Unlock the limitless potential of AI in the data universe in this riveting episode of Edge of AI. Join us as we delve into how AI revolutionizes data exploration, supercharges information discovery, and propels businesses forward. Our guide through this AI data frontier is Brad Schneider, founder of Nomad Data, and the mind behind groundbreaking AI innovations and an exceptional thinker on the future of data. In this conversation, you’ll discover how AI becomes the driving force behind data-related tasks, from finding and managing data effectively to eliminating the barriers to discovering valuable insights within vast datasets. Brad Schneider takes us on a journey to understand how this transformative technology can empower businesses, enhance decision-making, and ultimately change the way we interact with data. Tune in to learn the key takeaways from their discussion, and witness how the future of data and AI promises a new world of possibilities that will reshape industries and our understanding of information itself.
- AI plays a substantial role in data-related activities, offering assistance in efficiently discovering, handling, and utilizing data.
- Traditional keyword-based data searches are found to be inadequate as they often fail to capture relevant information due to the variations in terminology and descriptions.
- User-generated language serves as a conduit for describing real-world business challenges, with AI algorithms identifying datasets that align with these issues.
- AI systems continually learn and advance through iterative processes, with user feedback being instrumental in driving improvements.
- The potential of generalized intelligence, characteristic of human beings, lies in the versatility to adapt, reprogram, and think comprehensively.
- We need to be apprehensive about malevolent actors harnessing AI to manipulate data and fabricate deepfakes, intensifying the challenges of data security and verification.
- “The Myth of Artificial Intelligence” by Eric Larson serves as an insightful resource to gain an understanding of the current AI landscape.
- Brad Schneider: “Data is becoming the fuel of business, especially as we enter a world that has more and more uncertainty around it.”
- Brad Schneider: “It’s an exciting time to be involved in data and AI.”
- Brad Schneider: “For every product that required 15 clicks to get that what you needed, all those clicks are going away.”
- Brad Schneider: “AI is the ultimate generalist machine. It can be completely reprogrammed, it can adapt, and it has the ability to understand a wide range of knowledge.”
- Brad Schneider: “I’m an engineer by heart. Building things excites me.”
- Brad Schneider: “We are nowhere near an artificial general intelligence. That’s not a today problem. It’s probably not a tomorrow problem. It’s a way down the road problem, but it’s one we should start thinking about.”
Listen to the podcast here
Navigating The Data Universe: A Conversation With Brad Schneider, CEO Of Nomad Data
Welcome AI-enthusiasts. Join us, and we’ll introduce you to a data scientist with an intriguing habit, eating the same meals every single day. Discover how reading one book as a kid shaped his perspective. He will share insights on his belief on why consuming a wider range of information that you might suspect is the key to life. Fasten your seatbelts for a concise and insightful ride into the world of data science that will leave you with a fresh perspective on data and life. Don’t miss it.
I’ll be your captain for this voyage to the Edge of AI. As with most of you, I’ve embraced the spirit of exploration and entrepreneurship throughout my life, from starting my own business before graduating high school to traversing the world’s most challenging terrains. I’ve always sought out new frontiers and adventures. I built one of the largest award-winning custom home companies in Los Angeles. I’ve navigated complex regulations while founding and leading a public company dedicated to applying technology and training.
Buckle up and get ready. Let’s tackle uncharted territories in AI with curiosity as our guiding star. Our episode features Brad Schneider. He’s the Founder and CEO of Nomad Data, a platform to make data more discoverable and visible within any organization. Throughout his career, Brad has focused on using external alternative data to improve decision-making and prediction.
Prior to Nomad, Brad started, grew, and sold Adaptive Management, which was a data analytics company. Brad also spent more than a decade as a public market investor, investing in technology companies for firms, including Tiger Management and Jericho Capital, where he used external data to improve investment returns. For segment one, we’re going to kick this off a little bit. Nomad is a single AI-powered platform to run your data org and empower every business user to discover and manage data in minutes. Brad, maybe you can tell us about the trajectory of this project and a brief on how it works.
It’s probably not a surprise to anybody that data is becoming the fuel of business, especially as we enter a world that has more uncertainty around it. You need to enable people to find the data they need to understand the different pieces about the world and their business. Prior to Nomad Data, there was not a good way to do that.
People kept trying to solve this data discovery problem with the Yellow Pages model. That didn’t work for the internet and does not work for data. The main reason is that it’s very hard to compress a multi-terabyte data set into a one-paragraph description that captures what is possible with it. We built a natural language processing powered data search engine, which is the right way to think about it, where a user can literally describe in plain language, “I need data on wine sales volumes in the US going back five years.”
We have a network of almost 2,500 data-selling companies around the globe. Those include regular companies where their main business is something else, as well as many whose core business is selling data. That makes data a lot more discoverable, and we have a whole platform as you buy data and help to keep track of everything that’s happened in your firm around it, therefore, empowering people to get more out of the data once they do buy it.
Talk about a time that’s calm. We’re all afflicted with too much information these days. Typically, it is not the real information we need. We should have too much information and siphoning through it is nearly impossible. You said you have 2,500 sources of grabs you do on data, and that would be impressive on its own but useless if there wasn’t a way to manage all that and make it relevant to the user for the exact use that they’re looking for. There lies the magic. For me, if anything defines what AI can do and what it can do to create real value for us, that’s it. I don’t know if you want to elaborate a little more on anything I said there or make any comments in that direction.
AI is all about data, especially as you talk about cutting-edge LLM models. These are language models. Language models are typically operating on textual data. Textual data is probably the single largest category of data. Think about all the people filling out forms and typing things into spreadsheets. All of this data is not well-structured. Given that LLMs and AI are now at the forefront, this is the raw data you feed into these engines to create that structure or to look at the vast hordes of data you already have and do decision-making at scale.
A lot of these things were not possible before. Now, what is possible has changed dramatically. Imagine what the future will bring. You chatted about it in the beginning, but I used to be a technology investor. My job was to look at technology trends and understand how they would impact products and companies. I did that for quite a long time, and this is by far the biggest change I’ve ever seen with my own eyes. Also, I’m an avid reader of technology history. It’s one of the biggest changes that I can remember even reading about. It’s an exciting time to be involved in data and AI.
There’s no question about it. We’ve talked about it many times here at Edge of AI, and that is it’s the first birthday of ChatGPT, which is AI come to the masses, I’ll call it. You’ve been in AI a lot longer than that. Maybe you can back up a little bit and pick out the few bullet points that brought you to creating Nomad.
The real thing that I saw in the past was that most people would have a problem, and they wanted certain data that they may or may not know existed to solve that problem. There was no good way to find out if it existed. We found that once you showed someone, “Here’s a data set that solves the problem that you mentioned.” Their eyes would light up, and they would go ahead. They would purchase that data, build incredible products, and do amazing analyses based on it, but it wasn’t easy to find.
The problem was that the thing that didn’t exist was this descriptive information of not only what the data is but what can the data do. We started to collect that, and that is all textual. The challenge was, how do you look at the text of the person defining the problem and connect it to the text around the answer? That was a classic NLP problem, and we started working with much earlier versions of GPT to try to solve some of those problems.
At the time, it was not sophisticated enough. We ended up using things that people are still talking about now. Things around word vectorization and text similarity, but fast forward up until the beginning of 2023, we’ve re-evaluated everything and are now using GPT-4 to solve a lot of these issues but also to help us create structure out of the data that we’re collecting because one of the biggest engines in what we do is the text. People describe in text what they’re looking for. People describe in text their answer to the user’s inquiry. These questions and answers contain an enormous amount of useful information and modern AI. These LLMs basically allow you to get to that data in a way that you could not before.
When I think of Nomad, I think of an aggregator, but probably think of an aggregator in a couple of levels. You’re aggregating from, let’s say, 2,500 different data sources. I’m sure it’s probably even more than that, but you’re aggregating all that into one place, then you’re aggregating the request and doing a translation, I’ll call it, into something consumable. It could end up being one paragraph or one sentence or something small from all of that. That’s a use case that if you went to ChatGPT or some of these other AI services on your own to try and do that.
You’re not going to end up there. This is a niche product that, quite honestly, I believe once it gets out there, it’s not a small niche. It’s a very large niche. One thing I got out of Nomad is it’s not external data. If you look at a large corporation or large company, I don’t know whether that means from 500 employees to 50,000 or, in the case of some of the biggest companies, hundreds of thousands of employees, they have internal data that there’s plenty of gold in, but they can’t resource it or get to it. Maybe you can speak to that a little bit.
There are a few things to break down there. The word you used was the right one, which was translation. This is a translation problem. You have a set of language that people describe a business problem or a product problem, and you need to translate that into the data domain because data is not described in those terms. The heart of the Nomad engine is collecting that corpus of text of what is the business language to describe this data?
You talk a little bit about ChatGPT doesn’t work well for this. We announced a product based on GPT-4 that does just that, and we’re rolling this out to huge companies. They load in their list of data sets, whether they’re internal or external, it doesn’t matter, and then they expose a chat interface to any employee. They can say, “Do we have any information on roof ages in California?” It’s going to look through all the different data sets they buy, everything they have, and everyone’s notes. It’s going to answer them.
It’s going to say, “Yes, you have this internal data set. You purchased these three external data sets.” “How do I get access to them?” “Here’s the person that you would talk to.” “Who worked with it last?” You can get all that information all through natural language. We think that this is a future not only for us but for every product that requires fifteen clicks to know what you need.
I truly believe all those clicks are going away. Everything is going to be just describe what you need, whether it’s taking a picture of something or talking directly to the system or it’s typing in a command. It’s an interface people are familiar with. It gets rid of the need for every single person to learn something new. Everyone can learn how to type and ask for what they need.For every product that required 15 clicks to get that what you needed, all those clicks are going away. Click To Tweet
It’s amazing because everybody knows a lot of the big unicorn companies in the last couple of decades made money off of people’s data. They provided some service that allowed them to collect data on people, and they were able to capitalize on that data alone. That became the most valuable thing that they had, and a lot of value was created based on that. This is the next level to me that we’re going from personal data that is, quite honestly, now it’s in contention as to what they should be collecting and shouldn’t be collecting, what laws are out there, and what laws should be out there. Different jurisdictions.
If they pass something in one country in Europe, that tech company out of Silicon Valley that services the whole world now has to reprogram what they do because of one jurisdiction. It becomes problematic. What I hear that Nomad is doing is recognizing that the new generation of data is different. It’s not your personal data. It’s worldwide data. It could be city-wide. If you’re that roofer you talked about, maybe you want to target every roof that’s over 30 or 40 years old and be able to do that. That seems incredibly valuable, and I would suggest that most companies don’t realize that getting that level of data in a usable form is available yet. Do I have that right?
If I had to pinpoint the single largest challenge to the data market, that’s it. Most people do not believe that the data that they are looking for even exists. You have people giving up. Never connecting with the vendors that have what they need and that holds back the data market. There are a few mature verticals, let’s say, within the data market. One being financial data and the other being marketing data. We have, at least in our internal database, over 140 different verticals of data. It’s enormous.
Discovery is a critical piece. I use analogies in other markets. I think about the App Store. Before mobile phone app stores existed, you had apps living on all these different websites. You had to search around. You weren’t going to find anything. Even if you did, you would have a PhD to download that app onto your phone. People like Apple and Google come along, and they create true discovery. They came up with a standard for payment and listing, and all of a sudden, the market grew.
When I was covering that as an investor, it was a $100 million market. Now, it’s a multi-hundred billion dollar market. That is what happens when you solve these issues in a market. The funny thing about data is it’s not a $100 million market. It’s already tens if not over $100 billion. This is an enormous market. It is the largest, most dysfunctional market I have ever seen. That’s how important data is. Given the amount of friction, people are still willing to deal with it because this stuff is so important.
With LLMs, it’s only becoming increasingly important because a lot of the things that these LLMs do, people built entire companies around coming up with unique technology and unique algorithms. In a lot of cases, this blows up those business models because it can do so many things out of the box where you had custom applications and custom verticals built to solve.
A lot of these technology problems shift back to data problems. If you do not have access to unique data, your business is not going to exist because you will not be able to differentiate because the ability of anyone to build the most advanced features has catapulted and that’s only going to continue. The most obvious mode is data.
I’ve got a few questions that have gone through my head as I’m listening to what you’re saying, which is powerful. First, it was only a few years ago that we all used our first app. Try and imagine where the world is now. That’s not a very long time. There’s no question that AI is creating a faster and bigger snowball on AI.
As you said earlier, it’s happening at light speed. Talk to me a little bit about how this new situation with data and access to it is going to affect you. I’m looking at large companies, but I’m also looking at entrepreneurs and smaller companies. Is there going to be a bigger and better balance? Are they going to ride on their own rails, not touching each other? How do you see that dynamic?
There are a couple of different pieces there. One of the interesting things about what is going on is that typically, small companies tend to move faster, but the big companies have more trust, and they can spend more money on things. Usually, over time, that does create a meaningful competitive advantage. The one interesting thing about LLMs, though, is the idea that you don’t have unlimited access to an LLM.
If you’re using a model like a GPT-4 or you’re using Bard, you can’t make an unlimited number of requests to these services. You’re very much being rate-limited. If you’re a large company and you want to launch an interesting AI-GPT core-driven feature, you can’t even get access to the volume of calls to do that.
Even as a small company, it’s a challenge to get enough calls to serve your customer base. It creates this strange dynamic where larger companies are going to have to wait longer than they normally would have to embrace this new technology. Smaller companies can launch these features because they don’t need as much of that service upfront. Over time, as they gain more usage, they’re going to be able to train their own models using that data.
They’re going to be able to fine-tune their own models, lowering their incremental costs to deliver these services. It’s going to create a situation where younger and smaller companies will have a distinctive advantage in being successful here. The other issue around big companies versus small is, at least with people that we’re speaking to, there’s a lot of fear. There’s a lot of misunderstanding about these models from a large company perspective, especially when it comes to legal and compliance.
There’s this belief that all your data is going to disappear. It’s going to be used to train other people, and you’re giving away your IP. When I dig in and ask questions about, “What is it you’re concerned about?” Most people don’t even know. They read an article because the free version of ChatGPT can potentially use your data to train their enormous language model.
The likelihood of anything interesting you put in there coming out in that form is extremely low, but because there were some articles written about that. They’re so apprehensive to do anything. It’s almost like a double whammy. One, they’re scared of anything AI. It will take years to figure that out. By the time they do figure it out, they will be so far behind from a cost perspective because they don’t have the usage data that they won’t be able to compete. It’s a very exciting time to be a small company that can be nimble and can adopt this technology.
Lesson learned there. Adopt the technology, read plenty, learn all you can, and manage your fear. Know by taking action and moving forward, you’re going to be ahead of a lot of others, and that’s great. Walk us through your product a little bit.
I’ll show you a few of the pieces that we’ve got here. As I stated, we started out, whether you want to call it a data marketplace or a data search engine, but it’s all textual-based. Someone comes in, and they literally give a name and a description of the data they’re looking for. Here is an example search. Someone is looking for wine sales in the US. They want to track specifically high-end wine sold at retail, wholesale food, and beverage across the US. Ideally, they want to be able to see which winery and any information on the label.
I’m going to interject for a moment because some of the readers will not see the screen share of the video. We’re looking at one web page that shows four bullet points on where we’re going to end up. This is basic information. This is page one and literally requests name wine sales in the US. Under that, request description. A couple of sentences you type in, and you can go from there. It’s pretty straightforward. Go ahead, Brad.
You’ll see this fancy check description button. This was a very light GPT-4 feature we added, where we’ve taught it what a good request description looks like. It does a very good job of giving feedback on that. If I, for example, want to say, “I want wine data,” and you could be far more descriptive. It would still potentially complain about that.
We’re checking that box and it’s modifying.
It says, “Your search is too vague. Please type the type of data you’re looking for about wine. Are you interested in sales, data production volumes or consumer preferences? Are you looking for information on red wine, white wine or sparkling wine?” It does an incredible job of improving the quality of data we are getting. That was just not possible before these LLMs.
That’ll save you hiring about $300,000 to $400,000 a year prompter that people are hiring.
You can also hide your identity in a search. Maybe the fact that your company is looking for this particular data asset is sensitive and you don’t want a large number of people to see that. If there are data companies you already work with and know about. This is a made-up name. Let’s say you already spoke with WineUSA. You wanted to skip them. You put them in there so they won’t see this request.
This is bullet point two. We have basic information, then we have providers and now we’re on to requirements.
These are filters and are completely optional, but maybe you need a certain amount of history. For example, I need at least five years of history and I need the data updated weekly. I could put that in to specify to the provider. Lastly, we have the question section. The idea here is there may be people that loosely match this, but there’s some specific information you’d want those companies to answer so you’d know they are a good fit for what I’m looking for.
I might ask, what types of wines can you track? How far back does your data go? We have people ask very complex things. They might ask, are you HIPAA compliant? Is the data GDPR compliant? Where do you get the data? This is to save the searcher time. You put that together and submit it. It’s going to go on your dashboard of requests. This is going to list all the requests you have in process.
This is like on ChatGPT, on that left side, you might find all the different searches you’ve done. This is my request very similar, but I’ll say it is much cleaner and easier to deal with.
At this step, we use some pretty sophisticated AI. We have a computer look at the description of what you’re searching for. We then have it look through the profiles of our 2,500 data vendors, including information on what products they sell, how their data has been used in the past, and what example use cases. It picks the top roughly 5 to 10. It contacts those companies.
The AI sends those data requests to companies all around the world that loosely match that criteria. In those companies, a person logs in and looks at the request. In many cases, they might say, “I never thought of that before as a use case of our data.” Our data can do that. They’ll respond to you. You’ll get an email, log back into the platform, and you will see who got back to you. You’ll see the name of the company.
During this process, there was that other button that allowed you to be incognito, I’ll call it, and be anonymous. Would this still work here and you would be anonymous if you didn’t want to be exposed?
You could send out these requests anonymously. The data vendors would not know who you were. There’s a vague description of your business. We’re a wine merchant based in the Northeast United States. They would see the category of company you’re in. We’re a software or beverage company, but they wouldn’t know who you were. Even if you don’t hide your identity, it doesn’t say your name. It just says the name of your company. You’ll get responses and see who they came from.
The name of the person, the company and what does that company do. That’s the information you’d get in a data Yellow Pages, but the useful information is their response. This is a response where they read what you were looking for. They may have done some research and talked to their own internal data scientists to say, “Can we do this?” They’re telling you how they can do it.
I’m going to read this out loud for people. It says response. It says, “Our warehouse management solutions contract the flow of goods in and out of warehouses across the country. Using our proprietary data sets, you can easily see changes in the flow of goods. This will help you see into supply chain delays.”
This is not for the request we did, but it’s the same idea. They answer their request. They’ll answer any questions. Now, you have a short list of companies who say they can do what you need if you ask custom questions. They’ll have an even better idea and then you decide if it’s a firm you want to speak with. If it is not, we have a do not proceed button where you give a reason. Maybe data doesn’t go far back enough.
We had somebody answer that. They submit it. The vendor knows why they were rejected. There are also some pre-canned answers or the buyer says, “Yes, I’m interested in being connected with that vendor.” Once that happens, the two are put into a chat. If you choose to be anonymous, it still preserves your anonymity and it puts the two parties in a direct conversation.
It’s pretty incredible you’re going from the wide world of data all the way down to who can service your request and who you need. Ultimately, you’re conversing. There’s a chat button, a conversations button, and you’re going through all that. It’s amazing. Brad, can you tell me your clientele in general to whatever degree you’re comfortable saying who they are? That’s fine, but what size of companies is your main client these days? Is there any particular industry over another?
Data is something that serves every industry, but I’d say the early adopters for this product have been consultancies, whether they’re data analytics consultancies, management consultancies, helping people size markets and opportunities. We have a lot of investors. They’re looking at markets to make investments. They’re trying to understand the dynamics around a company. They’re trying to find companies that are behaving in a certain way.
On the corporate side, the heaviest users have been telcos, insurance companies, CPG companies, and retailers. Traditionally, data-heavy businesses. They already understand these problems. We’re starting to see a rise in technology companies needing data to either build products, maybe build a feature. You want a weather widget in your product. Maybe you need a weather data feed. You’re trying to train a visualization model on something. You need a bunch of images of that particular situation. Who has that? We work with a lot of name-brand companies to help them find the data.
When did you start working on this technology? Did this morph from other things you were working on prior? I know it’s a concept it did, but I’m talking about once you started coding.
The last thing we developed at my previous company was effectively a data list, a data yellow pages for preventers. That’s what everybody said they wanted. Even now, a lot of people who don’t know better say, “I want a big list of data.” We built a giant list of data, keyword searchable categories, and no one could find anything. Nothing is keyword-based.
Let’s say one data provider describes themselves as having data on home repair, then somebody searches for data on Home Depot. That doesn’t come up. It’s not the same word. It’s a different word. That’s the fundamental problem with a keyword search-based system. You typically don’t have the right keywords, and that’s why these LLMs are so powerful.
Seeing that problem, the first version, the struggle I had was how do you get these keywords? How do you get all this text? You would need to know that this data set does that particular thing. We didn’t touch on it, but in that flow that I showed you. A user is describing a real-world business problem. They’re giving us the language of business. We use algorithms to guess who can solve that, and at the end of the day, then we get the feedback from the data vendor.
They say, “Yes, I can do this.” Now, we have the language of business around the problem. We have their response in business language as well, and we have that confirmation where they said, “Yes, we have this.” That is the key feedback mechanism here. It’s not a static body of knowledge that we have. It’s growing every single day with every single search. We are learning more about what these data sets can do, and we use that to power a lot of other pieces of the application as well.
You’re using AI from your own business sense to analyze everything that comes in, everything gets returned back, and bettering your own system the whole way.
We use AI to make sure we’re collecting the right data in the first place. We make sure the AI does a good match, and what we learn from the match, we use AI to extract. We have even heavier features on AI. We basically have an internal chatbot where any employee at a company can talk to the thing and say, “What data do we buy on Indian grocery stores?” It looks through every data set you have and every data set you buy. It uses the knowledge that we have to answer the customer’s question. Those are much more complex AI features.
It’s mind-blowing. I love going down this rabbit hole a little bit because our readers have different reasons for reading. Some of this is potentially Nomad data to utilize your services and do all that, which we’re thrilled about. The other side of it is to get a general idea of what AI is capable of and why and how, then be able to relate it to their own situations, lives and companies. You described multiple points of that. That’s valuable information for people to take away because, let’s face it, AI is too big and moving too fast for any of us to absorb all of it. It’s just not possible.
I love that you brought all those things out. I think it’s going to help a lot of people. We’re going to head over to segment two now, which is AI wants to know. The whole concept here is AI is a product of curiosity and we’re going to stay curious here. We’re going to ask you ten questions that are designed to uncover the intriguing human mysteries that AI longs to comprehend, but can’t quite grasp. It’s a snack break in our journey, so keep the answers quick, but the safety belt, it’s off. If it feels right, we can occasionally roam about the cabin, exploring more of who you are and what makes you tick. Are you ready for it?
What’s the first thing you ever remember being proud of?
I must have been 8 or 9 years old. Back then, it wasn’t AI. Computers were the buzz. They were going from something the size of a refrigerator to something that you could fit on a large desk. I remember buying my first computer and starting to explore. Back then, it was very early Apple, like an Apple II. Maybe even earlier than that. Maybe it was a Commodore, but the default language was BASIC.
I remember buying a book to try to speak this foreign language. At the time, computer codes looked like science fiction. I remember buying a book. As a little kid, reading through it in probably two days and then being able to do something with that knowledge. It’s amazing in general. The coding, you could take what’s in your brain and make something, but as a little kid, you’re not empowered to do anything. All of a sudden, I was able to build something. Something that other people found useful and solved the problem. That was very empowering and very exciting.
I love that answer because your resumé is amazing. You’ve worked with some monster companies and both as an investor and within companies. You’ve done so much. I think you hit on it. You got rid of that barrier. There’s always a barrier that a lot of us have. In the case of the story you described, the barrier is, “I’m just a young kid and I’m not capable of, and later point in life, I will do.” You broke that barrier at a very young age. You said, “I need to get a book, apply myself,” and bang. No doubt the other kids around you had no idea that that was possible. That’s a very powerful answer.
Not to belabor the point too much, but the changes in AI feel somewhat similar. A lot of applications I would have loved to have built were just out of my reach. I didn’t have that skill set. Quite frankly, even if I did, the cost and time was prohibitive. All of a sudden, this new thing comes out. I can call an API and send it a list of instructions. If you’ve never worked with these LLM APIs, to me, they are mind-blowing.
Normally, an API is an algorithm and it has very structured inputs and output. It does something very specific. Whereas with these LLM APIs, you tell it what it does. You say, “This is the function you perform. Here’s the data you perform it on. Return me this answer.” You can get it to do anything. I feel that feeling that I felt as a little kid where all of a sudden, I’m empowered to do anything. It’s incredible.
For those newer to the industry, that’s where the term generative AI comes from because what Brad just described, when it returns you the answer, it’s a unique answer. Not every time, but it’s unique to your prompt and your question. We’ve never had that before. It’s sensational. I’m going to go on to question number two now. What do you need help with that you wish you did not?
As an entrepreneur, you are a generalist. If you came from another industry, you probably developed some skills. I spent years in finance, business analytics and business analysis. I understand that stuff as well as I can, but then there are areas like marketing. How do you make people aware of something where that wasn’t my background? That’s an area where we’re always looking for help. What are the best ways to reach people? What are the best ways to incentivize somebody that now is the time to embrace these technological changes and buy these extremely interesting products?
That’s always an area that I’m looking to develop. In my last company, I was a software developer, but I’d never built product for other people before. That was something I needed a lot of help on. Through that journey, I became fairly proficient at that. It was through getting help from folks and working with people much farther than myself in these areas that I was able to, through osmosis, become somewhat fluent.
Number three, what do others often look to you for help with?
A lot of times, people look to me for help, which is the big picture. I’m an average reader. I read anything from genetics to computer science to history. One thing I found is if you read a lot about a lot of different things, you start to see that everything is somewhat connected. Knowledge of biology matters in history sometimes. It matters in computer science.
By reading such a diverse base, I’m able to see the big picture a lot. I’m able to see how markets are going to unfold when I combine that with my experience looking at companies. A lot of people asked my opinion on those types of things because I’ve gotten a chance to meet with hundreds of companies. I understand the technology and some of the history around a lot of different events. I throw my hat in the ring and try to give advice where I can.
Time well spent. Makes a lot of sense, all that reading that you do. Sometimes, you don’t know what your takeaway is going to be from that endeavor, but it all matters. We’re going to head to number four. What do you treasure most about your human abilities?
We talked a little bit about AI being powerful because you can tell it what the algorithm is. Nothing competes with the human mind in that way. It has the ability to be completely reprogrammed from birth and even through your life to do different things. It is the ultimate generalist machine. Its ability to truly understand real generalized intelligence is something that’s incredible. That’s an area where, at least from what I’ve seen, has not been solved. There aren’t even well-established approaches to solve the problem of generalized artificial intelligence.AI is the ultimate generalist machine. It can be completely reprogrammed, it can adapt, and it has the ability to understand a wide range of knowledge. Click To Tweet
I’m going to head to number five, which is, throughout your whole life, what is the most consistent thing about you?
From anyone who’s worked with me or lived with me, the most consistent thing is my daily routine. I’m a creature of habit. I eat the same thing every single day. I wear the same thing when I sit at my home office pretty much every single day. People make fun of me. I have broccoli four meals a day every single day of the year.
You’ve brought consistency to yet another level. I appreciate that.
There’s so much to worry about. I’ve spent a little time on things that I don’t need to worry about as possible.
Doesn’t it go to Einstein in his ten blue suits or gray suits or whatever you have? All exactly the same, just thin out the decision-making. What has changed most throughout your life?
I’ve been somewhat of a generalist. I’ve been an expert in a lot of very different areas. I started out as a computer scientist. I did my undergrad at MIT, Electro-engineering and Computer Science. That’s what I did. I was building databases and writing SQL queries. The next minute, I was working at a long horizon-focused investment firm, which was a completely different job. It had very little to do with the databases and SQL and everything to do with understanding people and how to encourage them to help you understand the big picture.
That was an amazing experience, and then I switched roles within investing then I was an entrepreneur. I was an entrepreneur again in a slightly different area. My journey has been, in some ways, all over the place because they’re not jobs that one person has. Usually, most people aren’t a computer scientist, developer, and an investor. Those are two things that you don’t normally find together. At the same time, the thing guiding me has all been around adding structure. Data, at the end of the day, is structure. I try to create structure in any situation I get into, no matter how chaotic.
That’s connected to your daily routines being structured as well. It’s the way you bring things. Life is an adventure if you let it be and you’ve made it be. It’s fantastic. What do you find strangest about reality?
I don’t know if it’s strange, but it’s what perspective you look at it, but people’s decision making. I’ve found that collectively, humanity doesn’t pass on a lot of knowledge of anything that’s happened before to the next generation, especially frustrating when you read about history and you see the same mistakes. The same patterns of evil and mistakes being made over and over again. I meet with companies all the time that have an idea to start, which failed 50,000 times already.
They all fail for the same reason because nobody knows that. It’s crazy that we haven’t come up with a better solution to this knowledge transfer problem. The way school is structured, it teaches you basic skills, but it doesn’t teach you historical context for anything. People are doomed to repeat the same mistakes over and over again, even in a technology vertical.
Figuring out who made the same mistakes you made, it’s also all so random. You’re lucky enough to speak to the right person that can give you that knowledge and not that we’re solving the world’s problems. In data, that’s a problem that we went after. We saw that people, even in the same company, with the same data set. They would make the same mistake over and over again because they didn’t know anyone else had already solved it.
Part of our, as we call it, data relationship manager is about capturing that knowledge. There are so many places with humanity where knowledge is not being captured and transferred. Maybe artificial intelligence is the thing that captures it because I will say I’ll go to ChatGPT and ask it about the historical context of something. It has access to all the knowledge in the world. It can tell me anytime a certain type of event has happened. Maybe that’s the way we get over this. Maybe we’ve already solved the first big piece of it. I don’t know. It surprises me that we haven’t made much progress on that.
It’s interesting. I was going to use contradicted. I don’t know if that’s correct or not. I was going to say you contradicted yourself because, in the beginning, you made a comment about the lessons of history not moving forward. In the end, more accurately, it’s the lessons of the next generation or generations not looking to history. The information is there and certainly, it takes effort, depending on which generation you’re talking about and AI versus 50 years ago.
As to how much time it would take, but it is all there, but try and dissuade an eighteen-year-old from his great ideas because you’re double or triple that age and you’re trying to give them a lesson. They typically have to learn it themselves. It’s fascinating. Maybe there’s something here with AI to fix that. If we fix that and stop repeating all these problems that you described very well, maybe it is part of the solution.
It’s a piece of it. There are two kinds of knowledge. There’s the factual knowledge and for lack of a better term, there’s the muscle memory. If you have read about a situation, then you find yourself in that situation for the first time, you’re probably not going to be able to navigate your way out of it. You’ll have some pointers, but after you’ve gone through it, you can read a million entrepreneurial books. When you’re an entrepreneur and these things are happening, it’s a different memory that you develop, and then that, I don’t know a good way to transfer. Maybe it’s the metaverse or it’s downloading things directly to your brain.
I could tell you this. I didn’t come up with this, but I heard it and think it’s powerful. We know everything when we relied on Google over the last couple of decades, but knowing and learning are two different things. It’s the learning part that’s been missing. We could do another hour on this. I find it fascinating anyway, but we’re going to head to number eight. When most recently do you remember feeling alive?
I will say I’m an engineer by heart. Building things excites me. We released this new ChatGPT for data product. The product is exciting. I don’t want to focus purely on that, but we’re using GPT-4 in a way that I had never imagined before, which is having it write computer code and then we allow it to execute that code in a safe environment. Watching the computer understand the problem and watch it try to solve it, was just incredible. That, to me, was one of those light bulb moments. The world has changed. The way that we do so many things is about to change so dramatically. I was in awe, terrified, and excited. It was a lot of emotion.
Number nine, what is your most unique trait?
It’s probably, again, the diversity of experience, then the ability to look at things from a lot of different angles. As with most engineers, maybe I’m a little bit lower on the empathy scale than some people, so I can take a step back from the emotions of things. Running an early-stage company is a very emotional thing. People get very excited, very angry, and very worried. My ability to put a couple of different hats on and emotionally take myself out of that is somewhat unique.
It is somewhat unique and it’s also very important. This is number ten. I’m sorry I’m laughing because I find the question a little bit funny, but have fun with it. If you weren’t human, what would you be?
It could be something alive, just not human?
No restrictions on your answer.
I’d love to be an autonomous spaceship traveling the universe for all time to see everything that we don’t know, which is so much.
That’s that curious side of you. I’m going to throw in one last bonus question. I’m looking forward to asking you this specifically because, Brad, you’ve shown us you’re a deep thinker and in action. Those things combined have built Nomad. I’m very impressed with you. The thing that’s in the news a lot and on people’s minds is AI in our future. What are the dangers of AI moving forward that we need to be cautious about? Maybe even if you delve into the steps we should take to be on the path of enacting some caution.
AI is such a broad space. There are a lot of ways to approach this problem. A lot of people are worried about this idea of artificial general intelligence where a computer is going to somehow take over the world. Could that happen? Sure. From what I’ve seen in AI, we are nowhere near an artificial general intelligence. That’s not a today problem in my mind, but it’s probably not a tomorrow problem. It’s way down the road problem, but it’s one we should start thinking about. The thing I worry about more is the evil actors of the world using this in ways that are going to be challenging.We are nowhere near an artificial general intelligence. That's not a today problem. It's probably not a tomorrow problem. It's a way down the road problem, but it's one we should start thinking about. Click To Tweet
When I was at MIT, I took a class. It was basically a hacking class. It was an information security class at the end of the day, but a lot of it was hacking into things to know how to prevent people from hacking into things. Social engineering still remains one of the easiest ways to get at things. You call up a person at a company. You pretend to be somebody else. You send them an email. The problem is a lot of the attackers didn’t have the sophistication and the domain they were trying to hack into.
They couldn’t articulate it like someone who worked in a company in a certain voice or accent. All of that is going to be fakeable. That, to me, is terrifying. Protecting any information security system is going to be beyond complicated. Protecting my own parents from having their credit card number stolen once a month is a big challenge, and then you think about all the fake pictures we can make. All the fake newsreels we’ll be able to make. It’s going to be hard to discern what on earth is happening and who to trust. It’s going to be a messy world.
Brad, I’m going to direct you to the Edge of AI and look at some of the prior episodes we’ve done. I found it incredibly fascinating because we have had companies that have specialized in, for instance, voice AI recognition. It’s all about deep fakes and preventing from and being able to put a rating system in place so you know the chances of it being fake because I’ve had my voice deep faked and I can’t tell the difference. These are AI systems. I’ll call them the white hats. They can show you what’s real and what’s not.
We’ve also had another company that does the videos and still images as well. They’ll give you that percentage of probability that they’re a fake. They’ll also highlight where on that image or that video it looks like it is a fake. These threats that are in some part there because of AI, hopefully, will also help be solved by AI. It’s a fascinating world we’ve got now.
The ultimate cat and mouse game.
Let’s go to the resources a little bit. As you said, you read a lot and listen to a lot of things, trying to stay up on the industry. Any hints you’ve got for the audience, any show they might go to, books they might want to read or any resource whatsoever that’s helped you that might help them?
One book, as lack of a better word, revolution occurred over the past year in AI. I wanted to understand where we were and have a little bit more historical context. I read a book called The Myth of Artificial Intelligence by Erik Larson, which I found fascinating. It helped me understand a little bit better what we’re working on, which problems are out of our reach, and why they’re out of our reach. I would highly recommend that.
The thing with AI now, it is moving so fast. We’re coming up with techniques internally here that don’t even have names. Two weeks later, somebody writes a paper and they give it a name. Honestly, the best resource I’ve seen is LinkedIn. It’s being connected to as many folks as possible. There are a lot of individuals putting out papers on this stuff, which to me is the most useful.
It’s fantastic. We do this full time and it’s hard to keep up. The fact that you’re here, I consider it an honor to have you here, Brad. The fact you’re here means we siphon through a lot of options. We’re looking for the most relevant what you have developed and what you’re putting out there. Now, after getting to know you, you’re going to be staying ahead of the curve. You are one of those resources for us now and, hopefully, many of our readers as well.
Follow me on LinkedIn. I try to put out useful content when I can.
Anything we didn’t cover in cool ways to use AI, even outside of your company and what’s going on, in general or personal use?
One of the most amazing use cases I’ve seen, which we’ve built on and others, is the ability to go through large corpuses of documents, whether they’re your own apartment rental agreements, bank loan documents or vendor agreements. You can have an AI. You can ask a question, have it go through a thousand documents, and find exactly what you’re looking for. That is so powerful in so many different industries. That’s one I’m very excited about. We built an enterprise tool around this, and I am shocked that every time we use it, it is possible.
We’re going to head toward closing. You mentioned LinkedIn. Give us an idea of the best places to track what you’re up to and what you’re doing.
I’m just Brad Schneider on LinkedIn. That’s the best place to connect with me. I like a lot of content from people I believe to be AI thought leaders. Connect with me and you’ll be able to see some of those folks and connect with them directly. That’s where I see all of these papers being posted. I don’t know a better place than that.
Brad, I can’t thank you enough for taking the time out to do this for us and all of our readers. For me, it’s personally been fantastic, but now, it’s time for another safe landing at the Outer Edges of the AI universe. On behalf of our guests and the entire crew, I’d like to thank you for choosing to voyage with us. We wish you a safe and enjoyable continuation of your journey.
When you come back aboard, make sure to bring a friend. Our starship is always ready for more adventures. Head over to Spotify or iTunes now, rate us and share your thoughts. Your support and feedback means the world to us. Don’t forget to visit EdgeOfAI.xyz to learn more. Connect with us on all the major social platforms by searching for EdgeOf_AI. Join the exciting conversations that are happening online. Before we sign off, mark your calendars for our next voyage where we’ll continue to unravel the mysteries and advancements of AI. Until then.