Jeff Wang, MBA 20 – Diving Deep Into the World of AI

(Transcripts may contain a few typographical errors due to audio quality during the podcast recording.)

[00:08] Sean: Welcome to the OneHaas Alumni Podcast. I’m your host, Sean Li. And today, we’re joined by Jeff Wang. Jeff is the head of business at Codeium and also a host at the RocketFuel show. What would you… I’m sorry, RocketFuel…

[00:24] Jeff: I guess it’s like an education platform. So, it’s a host and it’s the writer. It’s the guy that just tries to keep everything afloat, which sounds like my role at Codeium, actually.

[00:34] Sean: I love it. Well, before we get into all that, Jeff, we love to start these podcasts hearing about your background, your origin story. So, let’s start with that—where you grew up, how you grew up.

[00:44] Jeff: Yeah, I guess I grew up mostly in Chicago, and then try to do… when I grew up, I wanted to be a film director, which is kind of odd. So, trying to do action movies, just experimenting with whatever the latest technology was. And whenever… and by the way, back then it was very hard to make movies. Right now, with the iPhone era and you can even edit on the phone, that’s just insane to me. Back then, I was using, I think of the 8-millimeter types of tapes. And I was connecting it to a very slow PC. I think the PC had maybe 20 gigabytes of storage. So, you couldn’t even store that much footage to edit. And then I was proud to have made some videos, but I could not even make more because there’s just no storage left on the computer.

[01:23] Sean: Wait, we have to rewind on that. I mean, what inspired that?

[01:26] Jeff: I guess it’s probably just from watching a lot of movies, and because the weather is terrible in Chicago, there’s not much else to do at that time. And technology was not great at that time, too. So, just being inspired by whatever was available to me and seeing if I could replicate that or do that, too.

[01:40] Sean: Nice. And so, what would you do next? Clearly, you’re not in film.

[01:45] Jeff: Yeah. I think, I mean, just hobbies-wise, film and martial arts, anything action-movie-oriented, probably, was my childhood. And then, going to college, I think my parents were like, “Yeah, you could be a director as long as you do engineering,” so basically went to Michigan to do a engineering degree.

And it’s funny because I, at this point in my life, was only exposed to bad weather. And then, finally, when I moved to California for my first job, I was like, “Oh, my gosh I’ve been missing out. Why the heck was I living in the Midwest?” And then, basically, now, I think I’m just going to stay in anywhere that has good weather, at least for the duration of my life.

[02:18] Sean: Did you end up going into engineering after school? What was this…

[02:21] Jeff: Not really. I think I was chasing titles when I was young, just really dumb and young. I think, actually, by the way, any sign of growth is you look at the previous few years and you’re like, “Oh, man, I was so dumb.” But anyway, I was really dumb when I was out of college and I was just going for those management or program management or project management kind of roles.

And I was working at Cisco at the time. And I think, working at a big company, you get those titles and you feel a bit of pride. And then after that, I went to Salesforce for a similar stuff. But I was always feeling like I was losing my soul. I was like, “Hey, I’m just in a big company. I’m not doing much. I’m not actually building net new GDP or net new products that I think are changing the world, right?” And I think that is what signaled me to go back to school. This is where we met again. By the way, we met each other in undergrad, which is interesting, but we encountered each other again at Haas Berkeley for our MBA. And that’s when I started my startup adventures after that.

[03:18] Sean: Oh, I didn’t know that. I thought you had actually had a foray in startups beforehand.

[03:22] Jeff: No, I was actually very risk averse. I was like, “No, I have a stable income now. What if I go to a startup and it fails, then I’m totally screwed, right?” And I have no skill sets, either, right? That’s what I was thinking in my head.

[03:33] Sean: So, I guess, how did Haas change that? And I guess, even before that, what made you want to go to Haas?

[03:39] Jeff: I think, again, the fact that I had so much free time and I didn’t feel like I had any skill sets. And maybe that’s not true. Maybe, it’s just internally how I felt. Going to Haas was something that I, at least, could open some doors, maybe add something to the tool belt. And I think, actually, in retrospect, Haas was probably the place to open doors, is probably the biggest value that it can bring. And meeting people that also were in the entrepreneurial spirit. And I think, at least two or three of the opportunities after just came from other classmates who were trying to build something. And I think that’s saying something that, yeah, opening doors is not because of just going to a class. It really is people that you interact with and have common values or common alignments of what you want to build.

[04:21] Sean: That’s awesome. All right, what’d you do after Haas?

[04:24] Jeff: Yeah. I think one of the classmates, Brett Li, he actually was at a startup called Tonkean. And that was an API automation startup. And I actually didn’t know what I was getting myself into, but it was, I think I was 25th employee or something or 20th employee, and basically, built a lot of stuff from the ground up.

They had just raised their Series A. And I was essentially, majority of my time was spent interacting with customers and trying to make them understand the technical capabilities, like half sales, half solution architect kind of role. And then, on the side of it, I was making sure everything was automated internally, filling in the gaps across the whole business.

Yeah, it was a wild ride. I remember this was also right before COVID. And I would sit down at home, just do a whole bunch of stuff on the computer, and then it’d be nighttime. It was kind of crazy, but it was also really fun at the same time. So, the opposite experience of working at a large company, but at least feeling like there was a lot at stake and a lot to be done. So, it’s weird how that works. Maybe, I overcorrected.

[05:20] Sean: That was your first experience in a smaller company, right? Because you went from Salesforce to Tonkean.

[05:31] Jeff: Yeah, exactly. Huge shift, complete polar opposite in terms of the type of experience. On the side, I always had side projects, too. So, another classmate, Steven Hubbard, right from your class, we started RocketFuel Education. And again, that was more of a hobby, a side gig, because I like to rant a lot. I like to talk a lot. Hopefully, that doesn’t come across here on the podcast episode. But the stuff I talked about had some uniqueness, I suppose, like some forward-looking insights, some predictive insights. And I think that kind of rant is valuable, right? People would want to hear about that. At least, that’s what the thesis was.

And from there, we converted that into lessons on the crypto markets. And now, it’s more crypto, macro, and AI markets and just really understanding what’s going on, and then, again, being predictive. And if I’m accurate, sometimes, that’s great. And sometimes, I’ll be wrong. But I think, over time, I think, just having the understanding and really understanding how markets work on RocketFuel Education, that is why people stick to it.

And obviously, my time is completely sunk in with my latest company at Codeium, but people still depend on that information to understand the world. So, at least, it keeps me up to date as well. So, I still write about it and still do videos about this.

[06:37] Sean: RocketFuel, it started out as a newsletter, right, if I remember correctly?

[06:42] Jeff: That’s right. I actually looked up my first issue and it was so bad. There’s so many typos. It was very short. And then I think it evolved into being very comprehensive and being… actually, not just comprehensive, but very concise with how much you covered.

So, just being very direct, very but predictive, like, “Here’s the prices where we think things are going. If it gets this point, it’s too expensive. If it drops this value, you should buy.” Obviously, not financial advice, but at least any other newsletter out there is not going to try to make those calls or expectations of what might happen.

And I think that’s, maybe, why it’s unique is to actually make those bets and then be wrong or be right. But at least, you really understand why that number is called out. And then, honestly, I think I’ve been very contrarian to the market as well. A lot of times, I’m very much opposite of what everyone’s thinking, but in the right way. And I think that’s constant theme of just any investment. You got to think a little bit different than everybody else, because what everybody else is thinking is the market price, right? And that strategy has worked out very well in terms of timing the market and we’re predicting where it’s going as well.

[07:43] Sean: I guess, to dig into that a little bit, from my memory back in 2016, ‘17, Bitcoin was, I feel like, blowing up in the public eye, right? And I just remember actually sitting in class, fall 2017, just everybody was… somebody, I felt like, and everybody was trading Bitcoins. Because for the, I don’t want to say for the first time, but it felt like, for the first time, Coinbase just made it so easy or more accessible. And most people just trade Crypto and consume this information.

But what inspired you to actually go write about it and share information? That’s a pretty interesting thing to dive into, to want to share that information that’s in your head.

[08:30] Jeff: I mean, the funny thing is, even when I was working at Salesforce, I was just always thinking about, hey, there’s a huge opportunity here, and how do we get an edge, essentially? If everybody has the same amount of information on an asset that’s blowing up, where should we be really looking?

And a lot of that was like, hey, if you join these crypto communities, you actually get some really good insights at what is upcoming that nobody else is going to be joining. If you go to these crypto conferences, you meet people that are actually the CEOs of these projects. And you can see if they’re for real or not.

Or, you could even meet the CEOs of projects that had not even been released yet. And you could actually invest in those companies as well. So, you can get an edge of just being very early, right? And a lot of those interactions compelled me to be like, “Hey, spending all this time, doing all this research, at least, I should put a brain dump of that somewhere, at least monthly, right?”

And that’s where it started, was like, hey, I’m becoming really involved in the ecosystem. I’m starting to invest in some of these companies early. I don’t want to have a crypto job, because in my head I’m like, “Hey, it could be a bubble. And if it’s a bubble, I don’t want to be working at a company that’s on the downside.” You want something that’s predictable, at least, even if it’s a startup.

 And I think that’s a big difference with AI, right? Crypto, it does come in phases. And it’s like, every few months, or every few years, I should say, there’s this rapid excitement everybody jumps on. And then there’s a moment where, for example, today, in today’s market, does seem like a lot of the excitement’s baked in. And then, maybe, the next year is going to be a little bit rough compared to the last year. And I wouldn’t want to get involved in an industry like that. And investing in that’s fine and writing about it’s fine.

But AI, I do think, is a little bit different because it does actually bring value to the user, the end user. And we’re still in the early phases of AI as well, because you can see a lot of these AI companies don’t even know how to monetize it yet. And they want to give it away for free, and they want to spend a lot of money to build these AI companies.

And by the way, very similar to crypto in the early days as well, where it’s like, “Oh, this technology is promising, but we don’t know how to monetize it yet. And we’re going to go big. We’re going to go big on marketing.” So, you hear a lot of noise on both crypto and AI. But honestly, just being deeply ingrained in both, I can see that the AI industry really has some legs to it. There’s going to be a lot of stuff that’s going to be coming out of it in the next couple of years.

[10:41] Sean: What was your first touch point with AI?

[10:42] Jeff: I think the first was actually Stable Diffusion, so you can generate an image from scratch. And I’ve used OpenAI’s GPT-2 the prior year before ChatGPT came out. And I didn’t think it was good enough. It was better than NLP stuff. Basically, NLP is using other methods to predict either sentiment or the next, like categorizing something, predicting if it’s A, B, or C, based off of just natural language.

And using transformers was a huge change. So, I started with GPT-2, I guess, didn’t really think it was that good, but it was still amazing. It just wouldn’t really work in the real world. And then I went to the fusion models, which is generating images. And I did that as a hobby. Even all my newsletter images are all AI-generated from my own computer.

And then, when ChatGPT came out, then it was a big turning point where I was like, “Holy crap, this is at the point where it’s real.” Anybody can use ChatGPT for, really, anything in humanity. All of humanity’s text is trained on this one little model that can fit on a USB drive, right? I mean, that’s really amazing. And even the best capabilities of ChatGPT or the GPT API have not even been discovered yet.

But in terms of why I jumped into the startup world, coding is definitely one of those use cases that’s extremely strong in how the technology works. And that’s where I place my bet of my time and energy right now.

[11:58] Sean: It’s so funny, I was just going to ask you this question, what is your favorite application of AI, so far? It sounds like coding is number one. What else? I guess, for listeners still trying to wrap their head around AI and ChatGPT and whatnots and all these things coming out, and I was just reading something about, I think, Meta trying to incorporate AI into an earpiece with a camera or something like that, in your eyes, what are the most useful applications of AI, aside from coding, obviously?

[12:26] Jeff: Yeah, I think there’s… you have to break it between enterprise and consumer. And what is actually good for startups is that the consumer is being targeted by the large companies like Google, Meta, and OpenAI. I mean, they’re focused on products that are really for the wide range of individual users, right?

And for enterprise, that’s targeted towards teams and how much value, business value, you get. So, just to split it up, like enterprise, I think the best use cases are going to be on the coding and the support phase. So, support would be answering tickets or categorizing tickets. Even the chat between customers that are coming in, you can probably get a lot more faster response times now, a much better experience for the customer. And for coding, right now, you can probably make a developer 1.3X faster or at least more productive.

And I think some of the capabilities that we are releasing soon and what others have announced will eventually 10X a developer. So, you can imagine a company is going to be rolling out features 10X faster than they were before. It’s really amazing to think about. Obviously, once you roll out one thing faster, the rest of the pipeline has to be faster, too. So, we’re also thinking about that.

Then, on the consumer side, I think there’s a lot of things about summarizing topics, like, “Oh, I don’t want to watch this YouTube video. I don’t want to watch this episode with Jeff and Sean. Summarize it for me.” And it will just do that. It will just save the user a lot of time by just giving them the bullet points of the conversation.

And there are things like therapy and having someone to talk to, with GPT 4.0 or 4.0, I should say, you can actually have a real-time conversation with ChatGPT, right? These are all consumer use cases that I think are very promising to start. I just don’t know the ROI of some of the consumer use cases, but I know that, if you go big, there will be some value, right?

[14:06] Sean: I’m going to ask you a tough question here, and you don’t have to answer this. What’s going to prevent AI, and this is hypothetical, from being misused, to actually figure out ways to, let’s say, hack the code? I have Chase Bank’s… somehow, I got Chase Bank’s code database and then, I’m just like, “Hey, help me look for vulnerabilities here.”

[14:30] Jeff: No, I think every AI tool can be used in however the user intends it to use, right? So, if you’re… and then you could go as granular as you want. So, if you say like, “Oh, write a script,” let’s just say you take a step back, “write a script to help me hack into JPMC.” That’s something that probably the AI has prompt guardrails in. And maybe you can trick it to still figure it out. But the user intent was there to do that. And then if the user couldn’t make that prompt, they could even go more granular. “Oh, give me a for loop that is looking through different fields,” something very granular. And then they go from there, “Oh, take this function and apply it to a login screen or something.” I’m just making up some ways to hack or make different call-outs to different IP addresses in this list. It doesn’t know that you’re trying to do a hack, but the user has very strong intent. And it’s just going more granular, AI is just not going to know what they’re trying to do. And they can accomplish the grander picture, right?

So, I would not look at this as AI is smart enough to break into systems or cause damage. I would look at it more like, hey, these are acceleration tools that can actually help the user of whatever they’re trying to do. And the user might be doing something very granular that might add up to something bad. For example, “write me an email that will make me vote for one political candidate over another, and then write me a script that sends emails very fast.” It does not know that it’s AI that’s… yeah, the intent is not clear from each granular action, but that grander action is not good intent, right?

It’s like, that, if I said, “Hey, help me make sure the Bay Area votes for another political candidate. What do I need to do?” That’s a higher-level prompt, but you can never really prevent someone from just making each of the pieces of the prompt. And again, it’s all user intent in that case.

[16:11] Sean: Right. It’s just such an interesting conversation, especially, I think, for most of the listeners, lay people like myself, we’re just trying to wrap our heads around this, but it is fascinating how it’s evolving. It’s fascinating how fast it’s moving. And in terms of its implications for society at large, I’m pretty optimistic about it.

[16:33] Jeff: I mean, this is, probably, maybe the biggest… the internet was in our lifetime, but this maybe has the potential to be even more important than the internet. The use cases now, almost seems like there’s no ceiling to what this can do. And I think, even with the internet at the time, there was like, “Oh, the internet can do peer-to-peer communication. That’s great.”

I think, right now, it’s only been a year into, or a year and a half, since ChatGPT went out. And there’s already been a lot of amazing products and things that really accelerate human productivity. And I think, when this starts getting applied to science, like research, even, again, mental health, let’s say, just all these things that it can be very good for society for that, I mean, these things have not even been really fully discovered yet. And I think, as the years pass by, the technology is just going to get even more grander, more better. And that’s also because compute capabilities are also accelerating.

You’ve heard of Moore’s Law, right? The amount of transistors doubles every year. Moore’s Law is out the window. I mean, with GPUs, I think they’re up 50X in terms of speed over the last couple of years. And you just don’t get that kind of jump in capabilities from the infrastructure layer and not get something amazing from the application layer. And we’re seeing results of that. We’re seeing the application layer now starting to show what you can do with the hardware.

[17:55] Sean: As a futurist here, I’m really curious to hear how you envision the future with AI. And I asked this question because there’s plenty of doom-and-gloom people can read. And I’m not a big fan of doom-and-gloom because it just ends the conversation, right? It’s like, “All right. Well, they’re going to take over. We’re all just not going to exist. Okay, end of story. There’s nothing to talk about there.” Other than just stop it and hide it, or some kind of ludicrous thing. But if we’re to really, as entrepreneurs, as futurists, what is your positive vision for AI and human race?

[18:34] Jeff: Everywhere that we are stuck in right now, even if it’s due to physics or if it’s due to just manpower, anywhere that humanity has slowed down, I think AI is just going to speed it back up again. Just think about, if I could add more headcount to any problem that humanity is facing, I think AI is that solution. And previously, we never had that capability to do that before.

And just from short-term, even, having a multi-step or agentic AI is a big deal. Meaning, right now, it’s a single prompt and a single response back. I think, just over… even starting today, even starting the next couple of months, you’re going to see apps that are going to be like, “Hey, I want to do something. Here’s a prompt,” and then there’s a bunch of things that the AI does to accomplish that prompt.

So, it’s like, “Hey, I need to book a trip to Barcelona next week,” let’s say. And then it’s like, “All right, here’s your itinerary. We already booked everything for you, and based off your preferences, too. I know you like to eat this kind of food. We already have the restaurant book for you.”

And that can be done today already. It’s just a matter of getting it more accurate. So, over time, the reasoning behind the AI is going to get better, too. The reasoning is not very good today, but I’m pretty sure, when GPT-5 comes out or the next iteration of LLMs come out, you’re going to find that the AI actually knows you very well and it can make decisions without too much friction or too much inaccuracy. And that’s, again, this is within the next few months, I’m telling you, this kind of capability. That is going to be the next “oh, my God” moment in AI. Right now, we had our “oh, my God” moment a year and a half ago. That next moment is coming up very soon. And I know this because we are already developing products that are accounting for multi-step inferences and multi-step capabilities.

[20:14] Sean: That’s so interesting. And I love that. And a lot of it, remember early on, the fear was, “Oh, are computers or AI going to dictate our lives and make decisions for us?” And always it’s not a clear cut answers. It’s not a yes or no, because most people, I remember reading somewhere, don’t even realize they already make a lot of decisions for us. And it’s completely beneficial.

Take for example Waze or any kind of map direction thing. They’re telling us exactly where to go, right? I’m not thinking about it. They’re telling me to turn right and I turn right. And that’s just a really simple example that’s been around for over a decade now, right?

[20:50] Jeff: Yeah, I think people actually ask me, what if people just get worse at basic coding? And then, my response, actually, to one person one time was like, “Hey, what if people get worse at reading maps because of GPS?” So, maybe we shouldn’t roll out GPS. Is that what you’re telling me? Is that your logic?

[21:07] Sean: Right.

[21:07] Jeff: I think that same thing, the skill set to read maps is completely irrelevant now, completely irrelevant. Nobody needs to know how to read maps anymore. And I think the similar thing is going to happen to AI where there’s a lot of basic skills that are just unnecessary anymore. And you can focus on the more advanced skills. And that might be harder, initially, because what if you need the simple skills to learn the advanced skills, right? That is going to be probably some degree of controversy of some of these AI tools that are coming out.

[21:31] Sean: But that’s the interesting thing, is that what Google did for us 15, 20-plus years ago now is it gave us access to information that billionaires didn’t even have pre-2000. And now, it’s just like you have all this access to all this information.

And then, for the past 20 years, it’s like, all right, I have all this information, but I can’t process it all of it fast enough to make any meaningful decisions or to learn faster and whatnots. And AI is really taking that next step.

One of the random ideas I had last year when AI popped up was around a personal cloud, in the sense that it’s self-hosted, just more private versus having my information, everything that I would want to feed, a ChatGPT, somewhere else. And so, it was just this fascinating mind exercise of, all right, in the future, are we going to have our own data feed into our own personal LLMs or AI, and how that’s going to work and an entire suite of products and businesses that’s going to arise from that, right? I don’t know. I’m just excited.

[22:37] Jeff: I guarantee you, this will be happening in the next 18 months, I bet you. So, the reason why is because Microsoft and Apple have already released research papers and results from smaller LLMs, meaning 3B or 1.8B parameters, just the sizing of the models that will fit on a phone or will fit on a laptop and actually be very capable, very close to GPT-3, not maybe in the future GPT-4 capabilities. But to your point, this is self-hosted. This is on your device. It’s not communicating to the internet on any information. It is using your calendar, potentially even your emails, your data to really be a personalized assistant or something that can be used on device, as you said.

So, your vision, I would bet it is in the next 18 months. The reason why I say 18 months is because the chipsets need to be designed and be good at power management and then deploy and then run these LLMs at a efficient way to not drain the battery. So, I predict this year, Apple will have an iPhone. They’ll get rid of Siri, actually. They’re just going to say a new name, or maybe it’s Siri 2.0, that is just using a OpenAI. And then, the following year, they’ll say, “Oh, now, we have the on-prem or the on-device Siri. And now, we’ve actually got the chipset for the phone that it can actually run it.”

So, your vision is not far off. And this is a problem with the AI entrepreneur, because all the ideas they have, you have a very short time to get it launched. because most likely somebody else is working on it or the bigger companies are already done and they’re just waiting to release it.

Every time OpenAI releases something, an announcement, I bet you, all this AI startups are shaking a little bit, like, “Oh, my God, are they going to announce our product? Are they going to kill us?” And to some degree, yes. Every time they announce something, it kills off a few AI startups.

And the same is going to happen with Google and Microsoft and even Meta, but the difference with Meta is they’re actually releasing all their models for free for anybody to use. So, they want everybody else to use their stack, but they’re going to have… they’re obviously going to utilize it for themselves on their own chatbots and social media.

[24:37] Sean: Yeah. Talk about startups shaking their boots. Just thinking or hearing about Google shaking their boots with OpenAI, it’s like, “I don’t need to Google anymore. I just ask OpenAI, right? I mean, I just ask ChatGPT. Why do I need to Google?”

And I saw them actually push out, now that my top Google search results, I just started seeing that this week, my top search Google result is Google’s AI response versus pages.

[25:03] Jeff: It’s not even AI startups, right? It’s even pretty well-established companies. So, for example, last week GPT 4.0 has real-time communication. It could translate in real time. Me and you could be talking in different languages and then this GPT 4.0 is just going to translate for us.

Duolingo, their stock dropped 5% immediately. Because it’s like, “Hey, we don’t really need a Google Translate or Duolingo anymore. We could just use this model.” And then, you’ve heard of things like Stack Overflow or Quora. All these things are in trouble, right? Any tutoring app is in trouble, because I could just go to ChatGPT and ask any question. And then any learning startup or company is going to lose. I wouldn’t expect them to gain market share, is what I’ll say, for some of these AI tools.

So, it’s not just startups. It’s really a whole ecosystem of tools and companies that are in trouble, as these AI tools get more powerful.

[25:51] Sean: I’m going to ask you another tough question. At least, I feel like it’s tough for me. As head of biz, how do you stay competitive, then? Following what we’re just talking about, how does an AI company stay in the chance and stay competitive?

[26:05] Jeff: It’s a good thing I went to Haas. I’m just kidding. I think that the main thing is to stay differentiated. And then the second thing is find where we are strong at, and then really push products that are leveraging that strength. And I can tell you our strength is infrastructure. So, there’s probably no better team at running very efficient models and having very low latency and running them at scale. So, how do we build a product that’s leveraging those factors? There’s no way that another company is going to build as fast as this team purely because of the number of hours they spend. And then, from a end product standpoint, if we roll out something that nobody else has and we push that and there is demand for that product, obviously, that is one way to continue to build market share.

And then, every month, we’re going to have an expectation that, in 24 hours, everything could change, like Microsoft or another startup rolls out a product that’s better or fundamentally makes our product look bad. And then we have had deal cycles where, “Oh, actually we just saw this Twitter demo. So, we’re going to wait. Instead of renewal, we’re just going to just pause and see how that one develops all the time.”

All the time, all the time, there’s a new startup with another idea that has a great demo, and then whether or not it becomes a real product doesn’t matter. It’s diluting the message of what our tool can do today and where we’re going tomorrow.

And that doesn’t matter. What we have to do is focus on… we can’t fight ghosts. We focus on what our product can do. We go to market. We talk to customers. We give them what they want. We continue to surprise them with our newer features as we roll those out. And then we just keep repeating that over and over.

And like I said, there’s going to be times where it looks like, “Oh, crap, another startup or another big company has something similar.” But again, we cannot fight something that doesn’t exist. And when it does exist, then we remain differentiated again. We got to just go in another direction where we can always win, right?

[27:49] Sean: I love that. No, that’s really astute. So, definitely, for our listeners, we’ll include a link. Go, check out rocketfuel.edu.com to subscribe to Jeff’s newsletter and just learn, just stay-up-to date about everything in Jeff’s brain. I feel like there’s so much in there that we still haven’t even heard about, but that’s wonderful. And if you’re a company looking to code faster, check out Codeium, obviously.

[28:14] Jeff: Codeium with an “E,” by the way. So, there’s another one without the “e,” C-O-D-E-I-U-M.

[28:21] Sean: Okay, actually, Jeff, before we go, Aside from your own newsletter, any other sources of things you recommend people read? Any books on AI that you recommend?

[28:33] Jeff: I actually have an AI blog, too. It’s just jeffwang.substack.com. It’s just a free place where I just, sometimes, I take notes of some experiments I’m doing, and I didn’t want those to go to waste. So, I actually jot down how to run your own models, how to do your own image generation, how to use some of those extensions as well. And then, also, some macro-level views of where AI is going and why NVIDIA and Microsoft are probably just going to eat everybody up. Stuff like that is in that blog, too. I just wanted a place where I could dump my thoughts and show notes of playing with other tools.

[29:04] Sean: I’m a geek about this. What note-taking tool to use for your personal knowledge base?

[29:10] Jeff: Well, for AI, it’s Substack now, but honestly, I don’t really have a good note-taking capability.

[29:17] Sean: You don’t use Obsidian or any…

[29:19] Jeff: Not really. I wish I was better at this, actually, creating a knowledge base of everything that I found was very interesting. But instead, I just, again, write the newsletters out. And that’s pretty much everything that I’m thinking about. It’s like, “Here’s what’s on my mind across all of the landscape of AI, crypto, and the markets.” And that’s my repository, at least, for when I look back at, like, “Oh, there’s this new startup. Oh, I remember there’s another startup I wrote about. Is it similar?” And I just search my newsletter and I’m like, “Oh, okay. I did write about it.”

[29:47] Sean: That’s awesome. Aside from all this stuff, what do you see for yourself in the near future?

[29:52] Jeff: That’s a great question. I think I was wandering in life without a purpose, found the groove with startups.

[29:58] Sean: That’s hard to believe, by the way.

[30:01] Jeff: And honestly, having a family now, I have two kids now, that’s keeping me busy. I don’t actually know what’s next. I think I’ll probably be at Codeium for a while. But then I don’t really know what’s next. I feel like, right now, I’ve just been finding something very interesting every few years in terms of technology.

And that always gets me really passionate. And I spend a lot of time just reading and researching about it. So, maybe, in another few years, there’s going to be another breakthrough in some other industry. And maybe, I get obsessed with that, too. But who knows, right?

[30:29] Sean: Hey, since you brought the kids and we’re talking about AI, I have to ask you this. How would you want your kids to use AI? We both have young kids. But do you envision our kids growing up with AI?

[30:42] Jeff: I think they’re going to ask a lot of help for homework and tutoring. I think, right now, in its current state, even GPT 4.0 is a good tutor. A kid can just ask a lot of questions about what they’re reading about and it’s going to give you pretty accurate responses. So, I think that’s step one.

Step two is more entertainment purposes, like, “Hey, read me a bedtime story or say something funny.” Those are things that can occupy a kid’s time. I don’t know if that’s a positive impact. I feel like the tutoring and even the knowledge sharing and learning is very positive. I wish I had that as a kid. I even wish I had the internet as a kid, right? We were old enough, where we only saw that later. 

[31:18] Sean: You can start with dialogue.

[31:19] Jeff: I would imagine the rate at which kids learn is going to be insane with these AI tools. And then, eventually, there’s going to be applications where it’s going to be even more interactive than just talking to it. So, I’m very positive about where AI is going to do for kids. I can only expect them to gain from it instead of lose.

[31:37] Sean: I mean, for me, personally, my outlook on this, and I’ve had this outlook since, honestly, since Google, is I find that the people that do well in life know how to ask Google good questions. It’s like, how do you become a better Googler is actually a skill. And I think that the next skill isn’t like, “All right, what knowledge can I retain, but can I learn to ask better questions to ChatGPT, to AI? Can I give better prompts so that I get better information in return?” I feel like that is going to be a really important skill.

[32:10] Jeff: For sure, for sure. Even today’s AI tooling, there is a skill difference between people that are able to use them and people that are just trying to, at face value, try to do what they’re trying to do, but not getting that much value from it. That just tells me that they just don’t really know how to use or prompt the tools.

[32:26] Sean: Right. that’s fascinating. Well, Jeff, thank you so much for coming on the show today. It was a real pleasure having you on.

[32:32] Jeff: Yeah. Thanks for having me. That was really fun.

[32:38] Sean: Thanks again for tuning in to this episode of the OneHaas Podcast. If you’ve enjoyed our show today, please hit that Subscribe or Follow button on your favorite podcast player. We’d also really appreciate you giving us a five-star rating and review. If you’re looking for more content, please check our website at haas.fm. That’s spelled H-A-A-S.fm. There, you can subscribe for our monthly newsletter and check out some of our other Berkeley Haas podcasts.

OneHaas Podcast is a production of the Haas School of Business and produced by University FM.

Until next time. Go, Bears!