Decoding the future of A.I. with tech visionary Jeff Huber

AI is the rage, and the question everyone wants answered is how artificial intelligence is going to change their lives for the better? Or is it a technology so powerful that in the wrong hands it could end life as we know it? If we're going to find out, we need to talk to those who were there at the beginning and lay the groundwork for what could be the biggest technology advancement since the dawn of electricity. Let's get started. Welcome to the Money Runner. I'm David Nelson. I'm very excited about today's interview. So much so I flew all the way out to San Francisco and then on to Palo Alto to sit down in person with today's guest. Giving a proper introduction would take some time, but let me just hit on some of the high points. If you have used or collaborated on the Internet, Jeff Huber has touched your life. He has worked on and led some of the most influential companies on the planet. He was the CEO and founder of Grail, a firm dedicated to early cancer detection and spun out from Illumina, where he served on its board. Prior to that, he was senior vice president at Google. He co founded the life sciences effort at Google X, and he and his team developed some of the more profitable projects at Google, including Google Ads, Google apps, and Maps. He was recognized as one of the hundred most intriguing entrepreneurs in 2017. Jeff holds a bachelor's degree in computer engineering from the University of Illinois, a master's degree in business from Harvard. He's a visiting scholar from Stanford University's Department of bioengineering, and he sits on the board of too many companies to mention. And finally, today, Jeff is a founding member of Triatomic, a leading venture capital firm focused on engineered biology, new energy, next generation computing, and engineered materials. Jeff, I know I left a lot out here, but welcome to the Money Runner. Thanks so much for being with us today.

Thanks so much, David. I'm excited to be here.

Jeff, I'm going to try something a little different here in this interview. Going to kind of go stream of consciousness and maybe a little out of order, but right at the top, I want to touch on what some of my listeners want to know about. You are on the cutting edge of technologies and services that in some ways are destined to almost change life as we know it. High on that list is artificial intelligence or AI. What are the dangers for me, my job, my children and grandchildren? So I put to you, what is the biggest risk we face with a technology that could disrupt just about every platform we know of?

So there is a lot of excitement around AI. And as you highlight, there is some fear along with that. In the scheme of things, I would say I'm an optimist. I do think that this is a time of unprecedented change. With change comes opportunity. But with change can come fear as well. On the positive side, I personally, and then our firm that you mentioned, Triatoma Capital, are big believers in the positive potential of AI to essentially give all of us superpowers where we now have access to intelligence and reasoning and capabilities that just weren't possible previously. I think the best way to consider or to think about the fear side of it is it is going to introduce substantial change. And my encouragement would be to lean into the change. If you look back on periods of historic change and evolution, the people that were part of the change tended to do better than the ones that were riding along or were ultimately impacted by it. So my encouragement would be for everyone, the listeners here, to lean in, to learn to be active users, to think about what are the ways that AI and the products that are available today. Things like OpenAI's ChatGPT or Google's Bard can impact how they think about the world, how they learn, how they do their job so that they can be part of the change.

Is government up to the task? Because even today the Biden administration announced a new executive order focused on AI. Is there a danger that they're going to overstep here and snuff this out before it's too late or the danger's real and something needs to be watched here.

So I think the key role that the government can play is one around providing the safety net as part of the change that is happening. So thinking about how can there be incentives for businesses to help employees become trained to kind of stay on the train as things are moving, I haven't today's been a super hectic day, so I haven't had a chance to actually completely digest the executive order and all of the implications of what they got right or what they got wrong. But my encouragement around regulation, I think there's a positive role that regulation can and should play. Just my encouragement for the United States, for other countries, governments that are considering things, is really to balance optimism and responsibility and take a forward leaning approach of what are the positive implications of this. But then with that element of responsibility to make sure as many people come along with the change as possible.

This is where AI can get a little scary and maybe where we really do need some regulations. Let's drop in on the conversation where Jeff weighs in on the Industrial military complex. One of the fears that I have, the military is going to want to use this. They probably already are. Maybe first to find better ways to protect us. But let's be honest, better ways to kill the enemy. And if we're using it, so are they.

Yep.

What's out there? I mean, I worry about some maniac being able to, you know, bioengineer, you know, some kind of new, new disease that's going to kill us all. There are enemies out there and if they have this, how afraid should we, should we be?

Yeah, I mean, I think it's the, the reality of the modern era and AI is being used already on the, on the battlefield. If you think about applications of, of computer vision and the automation of drones and applications like that, it's already here. And that said, I think it's going to be nearly impossible to regulate that because turns out that bad actors don't pay attention to regulations or laws or guidance. So it is the reality. Again though, I think the positive side of it is it can be used for signal intelligence for, for defensive applications. So there is on the positive side of it, sort of a positive arms race of the ways that these can be used for defensive and peace enabling solutions as well.

Reagan's dream back with Star wars some 30, 40 years ago, even longer was that there be a system in place that could end the idea of, you know, MIRV warheads flying all over the world and knocking out dozens and dozens of cities at a time that we'd be able to defend ourselves. Is that a reality? Could that even happen.

In limited applications? Potentially. If you look at Israel's now famous Iron Dome, it steps that direction. I think the reality of it is it's hard for any system to ever be perfect or impervious. So I would say it's still a work in process. That said, unlikely to ever be completely accomplished because adversaries come up with more advanced capabilities as well. That's the literal definition of arms race.

A lot of AI and I'm learning it as well, I'm using it in my work and it's pretty fascinating. But I'm trying to understand the difference between there seems to be artificial intelligence designed for specific tasks, then there is generative AI. And if I'm understanding that right, it creates new content and data and in some ways mimics human intelligence that scares a lot of people, including me. Are we going to get there? Are we going to be able to tell the difference between the two?

So I think it's useful to do a Little bit of defining of what we mean by the term AI and then how AI has evolved over time. So interestingly, AI as a term is one that I've historically resisted using because I'm an engineer by background. And to me AI has been much more of a, essentially a marketing term than something that means anything in actual underlying technology.

And what would you call it if you were going to call.

So I have finally conceded basically in the last year, because AI is so broadly used, but my personal definition of it is I use it as an umbrella term that encompasses underlying capabilities and technologies. And if you look at the kind of last 20 years of AI, I think there's been three pretty distinct chapters of AI. 20 years ago, 2003, 2004, when I was at Google, my teams built some of the first AI systems at Google. And those were machine learning systems that were really kind of statistical regression based models where they were predicting future events based on historic data. So call that the era of machine learning. And there you had kind of specific applications developed for specific purposes. So after we developed the first one, there were then 30 or 40 machine learning systems that proliferated across Google for things like spam detection in Gmail or fraud detection in transactions, building on that first system that we built for ads quality or click through rate prediction. A second era was about a decade later, which was the era of deep learning. And that was using neural network technologies that in some respects, yes, they were still big complex systems, but. They were a level of conceptually more simple than the very purpose built machine learning systems.

You'll get a kick out of this. Listen to Jeff's comments about venture capitalist Marc Andreessen.

And that led me to coin a phrase at Google that built on another observation from industry. I don't know if you remember from kind of 2010, 2011 or so, Marc Andreessen and Andreessen Horowitz was very famous for coining the phrase that software eats the world, that everything was kind of being software platformified. My observation at Google was that AI eats software because we went from having 30 or 40 different uniquely developed systems to being able to have a common system with deep learning based systems where it was common code still a specific instance for each application and tune for each application, but common code underneath. So AI eats software. The next chapter that we've just entered into or has exploded now into broader public consciousness is as you mentioned, is generative AI using large language model systems. And if you look at those systems, they're the next level of conceptual simplicity. But the Big difference is the amount of data and the amount of compute that you throw at the problems.

Is it bigger or larger?

It is bigger and larger. It's far, far, far more data and far more compute that's being, being applied. And that's led us at Triatomic to kind of go the next chapter of observation, which is in these AI systems now the output is entirely defined by the data that you feed it. So we've encapsulated that as in AI systems now, the data is the code. So back to your question about AI, kind of, that's the umbrella encapsulating that machine learning, deep learning, now generative AI, large language model systems. There are other variations underneath those, but those are the largest categories. And in each case, as with the evolution of media, television didn't replace radio, it supplemented it. The Internet didn't replace television, it supplemented it. Each of these models, as it goes along, or eras as it goes along, supplements what was there previously.

Marc Andreessen also said that this would be bigger than the Internet. Do you think that's true?

I will echo. Actually, the most famous observation that resonates for me is John Doar, in the mid kind of 1990s or late 1990s with the first Internet boom, said famously that the Internet was under hyped when everyone thought it was a bubble. And he ultimately came to be proven true.

The only thing that seemed to have become a bubble was the stocks from that period of time. Correct. But the Internet.

But the implications of it are significant. And I, and we as a firm at Triatomic, think that indeed AI is a big deal. I mean, we've compared it to, if you think in more kind of century terms, century defining terms, the introduction of electricity in the. Not the introduction, but the broad scale of adoption of electricity that happened across the 1920s and 1930s. Electricity had existed previously, but that's when the penetration really took off, both residential and commercial. And Certainly by the 1930s, 1940s, companies didn't say they used electricity because it was so obvious and implicit and had value. We think that AI is that kind of inflection now happening in the 2000s, where just about every company is going to be an AI company, whether they recognize it or not. We do think that there will be significant differentiation between the ones that do it well and the ones that lag.

I hope you enjoyed today's interview and of course you know what comes next. This is the part where I ask for your support and know it doesn't cost money, if you like today's podcast. Hit subscribe and let us know what you think. Also, don't forget to visit me on Substack, where I publish my blog and research. You'll find articles, charts, audio and video. Thanks for joining. I'm David Nelson.