
He's computer engineering at Michigan, who graduated now, is working at Microsoft. When we had that COVID situation, everybody is going to class behind a screen. He was wondering, what can I do, right? I mean, we don't go anywhere, we don't do anything. So he applied to be a TA. And he loved it. You know, he never realized that the value and the fun. Well, it was during COVID, right? You have to do the best that you, you can behind a screen, but he loved it.

Hanh: 00:23:25
He learned so much. And then the point that you made about the networking and so forth. So he's over at Microsoft now and the people that he helped, Proctor classes, Want to do the same thing. So now the company that he's with open up for multiple qualified students. So it's true. It's so true. So great advice.

Deedy: 00:23:47
Exactly. I work like a charm.

Hanh: 00:23:49
Yeah. So, you know, in the world of technology and investing, pivots and adaptation is so important to stay ahead. So can you discuss a time when you had to shift your approach to a project or investment? Like, you mentioned unstructured, pinecone, or anthropic. And what insights you gain from that experience?

Deedy: 00:24:13
Let me talk about the Glean example first because I have more knowledge of that. In late 2022, we were at a pretty critical juncture at Glean where large language models were clearly powerful and extremely interesting and cool. And they went very far to solve the mission sort of that Glean was going after, which Um, to organize it's Google's mission, but for work. So to organize all your information at work and give it to you when you want it. At that time, because a lot of our leadership was ex Google search people,

Deedy: 00:24:51
when you present them with this alternate view of the world, Their first knee jerk reaction is, as expected, we've been working on search for 20 years, and if you give me a penny for every time I heard this new technology will disrupt search, I would have a lot of money. And so when they came to this idea of LLMs, the first reaction was, yeah, I don't know about it. It's like, it's cool. We're doing search. This is just not our business. And I think that was a moment where, It was interesting, it's an interesting

Deedy: 00:25:25
time was interesting for me and some other people at the company because not only do you have to dig deep within yourself and find your own conviction in order to fight this battle, because you're like, Oh, these guys are probably right, like they've been working at Google for they built search that they're, they're not dumb. And so you first have to dig deep for your own conviction. And then you have to strategically make sure you can convince. And it was a challenge. It took about a month, maybe more of

Deedy: 00:25:51
many people and many examples and demos and building stuff and demonstrating what it could do before it sort of was greenlit by the company as, Hey, we should do this next product. And looking back on it now, now the product is, you know, a significant part of, of Glean's revenue. And it was one of the best decisions that they could have made, but it wouldn't have happened if you're not constantly willing to fight the fight and willing to question things at the right moment. Menlo is similar. I've only, again, I've only

Deedy: 00:26:25
been a part of some of these and they say it all the time. Menlo's existed for 48 years. It's one of the oldest venture firms in the valley. It's running on its 16th fund right now. Most venture funds, I think the data says that most venture, 65 percent of venture funds never return capital. So if you've done it 16 times, you've clearly done something right. Um, when you do it for that long in venture, there's. And these are, this is advice coming from people who've done this for 20 plus years.

Deedy: 00:26:58
And I'm just telling you what they tell me is you can always have to reinvent yourself. If you're not reinventing yourself, you're not ahead of the curve. If you're not thinking deeply about, hey, what am I doing differently that the industry isn't, then you're just going to follow whatever everybody else is doing. You have no actual alpha on the industry. So we constantly think about our approach, whether it's the kind of checks we write, the volume of checks we write, the, the, the size of checks, the kind of companies we invest in,

Deedy: 00:27:25
um, what stage we want to invest in. All of that stuff you constantly have to think about and say, can we do better here? Can we do something more interesting here? And the Anthropic example is such a good example of that, which is such a non-traditional deal for. Um, Menlo to do, and it was, you know, an SPV structure into Anthropic that did a huge round that, that, that, you know, is keeping it running now. So I think that's a, an example of innovation is, is everywhere.

Hanh: 00:27:58
That's very exciting and, you know, balancing risks and practical consideration is an ongoing challenge. Now, how do you go, how do you approach balance when investing in or developing cutting edge technologies?

Deedy: 00:28:15
It's different for when you're operating and versus when you're investing. Um, when you're operating, I think risk mitigation is, you know, You know, it's, it's a little bit easier, I think, because it's resource constrained a half the time. Um, you, when you're operating a business or a startup, especially at small scale, you want to remain focused. You also want to try something new. The philosophy we always had was sort of 70, 20, 10, just 70 percent is the core work. 20 percent is some new work.

Deedy: 00:28:54
And 10 percent is completely experimental stuff. I think that framework broadly generally works and that already balances out the risks. It balances out the risks because you're only focusing on one thing at a time. And that is what you need to do. So that's, that's how it works when you're operating. You can't really, and then as you do that focus, the way you balance out different kinds of risk is more strategic and you'd have to get into the details. Like how do you.

Deedy: 00:29:19
Solve for market risk. How do you solve for tech risk? How do you solve for all these other types of risk in the business? And you need to, as a startup, just slowly eliminate all of this risk so you can make something successful as an investor. It's a little bit different. Um, managing risk involves number one, understanding the reward. Which means doing diligence and really understanding a space as best as you can. It's going to be imperfect talking to people, talking to experts, really diving deep into the

Deedy: 00:29:45
numbers and, um, the technology. And so you need to understand the risk, the reward to understand how the risk plays into it. But unfortunately, the hardest part of investing is you don't control. The risk is completely dependent on how much you pay. Like if you pay less than there's, there's no risk. So, but the price is completely dictated by the market. So if the market is frothy and people are overpaying for stuff. It's a tough place to be

Deedy: 00:30:13
because what are your options? You can sit out of every single deal because people are overpaying, but then you're not doing your job or you can participate and overpay, but that price is just not dictated by you. So all you can do is say, maybe I have an asymmetric or a different view of the reward. So I think the overpaid price is okay, but that's about it. Um, I, I think that's what makes it harder. You have much less control as an investor than you do as a startup, um,

Deedy: 00:30:44
founder, because as a startup founder, you can kind of move that, stir that ship in a lot of different ways. Um, as an investor. You're kind of at the behest of the market.

Hanh: 00:30:55
Great insight. So now, what do you see as the most significant opportunities and challenges facing AI over the next, let's say, 5, 10 years? And how do you incorporate these considerations into your investment decisions?

Deedy: 00:31:10
There's several. I'll talk, I'll start A lot of the generative AI products that we see today are quite impressive to demo, but they are not reliable. So we know this, they're non deterministic and non reliable, and that's a problem that if we can solve, I don't know how, but if we can solve it, it makes it way more effective. Number two is, I think, In this sort of hype cycle or a frothy cycle, there is a lot of people building with AI tools that are not focusing on product from first principles.

Deedy: 00:31:46
As a result, a lot of products that we see have really poor retention and they might be nice and hot right now, but Who knows whether people will continue to use those products in six months, unless you fundamentally think about first principles. How do I build a good product? And that does nothing to do with AI. That has basic product principles. That's number two. Number three, I think is we are at a point where so much of the next generation of power of what AI can do is bound by compute.

Deedy: 00:32:22
Um, when you're bound by compute, There's only only a couple of different options you have. Either you can make chips better and more efficient, more throughput, faster. Um, and that takes time, or you can figure out ways to make the, uh, underlying algorithms more efficient. Now, historically, if you look at it, typically it's been the latter. Sometimes there's, uh, innovations and algorithms, but almost always like the innovation is driven subtly through an increase in computation. I'll, I'll forgo as an

Deedy: 00:32:52
example, when they thought go. Yeah, the technique was really cool, but you couldn't have even done that 20 years ago. You just would not have the computational power to do. So that's, that's the third. I think about that a lot. And fourth is, um, less about the macro industry, more about like specific investing, which is a lot of people have great ideas, but they're, they don't think that, or they don't think that much about how their right to win in a market. And I always like to say.

Deedy: 00:33:27
I love ideas that are unsexy because if you're trying to work on something that's a really sexy idea, you have to also convince me that you, a small startup with three people are working on something that Google and Facebook and OpenAI and all these companies are not working on. So that's a challenging question and it's not like you're building a product, you're building a technology and these guys have talent, they have money, they have GPUs. What are you going to build that they can't build? What is your core insight? That they're not going to do already.

Deedy: 00:33:57
So those are a couple of things that I think about when I'm thinking about my investing thesis. And a lot of the companies I end up really liking are companies that look, they have a pretty strong play in AI as a startup, but their underlying technology is actually way more. And if by chance AI stops being important or stops being as hot, or it's captured by other players, I think the risk is generally mitigated because these guys can do so many other things. So a lot of focus on that sort of strategic thinking, but ultimately I think

Deedy: 00:34:31
the biggest advice is team as an investor. The most, most important thing at early stage is team. There are just some people who can, as they say, like run through walls, hire great people. And for them, it's almost. Unimportant what they're working on, they usually succeed no matter, no matter what.

Hanh: 00:34:49
Let me ask you about the team dynamic and hiring people, you know, with this AI ecosystem and agents. Okay. How many people, I guess the question is what impact you think that will have in the workforce? Let's say we used to think about five, 10 people, startups and so forth. How many people do you think we really need now? With the ability to create AI agents.

Deedy: 00:35:15
It's an interesting question. I think a lot of functions I generally think are getting automated and that doesn't mean people lose their jobs overnight. Uh, and sometimes it doesn't mean anyone loses their job. If the company can afford the people, then you're just getting more work out of the same people. But how does that change what we do? Let me just jump in. Be specific on certain things that I've seen, um, and TBD, how this all plays out, but for sales and outbounding,

Deedy: 00:35:45
there are tools that completely replace that of varying quality. Some of them do it really, really well, where they can just, they pick your ideal customer profile, the kind of target persona you want to reach out to, and then they will handle emailing them, responses, everything to when the meeting is on your calendar at whatever time you have. That is a huge time saver. There's a similar sort of tool for recruiting. You can do the same thing for recruiting candidates. Uh, get them, get them in the door.

Deedy: 00:36:17
No need to send emails anymore. There's a lot of work going on when it comes to doing engineering. So, tons of coding agents of various forms that can do interesting things. Like, there's some variants of coding agents which is like, just put this in your GitHub repository and we will automatically find issues and we'll go and fix them for you. And that's just work you wouldn't have even thought of. So that's like a different kind of, uh, extension. And then of course there's Copilot

Deedy: 00:36:44
tools where you're trying to develop and as you're developing it sort of helps you write code. Um, it helps you review code, helps you test code, various forms of that. Um, there are even voice outbounding tools or survey tools that help you. You know, do that better tons and tons of summarization tools, which help you take a lot of information and put it and make it really, really, um, digestible and easy to digest. Glean, obviously I'm biased, but I think that is a huge one because even at Glean, we would sort of get jaded when you work

Deedy: 00:37:18
at a company for a long time and you forget how, how cool the product you built is because you're always in it. And we'd have, we had a customer one time tell us. You just don't understand that most of my job as a senior person was being in these slack channels and answering the same questions. Day in and day out people ask me the same stuff. Big company, Hey, what do I do here? How do I fix this? How do I do that? And I would always be redirecting

Deedy: 00:37:43
them to a wiki and telling them, Hey, look at this answer. And I'd feel like a bad person for doing it. Because I wasn't just giving them the answer and then when Glean showed up and Glean Assistant showed up We would literally deploy these bots in a bunch of slack channels Say for HR or IT or even generic stuff and he's like my job is 80 percent lighter I just don't have to do that anymore. It goes it searches through the information and it tells them the answer Without me having to do

Deedy: 00:38:09
anything, there's completely autonomous. So I think there's a lot of power in what these things can bring. And I think it's incremental. So we're not going to see a change overnight. But yeah, I mean, it's already true that there are businesses of 10 people, 15 people who can make hundreds and millions of dollars and operate efficiently, which is not really possible like 25 years ago.

Hanh: 00:38:33
Yeah, that's true. So these AI agents augments your workflow. But it's like a GPS. It still needs a driver. And that's you.

Deedy: 00:38:44
Yes and no. I think there comes a time when it can, it can work autonomously. I don't think we're necessarily there yet for each task. I mean, take the outbounding tool too. Like, yeah, you have to drive it and say, Hey, these are the kind of people I'm interested in. Once you do that, then for the next month, two months, you don't really have to do anything. You just have to take the meeting. And so that's almost fully automated. So yeah, some driver, cause you

Deedy: 00:39:07
need to give it some direction, but yeah, a little sometimes.

Hanh: 00:39:12
That's great. You mentioned earlier about, I think it was topic one or point one that you were making about the products are not reliable or sustainable. What do you think that's due? Okay. Due to what?

Deedy: 00:39:26
Oh, well, that's just due to the inherent, um, way that large language models work. There's a lot of, uh, interesting research. Work on this, which is by design, it is trained on a large set of data and it is meant to be probabilistic in nature when it generates new text. And because of that probabilistic nature, it sort of imitates, um, I guess the core of the question is if it didn't have that probabilistic nature, it would also be less useful for many tasks. So, if you tuned it to hard determinism,

Deedy: 00:40:02
you would lose all the creativity, and it's very difficult to understand what kind of tasks are creativity focused tasks and what kind of tasks are. It's not a clean line. There's tasks where you answer a question and based on how it reads the information, there's a little bit of creativity, quote unquote, involved in understanding what the answer is. And determinism won't work there. But there are other tasks where you really want it to always work the same exact way every single time. And I think that's the sort of trade off

Deedy: 00:40:30
in large language models is hard for, just on a technical level, to get at.

Hanh: 00:40:37
Very true.

Deedy: 00:40:39
Induce it as well, but um, yeah, that's kind of where we're at.

Hanh: 00:40:43
Well, that's great. So, as AI continues to advance and become more integrated into our daily lives and businesses, an issue comes up is ethical consideration, which is so crucial. So what specific ethical challenges did you encounter at some point or anticipate facing in AI development and deployment? And how do you, let's say, contemplate in navigating these complexities as a tech leader?

Deedy: 00:41:13
They're tough. I mean, I'll give you a list of ethical considerations that are just really tough to wrestle with. One of them is. You're training on all this people's work. Maybe that's legal. Maybe that's not in some domains. It definitely is not in some domains it is, but you can't prove it. So that mean for how law works in this case, artists are very annoyed at this when it comes to music generation, uh, artists are annoyed when it comes to image generation, text, I think

Deedy: 00:41:40
less annoyed, but you know, for books and stuff, it's very clear that these large language models train on books. You can kind of trivially trivially figure that out. But. Again, you just can't pinpoint and say you definitely train on copyright material. So, that's an ethical consideration in that I don't know how to navigate, frankly. Here's another one. The second thing is, say this is a minutiae of things, but OpenAI released this amazing voice feature with 4.

Deedy: 00:42:13
Once that API releases, a lot of people are going to be using that voice feature to make phone calls to other people. And I think by law, if I'm not wrong, it is not. Legal to make a phone call with an AI without disclosing that you are actually an AI But we know that everyone's not gonna follow that it's not really gonna be cut down on and ethically That's just pretty wrong and you can't do it similar considerations on recording tools There's a whole suite of hardware stuff that just kind of always on recording Not all of them ask for consent of every single

Deedy: 00:42:47
user or every single person in the room. That's an ethical consideration.

Hanh: 00:42:52
You mean transcription? Transcription, yeah. That's true.

Deedy: 00:42:57
And the last one I think is the one, I think it's a borderline ethical, but it's also an ecosystem consideration. And let me put it this way, the reason I Sometimes care less about the ethical considerations because I feel helpless. I'm like, what am I going to do? This is just the way it is. I'm not in a position where I'm a regulator or I'm a law enforcer where I can make a difference here. Um, I'm investing in the space. Like that's just not something I can really focus on if it's something egregiously wrong.

Deedy: 00:43:24
I'm not, we're not going to invest in it, but I can't fix bad actors. So that's why I think about it less. The one I think about a lot. Is the ecosystem dynamic that the what happens to the ecosystem with these AI tools? And what I mean by that is the web ecosystem used because there was a feedback You would create a website people would go search they'd find your website. You'd get paid some ad ad revenue And you have incentive to create more websites, there's more people will create more websites and the ecosystem

Deedy: 00:43:53
is nice and flourishing and always has content and people can find content and Google can make money, everyone's happy. Now, with Gen AI, because it's trained on, you know, content up to a certain point in time and then just shown to you, the people who are making the content just don't get paid out. When the people don't get paid out, there's Dying incentive for anyone to contribute anything of value. We see this in journalism already, like, right? You know, in journalism, everyone just writes click baity articles because no

Deedy: 00:44:22
one will read the actual other stuff. So, what does that mean for all of the content in the next ten years? Like, who writes that? What did all of the information, all the news, all the stuff we want to know about the world, who's gonna write that? Why do they have any economic incentive to do that? That's a huge question. Um, and that's kind of a more existential question and people have answers that I think are just not, you know, Great. Like some people say, Oh, people will just, I don't know, tweet about it

Deedy: 00:44:49
or post about it, but no, they're not going to post about what's happening in, you know, a war in Mongolia, that's a journalist job. They have no incentive and no economic incentive to do that because no one's going to get paid them for their work. And if some AI is going to train on it and, and spit it out, that's a problem for the ecosystem. So. Those are the cuts. That's one I really, really worry about. Um, but yeah, there are, there are plenty of it.

Hanh: 00:45:14
Yeah. You know, great power comes great responsibilities, right? And, um, it's not always a clean cut, but I think, I guess just like anything, you know, you embrace it with caution.

Deedy: 00:45:27
Yup. I mean, what are the options you have?

Hanh: 00:45:30
Yeah. So even some of the most successful leaders face setbacks and failures, failures, and So, can you share an example of a significant challenge that you encounter and, and I guess how did you overcome that and what lessons did you take away from that experience?

Deedy: 00:45:49
Um, I would say one of the most significant challenges, uh, in, in my career was at the later stage of, of Google. I had a, um, a boss that was frankly just bad. Um, he was egotistical, power hungry, less concerned about solving the problems and, and driving, uh, the product. He was more concerned about control. So naturally this led to some tension and. I can't say I walked away from that with some amazing revelation of how I should deal with conflict. But what I can say is I, I realized

Deedy: 00:46:37
at that point, you know, I was going through some, some tougher times at that point, I was talking to a lot of friends and people, I realized the most liberating thing Is to not care. You have principles, you have values, you have what drives you. And I think what really drove it home for me was when I was talking to, you know, a mentor of mine and they said, what's the worst that can happen? He'll fire you. Are you going to be really that sad if you were fired for standing up for your values?

Deedy: 00:47:11
No, I'm like, no, I'm not going to be sad at all. In fact, I will be sad if I stay, stand up for my values. Cause then I'm like, what am I doing? I think what I'd say is, what I took away from that is, you should know when to fight the fight. You don't always, think, sometimes people disagree, you're, most of the time you're probably trying to make sure. Everything is nice and happy and people are under control, but when there's a time when your values are coming to question and the way you

Deedy: 00:47:39
operate come into question, yeah, you need to fight the fight because you will feel much worse if you don't. And then if you're not standing up for the things that you, so I think that's what I took away from from that.

Hanh: 00:47:52
Yeah. And you know, it's so true as you navigate in the later years. Whether it's professional or personal, we all develop that core of conduct. It's, it's that solid rock that over the years of experience that have formed and shape and where you are right now. And it's your core of conduct and that does not deviate, you know? So I, I agree with you. That's so important. So now I'm sure there have been many difficult decisions that perhaps That you made that are against popular opinion,

Hanh: 00:48:22
aside from your ex boss from Google. So can you share with us a situation where you face this challenge, whether it's Glean, Menlo, or any other physicians, and how would you navigate it, and what would you do to overcome that?

Deedy: 00:48:39
Multiple times. I mean, I think the hardest, hardest time is. I think some of it is personality dependent. Sometimes I'm generally a more of a debater persona. So I generally don't hesitate to express my opinion. If I disagree, I think there's an art to it. Like you don't, you don't disagree and then attack the person you disagree and try to arrive at a solution together. Can't say I'm perfect at that, but

Deedy: 00:49:00
I try to do that at most times. The hardest times I think are when you're, when you disagree with someone you respect immensely and you're like, this guy has every right to be right. Like he's done this longer than I have. He knows more than I do, but I just don't buy it. Like, I just don't understand what he's trying to say. And I think in those situations, what I, what I try to do is it's really, I guess, not that fascinating advice is you have to lead with respect. My view is.

Deedy: 00:49:32
If you know the person has way more power or you respect the person it's not even about power It's that person if you respect the person he has more control over this and more experience Then your first approach has to be look I may be wrong But i'm telling you this because this is what I feel at the end if you disagree with me Your, your opinion goes, and I always leave it, I'm like, look, your opinion goes at the end, I'm just telling you what I think and why, but your opinion goes, and here's why I think what I think, so let's talk about that, here's what

Deedy: 00:49:59
evidence I have, here's the data I have, here's as much, do the research, like do the work, nobody likes to be somebody. You know, a lot of these people are busy and they don't have the time for you, honestly, a lot of the time. So when you are going to disagree with, with, uh, with them on something, do the research, do the work, bring it up. And then, you know, it comes to a point where they actually respect you for that. And they're like, you know, I'm really glad. I get this all the time where a lot of people are like, You know what?

Deedy: 00:50:30
This is why we hired you, man. Like, you, you disagree, and I, I think you have a fair point. And so, that's, that's, I, I try to nurture that a lot. And there's a healthy balance. Again, you have to know when to disagree. You can't disagree every day. Then you just become a disagreeable person.

Hanh: 00:50:45
Yeah.

Deedy: 00:50:46
You know this is a battle I don't want to fight. I don't care about, this is the one I do care about. And then you say, the ones I care about, I'm going to express respect.

Hanh: 00:50:54
That's great. Communicate so that the two lines stay open, right? Because it's based on respect.

Deedy: 00:51:02
Yeah.

Hanh: 00:51:02
That's awesome.

Deedy: 00:51:03
And in addition, I'll say I also really believe strongly in bonding. Like the fundamental reason why people get into conflict is because they don't know each other. And so you need a little bit of that bonding. You need to know like, hey, where is this person coming from in life? To really understand and have empathy when you're dealing with them. And then you're like, okay, I get it. Look, we disagree, but I, I still, I still think you're awesome. Like, so.

Hanh: 00:51:27
You know, I agree with that, you know, and, and I think it's unfortunate sometimes, at least just me. I don't see it enough where people invest in the time to understand the story behind it, because everybody has a situation. Everybody has, whether it's a bad day, bad scenario, uh, whatever they come to That perhaps that you are in alignment. It could be in the actual scope, the topic, personality, or just the history and the luggage that they come with. And it takes, like you say, time to invest, to explore, to understand with empathy.

Hanh: 00:52:01
Maybe it's me. I don't see enough of it. Do you?

Deedy: 00:52:05
I don't. I don't. And I like this quote that one of my friends always tells me. It's like, Everyone has a struggle that you don't see. Exactly. And, and they're not going to tell you about it sometimes, but just like, just leading with knowing that everyone has that. Their stuff just makes a little bit nicer.

Hanh: 00:52:21
I agree. So now for those who are starting out in their tech industry, what advice would you offer based on your experiences as an engineer, manager, and venture capital principal?

Deedy: 00:52:33
I would say don't be too swayed by what's the hot thing right now. So learn it, know it, but. Know that, like, things come in waves and your incremental alpha is knowing what, knowing something that other people don't, not knowing something like hundred people do. Um, so focus on your curiosity first. Um, I would say learn how to love the craft and always stick to that guiding principle because work doesn't become work when you just love the craft. So those two things I advise. And then the third thing I would say is.

Deedy: 00:53:10
You know, those are easy things to say, but in reality, you know, everyone has to recruit and get a job and do all these things. Generally, I would say that don't be idealistic about how a career works. And I see a lot of this. I see a lot of people who say things like, Hey, You don't need to go to college to get a job and they're right you don't but I mean you also don't need to There's a lot of things you don't need to do to to get somewhere But they help and they help a lot and I think what's more important to me is I like to tell Especially like

Deedy: 00:53:46
my brother is in college right now Like I like to tell my brother and kids that that I advise I'm like, hey, man, like look It's great to be idealistic, it's great to be curious, it's great to lean into what you like, and you should always do that. But then always have your reality hat on, which is, this is just how it works. And don't let the, don't lie to yourself about how things work. Like, I'll give you an example, like, people lie to themselves all the time. Stanford, if you go to Stanford, you are going to get, have it easier to get a job and to start a company.

Deedy: 00:54:16
That's just how the world works. I think most people know that. But very, once in a while, they're like, Oh, actually, it doesn't matter. I can go to, you know, Cornell and I can do the same. You can do the same. Be honest. Stanford helps. And I think having that reality check and knowing this is just how things work. Right now, and this is how I like to operate, that in itself will go a really long way in, in your career.

Hanh: 00:54:42
Very true. Many years ago, when my kids were in college, they all graduated now. I always advise them, the branded school is always nice. Is it mandatory? Well, I think what matters more is that what you do with what you know, As opposed to a branded school and just bank on that, you know, and to me, knowledge, your, your thrive, your execution, your heart, your conviction, and a branded school is always nice. No, those are great advice. So thank you.

Hanh: 00:55:16
So now we're getting to the end, another question. So as a decision maker, our personal philosophies and values. So what core principles have you or have shaped your approach to your roles at startups and your principle at Menlo Ventures?

Deedy: 00:55:39
The values that really matter to me, I've thought about this a lot in my just general in my life is there's both personal and professional. I don't think they're different values. And there's two. Um, there's only two. Um, the other ones are like common. Everyone has those, but the two that matter most to me are humility, which I think no matter what, where you are, what you do, what you achieve. You need to be humble like you haven't done it all because every next battle is a separate battle

Deedy: 00:56:07
And you don't have any right to win that one unless you're humble. The second one is curiosity Um, I just think that you know for me that's like why I do Whatever I do. I don't seek any title or I don't really care about the title Money and status and I don't have like very rich, like expensive hobbies. What I really care is like, I'm a curious person. I like to read and learn about things and, um, as, as much as I can make an impact in the, in the way that I, that I can. So I think humility and curiosity will

Deedy: 00:56:40
always be sort of my core driving values and everything else comes from that. So yeah, you have like the honesty and you have like be, be disciplined and be hardworking and put in the effort. All of these things come from that, but I think those are, that's maintained that sort of throughout.

Hanh: 00:56:59
That's wonderful. So we're at the end now and I appreciate your time so much. Would you like to add anything else before we wrap up?

Deedy: 00:57:09
No, I, I think the only thing I, I I'd like to say is. When it comes to, uh, like the, the, the advice question got me thinking is when it comes to approaching a new thing, an industry or whatever, trust your gut, you know, trust your gut, what you, what do you think that's different from other people and follow that. And I can give you so many examples. Of people who have just trusted their gut long enough And those are the ones who are the most tremendously successful. Everyone else told them don't do that. That's a stupid decision Why

Deedy: 00:57:46
are you starting this company? Why are you spending all your time in this dumb project that doesn't work? Just trust your gut like that's fine And I don't believe in this. Oh all your friends if your friends are haters, then they're not your friends No, sometimes your friends will be like that. Sometimes i'm like that people are like that. Um, and that's okay, right? And that's okay It's your job to be like I hear you but not I'm going to do it anyway. And that's, I think how most people

Deedy: 00:58:11
should, people should have the confidence to just trust their gut on things.

Hanh: 00:58:17
Well, I want to thank Deedy for joining us today and sharing his incredible insights and experiences from the world of AI and infrastructure from his early days as an engineer to current, his current role as principal at Menlo Ventures. His journey is a testament to a power of curiosity, adaptability, and strategic thinking. In his stories he shared today and lessons on building high performing teams, navigating difficult situations, and staying ahead of industry, industry trends are so invaluable for anyone looking

Hanh: 00:58:50
to make their mark in the tech world. And as we discussed, it's clear throughout our conversation that his passion for innovation, combined with his deep technical expertise and leadership skills. It's made him a driving force in shaping the future of enterprise technology. So we're excited. Can't wait to see what groundbreaking startups and technologies that will help bring to life next. And to the audience, thank you so much for tuning in. And don't forget to follow AI50 LinkedIn page.

Hanh: 00:59:26
Subscribe to our podcast and YouTube channel for more engaging discussions. About tech venture leaders, aging products for the 50 plus demographic. Thank you for your attention.

Deedy: 00:59:40
Thanks, Hanh.

Hanh: 00:59:41
Thank you. Yeah.