Episode #56: Three Breakthroughs: OpenAI’s Latest Release

Frequently Asked Questions

FAQ

• Where is the transcript for this webinar?

  Sam Herrick:
  Using AI to fight future pandemics, ChatGPT’s newest release, 6G and beyond. Let’s go.

  Michael Peri:
  The fact that we have the ability to think about these things differently even a few years later will fundamentally change how we think about what would be a perceived limitation.

  Mike Collins:
  Who basically said, “Okay, this thing, I spent an hour and it did what it took me a year to do.” And so I think it’s just adding to the mix. They forget what 3G was like.

  Michael Peri:
  This is global connectivity across all industries.

  Sam Herrick:
  Hello, everyone. Welcome back to another really awesome episode of The Tech Optimist. We have a Three Breakthroughs episode for you today. And of course, these guys need no introduction, but they’re going to get one anyway.

  We have the legendary Mike Collins, founder and CEO at Alumni Ventures, on one side of the table today. And then Mike’s savant in today’s episode is going to be Michael Peri, a partner within our walls here at Alumni Ventures. And then my name is Sam. I’m the guide, editor, and sort of narrative writer for this show.

  So I’m going to give you a quick sneak peek into the topics that are talked about today. Topic number one, Mike Peri brings up AI and how it can be utilized to fight future pandemics, epidemics, and just global crises. There’s a really awesome article that we dissect that talks about how AI might be able to analyze social media and, therefore, how it can process other scientific papers that help scientists, doctors, and government officials get further along in their process of helping the rest of the population during such crises.

  Then we talk about ChatGPT’s newest release, Strawberry, and how insane it is—how innovative and groundbreaking it is—and the thought process behind Strawberry and how it works is just absolutely bonkers. We have a video that comes along with it that helps describe how it works and how incredible this new technology is.

  And then we round off the show with the concept of 6G. We all know about 5G on our Verizon phones and LTE connectivity and all that fun stuff, but we try to dissect what it would mean if we had 6G connectivity across the world. So let’s dive in.

  Disclosure:
  As a reminder, The Tech Optimist podcast is for informational purposes only. It is not personalized advice, and it is not an offer to buy or sell securities. For additional important details, please see the text description accompanying this episode.

  Mike Collins:
  Hello, everybody. Welcome to Alumni Ventures and our Three Breakthroughs podcast. I’m joined by my partner, Mike Peri. I’m Mike Collins. I’m the founder and CEO. And we get together every week and talk about what’s going on in the world of technology—what are the three big breakthroughs that you should know about, people are talking about, and people should be talking about. Welcome to the show, Mike. How are you doing?

  Michael Peri:
  I’m doing well. I’m glad to be on.

  Mike Collins:
  Great. You’re kicking it off this week. What do you get? What struck your fancy?

  Michael Peri:
  All right, I have one that’s pretty timely. Mike, you know this—I’m a new parent. So when September-October rolls around, it’s like a flu shot, COVID shot, every shot. The kids are always sick. And it just had me thinking about how far we’ve come in terms of COVID.

  I saw a really interesting article that’s basically saying, how can AI help fight the next pandemic? I think the reasons in which you think AI could be helpful are both prominent and easy to understand, but also at the same time pretty nuanced.

  A lot of researchers, epidemiologists, and academics warn that the next pandemic, which they call Disease X, is likely to be here in the next decade. And as they’re thinking about this, what’s new today versus even four years ago with COVID is how AI is being used to really understand and prepare us for it.

  The research and ability for them to use AI is really cool because it’s breaking down a lot of complexities as you think about global health. There are aspects of predicting the outbreak that they’re getting much more precision on. There’s the speeding up of vaccine development, which I think we all thought during COVID was moving mountains. I think that’ll only be…

  Mike Collins:
  They called it warp drive, I think. But it was fast. It was remarkably fast.

  Michael Peri:
  They’ve talked about this specifically where it’s like AstraZeneca, for example, are using AI models now to identify and screen antibodies for different mutations. They’re running AI upfront on the mutations and then behind the scenes on the antibodies. And they’re saying that this is cutting down time for vaccine development to literally just days.

  So you have the hard science. And then what I also liked about how the article presents it is there’s also a view of cutting through the noise of public sentiment, unusual patterns, obviously fake news, and people just saying everything.

  To me, it was just this notion of like, wow, it will happen. It’s not a matter of if, it’s when. And I think the tools that we have in our toolkit are going to be fundamentally different in all aspects of how we would address a pandemic moving forward.

  Michael Peri:
  Hopefully we don’t have to see it in our lifetime, but if we do, I like that AI is now being used front and center—the full value chain of how we think about this stuff.

  Mike Collins:
  Yeah, a couple of thoughts. One, I am just struck every day by how short people’s memories are. Where we were in the spring just four or five years ago—how quick it happened, how disruptive it was, how scary it was—and now it’s just like, oh, this, that, or the other thing, and millions of people died. Arguably, it could have been 10, 20 times worse with just a couple of little differences.

  And so that’s one point. Two is, I’m very excited about the modeling of these things. We’re moving from human, physical models to digital models with drugs, drug discovery, drug testing—all the things where we need trials. It’s like the digital wind tunnel to test aircrafts. So much of healthcare is moving into this digital realm, which is just way, way better.

  But I do think we live in a global world. People intermingle, people travel, and we are still captured in a very fragile biological body. These bugs don’t give a shit. Good. Tech Optimist, I am glad people are working on this problem.

  Michael Peri:
  Yeah, yeah. I mean, they did a nice job at the end where it’s like, listen, this isn’t a cure-all, but the fact that we have the ability to think about these things differently even a few years later will fundamentally change how we think about what would be a perceived limitation and breaking through those barriers moving forward.

  Mike Collins:
  And let’s just give a nod to the people that get up every day to work on these problems while others are worried about more trivial things. Tip your cap to these scientists.

  Michael Peri:
  We talk about patient capital as investors, right? I mean, this is patient innovation. These are the people waking up every day that have a calling for a global good. But it’s going to be really fascinating to see just how this develops moving forward and certainly as AI continues to advance. I think really and truly, everyone talks about, “Oh, here’s this big issue.” It’s also going to be a huge key factor in safeguarding humanity for the next pandemic, let alone global catastrophe.

  Mike Collins:
  So let’s be sure the good guys have the best.

  Michael Peri:
  Yes, exactly.

  Sam Herrick:
  Okay, you guys know the drill. Now it’s time. Before I get into my little spiel about this article that Mike Peri found, we’re going to hop into an ad. And then after my tech note and everything, we’re going to get back into the show where Mike and Mike start to talk about ChatGPT’s release of Strawberry and what that means for the AI world. So we’ve got some exciting stuff coming up for you. Don’t go anywhere. We’ll be right back.

  Ludwig Schulze:
  Hi. Just a brief interruption to introduce you to the Healthtech Fund from Alumni Ventures. Alumni Ventures is one of the most active and best-performing VCs in the US, and we have raised over a billion dollars from more than 10,000 individual investors.

  With our Healthtech Fund, you’ll have the opportunity to invest in a portfolio of around 20 healthtech startups, from transformative healthcare services to groundbreaking diagnostics. Our founders are paving the way for a healthier future. To learn more, visit us at av.vc/funds/healthtech.

  Sam Herrick:
  Okay, we are back. So now, this article that Mike Peri brought up was from the BBC. It was written by Suzanne Bearne, a technology reporter, and it was released last week. This article does a really good job of exemplifying different ways that medical professionals and government officials can use AI to help prevent or tackle the next global pandemic.

  There’s a section of this article that I want to read. So the question is: How could AI help to alleviate the frightening prospect of another global pandemic? Researchers in California are developing an AI-based early warning system that will examine social media posts to help predict future pandemics. The researchers from UCLA and the University of California, Irvine, are part of a US National Science Foundation’s Predictive Intelligence for Pandemic Prevention grant program. Quite a mouthful, but they’re doing some awesome work.

  This funds research that aims to identify, model, predict, track, and mitigate the effects of future pandemics. The project builds on earlier work from UCI and UCLA researchers, including a searchable database of 2.3 billion US Twitter posts collected since 2015 to monitor public health trends.

  The tool works by figuring out which tweets are meaningful and training the algorithm to detect early signs of a future pandemic, predicting upcoming outbreaks, and evaluating the potential outcomes of specific public health policies.

  “We have developed a machine-learning model for identifying and categorizing significant events that may be indicative of an upcoming epidemic from social media streams.” So that’s one pretty fascinating way of looking at this.

  And then this is where AstraZeneca comes in. AstraZeneca’s Vice President of Data Science and AI R&D, Jim Wetherall, brings up a point that the pharmaceutical giant uses AI to help speed up the discovery of new antibodies.

  Antibodies are proteins used by the body’s immune system to fight off harmful substances that can be used to make new vaccines. Mr. Wetherall says the firm can generate and screen a library of antibodies and bring the highest-quality predictions to the lab, reducing the number of antibodies that need to be tested and cutting the time to identify target antibody leads from three months to three days.

  So it’s fascinating—just the two polar-opposite sides where AI can be used to help such a vast issue that could plague the planet in the future. One tackles social media and what society is saying or sharing, and the other is on the science side, where AI can reproduce hundreds of man-hours in, as you can see here, just three days.

  All this stuff is very, very exciting. But now let’s hop into the next subject, which is Mike’s topic—Strawberry from ChatGPT. So let’s get into it.

  Mike Collins:
  Okay, so on a related note, my breakthrough of the week was the release, ChatGPT o1. Again, talk about a company that can’t name things to save its life. Use your own product to come up with some better names, please. This is rumored and affectionately called Strawberry because ChatGPT-4o couldn’t count the number of Rs in Strawberry.

  So they released a model built differently, whose strength is in harder reasoning problems and some really clever examples. Again, for me, I’ve played with it a little bit. It seems to be very different but really good at some things that the other models aren’t, and vice versa.

  One of my takeaways is I think we’re going to be in a world where there is not one model—one God model—for a while, but various models that do different things really well. And I think over the past six to 12 months, people that have gotten invested in Gemini or Claude or ChatGPT-4o have learned, “It can do this, it can’t do that.” Here’s another one that is very good.

  I used it just personally to outline a tech spec that required pretty significant reasoning. I also gave it a business problem, and it was thinking. And again, some of this is a good design interface, I think, by the OpenAI guys—where the time it’s taking, the words that are showing up on the screen—it’s like, oh, this thing’s thinking really hard about my problem. The answers are way more involved, showing their work, showing the logic, showing you their homework.

  So again, not a lot of time—just released maybe a week ago—but I think it is a big deal. I think there are people that are working on really hard problems. I saw one posting from some PhD student who basically said, “Okay, this thing, I spent an hour and it did what it took me a year to do.” And so I think it’s just adding to the mix.

  And then again, the third point—I’ll underscore again, Mike—is just the speed at which these things are coming out. Every quarter, there are new capabilities.

   

  Michael Peri:
  It’s exponential.

  Mike Collins:
  It’s exponential. We are on the Ray Kurzweil—we are on the steep part of the geometric curve. People don’t see it, people don’t get it, but we are on it. And hang on.

  Michael Peri:
  I’ll give you some examples too, and maybe for the audience out there, how you think about it. We could all read, right? They put it through International Math Olympiad-type questions. It’s performing eight times better than their current model. It’s really excelling in math and coding and whatnot. And I was tweaking it a little bit in the perspective of what we do day-to-day.

  I was positioning questions as a venture investor more through the lens of the scientific method, where you take what a company was trying to do and I reframed it a bit as a hypothesis—like what would have to work, what wouldn’t? And I saw that this model was just… It responded like I was talking to someone versus just regurgitating something. And you could really see that the model was—I don’t know the actual terminology here—but it was thinking.

  It was really thinking about it from a hypothesis standpoint—how you would accept or reject it through the lens of an investment. And then it got me thinking about this notion we see in all these companies, which is an AI co-pilot or agent for a specific task. I think, very subtly and in an elegant way, OpenAI is almost positioning these models this way.

  If you work in a creative field and this is what your content needs to be at the end of the day, use model 4o—that’s actually going to be great for you. Now, vice versa, if you’re doing something much more technical and/or reasoning from a math perspective, you’re going to be living here. So I’m starting to see and maybe connect the dots on how they’re building this as a true enterprise function that meets the needs for everyone.

  And then of course, at the end of the day, all of these models are reinforcing each other. I mean, I consistently have been blown away by what they’ve been able to do, but this had a different rigor coming out of it. It was like this is speaking as though Mike and I would be about an investment.

  Mike Collins:
  Yeah, no, just a couple things. One is, I’ve seen… They put these models through, I think, the Finnish IQ test. And again, a lot of them have been at the 80 level. And listen, an army of 80 IQ helpers can be incredibly valuable. This is like 120. So getting smarter on just that dimension I think is noteworthy.

  My personal heuristic right now—and again, I’m sure in 30, 60, 90 days this will be totally different—but today, September 18, 2024, I would say I have my own personal agent where I’ve trained it on the way I talk, my job, our company, how we think, our culture—private. That is my personal AI on the shoulder that I use as a generalist assistant, chief of staff.

  Then I have a series of task-specific ChatGPTs where it’s like, I’m creating a model, a template. I do this repeatedly, and I just want you to save me time and increase the quality. And I’ve got an army of those—small army, but an army.

  And then I think over here, I’ve got an arsenal of tools—I’m going to use Google, Gemini, Perplexity, Claude, I’m going to use ChatGPT-4o, o1—and I’m just going to apply them where I have what I’m going to call… I want to learn something. I want it to tell me how to do something. I want it to help me solve a problem or figure something out or scope something. I’m going to try to match what I need and what they’re good at.

  So right now, it’s not just “turn on my God agent and it does everything for me.” I don’t know—I think we’re a bit away from that. Right now, it’s still, I need to orchestrate it all.

  Michael Peri:
  And I will tell you, from the perspective of the economy, any job—if you’re not using this stuff today, I don’t know what you’re doing, really and truly. If you’re not a leader of an organization saying, “Every person here should be leveraging these tools to find efficiency.” We’ve engaged with big corporations our portfolio companies have, and it’s like red tape all over the place—regulatory, don’t do this, don’t do that. You are going to fall so far behind so quickly as an organization…

  Mike Collins:
  You’re going to get killed—your organization, your team, you personally and professionally. You’ve got to lead or get run over. I’ll give you another example. I had an idea. I thought, wouldn’t it be neat if I could type in a job and a year, and what would come back is the likelihood that that job would be dead, materially changed, or booming, and why.

  And while I ate lunch—with ChatGPT-4o and a little bit of coding software—I was able to knock that out in 15 minutes. And I’m not a programmer. This was all just natural language, back-and-forth, doing it while interacting with it. It’s very exciting. It’s a little mind-bending, but I do think, at the end of the day, it’s very exciting.

  To where you started out—which is scary—you can be a Luddite, or you can embrace it and just say, “Hey, let’s lean in and make the future better.”

  Michael Peri:
  Yeah, it’s going to propel humanity in so many ways. And of course, you have to always take a commitment to safety and recognize what that means.

  Mike Collins:
  But it can help you do that, right? I mean, that’s the whole point.

  Michael Peri:
  Yeah, that is the whole point. I mean, you articulate it well. Sometimes it’s mind-bending because you could visualize something in your head, you could attempt to articulate it with a friend or a colleague, but you don’t know where to start. All of a sudden, at your fingertips, you don’t just have a place to start—you have something that can totally execute on what the vision is for you.

  Mike Collins:
  You have a tutor—and a tutor who is going to meet you where you’re at. If you don’t understand, it’ll tell you, “Okay, give it to me at a lower reading level, break it down into more steps for me,” and you’re not embarrassed by it. So again, I also believe this has profound impacts on education and learning.

  Michael Peri:
  Yeah. Oh yeah, definitely.

  Sam Herrick:
  Okay, y’all, I have a little interjection here regarding this. So I was just doing a quick search on the changes of ChatGPT-o1 and Strawberry and the Mini—all the stuff that has come out recently from OpenAI. And I’m a huge fan of YouTube and visual explanations of things.

  I came across this really awesome YouTube video by Matthew Berman. He’s this AI technology extraordinaire on YouTube. He’s got a really cool channel with almost 350,000 subscribers. So he’s got quite a following, and he put out a video on ChatGPT’s newest release. He explains really well how it works, what it does, why it’s super important to the world of AI, and why this is an awesome innovation.

  The whole video is 20 minutes long. I’m not going to play the whole thing, but I’m going to cut it up here and there to give you the highlights from it. I think the information it provides is really powerful, and it’s fascinating to understand how it actually works. Whereas we’ve heard from Mike and Mike about their real-world implications and how they use it and how important it is for the future, I want to provide more information on how it actually works.

  There’s some really awesome terminology that Matthew uses here in his video.

  Sam Herrick:
  But yeah, I’m going to cut it up, get you the highlights, and just dive in from there. After that, we’re going to do our last ad, and then the last topic for today is going to be about the concept of 6G. You know 5G and LTE wireless capabilities. But have you ever heard of 6G and what’s after that? Okay, if you haven’t, stick around.

  Matthew Berman:
  So the new model is called o1, and it’s a series of models. We have two models available today: o1-preview and o1-mini. We developed a new series of AI models designed to spend more time thinking before they respond. They can reason through complex tasks and solve harder problems than previous models in science, coding, and math. We trained these models to spend more time thinking through problems before they respond, much like a person would.

  Through training, they learned to refine their thinking process, try different strategies, and recognize their mistakes. In our tests, the next model update performed similarly to PhD students on challenging benchmark tasks in physics, chemistry, and biology. We also found that it excels in math and coding in a qualifying exam.

  In the International Mathematics Olympiad, GPT-4o correctly solved only 13% of problems, while the reasoning model scored 83%. That is a massive, multiple-times improvement over GPT-4o in math. Their coding abilities were evaluated in contests and reached the 89th percentile in Codeforces competitions. You can read more about this in our technical research post. I’ll get to that in a moment.

  As an early model, it doesn’t yet have many of the features that make ChatGPT useful, like browsing the web for information and uploading files and images. For many common cases, GPT-4o will be more capable in the near term, and I suspect it’s actually going to be a lot cheaper and more appropriate for the vast amount of use cases for a while. For 99% of use cases, we just don’t need PhD-level reasoning. But for complex reasoning tasks, this is a significant advancement and represents a new level of AI capability.

  Given this, we are resetting the counter back to one and naming this series OpenAI o1. They’re also releasing o1-mini. And by the name mini, you know it’s going to be a smaller, faster, and much cheaper model. So let’s read about it.

  The o1 series excels at accurately generating and debugging complex code. This is going to accelerate the transition to AI writing the majority and then eventually all of our code. To offer a more efficient solution for developers, we’re also releasing OpenAI o1-mini, a faster, cheaper reasoning model that is particularly effective at coding. As a smaller model, o1-mini is 80% cheaper than o1-preview.

  All right, so that is fantastic. And by the way, you know I’m going to make test videos for all of these models. So what’s next? This is an early preview of these reasoning models in ChatGPT and the API. In addition to model updates, we expect to add browsing, file and image uploading, and other features to make them more useful to everyone. So it seems like this is the future of OpenAI’s model family.

  And of course, they released a demo video of o1 creating the game Snake. Let’s take a look. The prompt is: implement Snake with HTML, JS, and CSS. The entire code should be written in a single HTML block with embedded JS and CSS. Don’t use any remote assets. Then it continues. So let’s see how it does. It’s thinking, thinking, thinking, and then boom—it outputs it. And it seems lightning-fast.

  Now, granted, GPT-4o and even GPT-4 were able to do this. So it’s not that great of a demo, but it’s still fun because that was the test that I used to test every previous model before I switched to Tetris. And you know I need to test Tetris with this new model. And there’s the Snake game. There you go. It’s good. It’s fine.

  All right, let’s read a little bit of the technical paper. Again, don’t get too excited because they’re not really releasing too many details about how it works. o1 ranks in the 89th percentile on competitive programming questions in Codeforces. It places among the top 500 students in the US in a qualifier for the USA Math Olympiad and exceeds human PhD-level accuracy on a benchmark of physics, biology, and chemistry problems.

  Wow! Just imagine spinning up hundreds, thousands, millions of these to run 24 hours a day to discover new science. This really— I know I said it before—but it feels like the intelligence explosion. Our large-scale reinforcement learning algorithm teaches the model how to think productively using its chain of thought in a highly data-efficient training process.

  It seems a chain of thought is built in to run at inference time. That is why it takes longer. That is the thinking it does. We found that the performance of o1 consistently improves with reinforcement learning (train-time compute) and with more time spent thinking (test-time compute). The constraints on scaling this approach differ substantially from those of LLM pre-training, and we are continuing to investigate them.

  So here we can see this is train-time computation, and the accuracy just increases with the more train-time they put towards it. And then, of course, test-time—same thing. The more time they put towards it, the better accuracy it gets.

  One last test for this video—but you know I’m going to be testing it in full, so stay tuned for that. The prompt is: How many words are in your response to this prompt? I wonder how it’s going to do because it has a bunch of chains of thought in the background. Is it going to include all of that in the answer? We’ll see.

  Addressing paradoxical query, figuring out word count, identifying word patterns… Look at that: “There are seven words in this sentence.” One, two, three, four, five, six, seven. Wow! This is actually the first time where I think it got it right. Officially got it right. This is crazy.

  Addressing paradoxical query… I’m working through a paradoxical question, which involves self-reference and determining the response length. Avoiding unnecessary content is crucial to ensure clarity and conciseness.

  There it is. Unbelievable! I’m so excited to test this model out more. If you’re not already subscribed, make sure you subscribe because there are going to be a lot more videos about this new set of models from OpenAI. This might be the beginning of the intelligence explosion. If you enjoyed this video, please consider giving a like and subscribe, and I’ll see you in the next one.

  Ad:
  Do you have a venture capital portfolio of cutting-edge startups? Without one, you could be missing out on enormous value creation and a more diversified personal portfolio. Alumni Ventures, ranked a top 20 VC firm by CB Insights, is the leading VC firm for individual investors. Believe in investing in innovation? Visit av.vc/foundation to get started.

  Mike Collins:
  Okay, so what’s your last one?

  Michael Peri:
  Okay, all in the next wave of technology. This is a little funny because I was reading and it was like, what comes after 5G? And I was scratching my head like, are we at 5G yet? Where are we? But the Pentagon is literally already prepping for 6G. For those out there, this is going to be the next frontier of wireless communication technology. Of course, through the lens of the Pentagon, obviously it’s going to transform military operations, global defense strategy, and communication.

  But really what this is, in essence—and I think this is a good part, defining what this stuff means—at the end of the day, 6G is going to yield promising faster speeds, lower latency, and much better and more enhanced connectivity. This can be applied to so many different use cases that I think we tend to struggle with because it’s very Pentagon-driven, US military applications.

  But if we talk about any of these use cases we’ve discussed over the last six weeks—improved healthcare: now you have much better vital parameters through wearables. You have high-speed data transfer that would allow remote surgeries that we’ve talked about, and real-time triggers like, “Hey, this is going to happen,” enabling earlier interventions.

  These use cases pull through to every part of the economy and every industry—from realizing what virtual and augmented reality is going to be to an Internet of Things operating system that actually works how you think it should. This stuff, to me, is often passed on by most individuals. You think, “Oh, I have a cell phone. Great. It’s much better connectivity.”

  Mike Collins:
  They forget what 3G was like.

  Michael Peri:
  This is global connectivity across all industries. And to see that they’re already pushing, “Hey, try to get this out in the late 2020s, early 2030s.” In the same way we’re seeing AI unlock a ton, this type of connectivity is also going to unlock and just reinforce each other. So I think it’s very timely you’re starting to see some other pushes for innovation broadly that would otherwise be perceived as not as innovative.

  Mike Collins:
  I mean, I think this global coverage—I think satellite coverage, which obviously won’t be at 6G levels—but I think we’re going to have broader coverage, we’re going to have smarter coverage, we’re going to have lower latency, higher bandwidth, these enabling things that are just not as sexy and that we very easily, all of us, take for granted. But this is very much a part of infrastructure in a society every bit as much as water and roads and things like that.

  Sam Herrick:
  All right, I’m going to pump the brakes a little bit. So I looked into how 6G works and what it is, and I found some really interesting factoids that I want to provide before we round off this episode.

  So 6G, the sixth generation of wireless technology, is expected to revolutionize communications and connectivity in ways that go far beyond current 5G capabilities. While still in the research and development phase, 6G is anticipated to launch around 2030, just like Mike was saying.

  Here’s an overview of how 6G is expected to work and what it can do.

  How it works:
  It uses advanced spectrum usage, AI-native networks, advanced antenna designs, and edge computing integration.

  So now, if you’re asking what the heck does any of that mean, I’m going to tell you. 6G will utilize higher frequency bands including sub-terahertz frequencies up to at least 100 gigahertz. This expands the use of radio spectrum and will enable much faster data rates and increased network capacity.

  6G networks will incorporate artificial intelligence and machine learning at their core. These AI-based air interfaces will adapt in real time to optimize performance and energy efficiency.

  Antenna designs:
  Innovative multi-antenna technologies like massive MIMOs (multiple-input, multiple-output systems) will be further refined. These antennas will enable simultaneous transmission and reception on the same frequency, improving spectral efficiency.

  Edge computing integration:
  Mobile edge computing will be built into all 6G networks, unlike current 5G where it’s an add-on. This integration will provide improved access to AI capabilities and support for sophisticated mobile devices and systems, which are coming out more and more every day and every year.

  What 6G could do:
  It can provide ultra-fast speeds and low latency, just like Mike was saying. It can enhance IoT and M2M communication. 6G aims to support up to 10 million connected devices per square kilometer, greatly expanding machine-to-machine connectivity.

  Advanced sensing and imaging:
  This is probably where the Pentagon comes in. 6G’s higher frequencies will enable more precise wireless sensing technology. This could lead to advancements in threat detection, health monitoring, and air quality measurements.

  Other capabilities:
  Seamless integration with physical and digital worlds, space communications, autonomous systems, and smart infrastructure. Looking into the future, 6G could enable more sophisticated smart city applications, improving urban management and sustainability.

  In essence, 6G aims to create a seamless intelligent network that blends the boundaries between the digital and physical worlds, opening up possibilities for applications and services that are difficult to fully envision and implement today with the software and infrastructure we currently have.

  Michael Peri:
  Yeah, it’s well said, right? I mean, it’s almost the infrastructure that’s somewhat forgotten, but as you think about this—I mean, sure, in the military context, obviously faster, more reliable connectivity across soldiers, satellites, drones, vehicles, etc. That totally makes sense.

  But as you think about what it can do in playing a critical role for AI-driven operations or enabling almost near-instant data sharing and decision-making—talking about that exponential growth curve—you see that type of connectivity start to happen. Endless possibilities in what we could start to do in real time with AI.

  Mike Collins:
  And as venture capitalists, it’s just a reminder too of how well you can do in infrastructure—tech infrastructure. Everybody’s always focused on the sexy, high-level names, but there has been enormous value creation in tech infrastructure—putting together the picks, shovels, roads, byways, highways of these new frontiers. You can do really well.

  Michael Peri:
  We’ve built a strong portfolio across AV and a lot of these picks-and-shovels-type plays—a lot of big infrastructure-type companies that are going to enable a whole lot. So this is one I always tend to keep an eye on. Like I said, it might be funny for some people to think, “Oh yeah, like my Verizon phone.” But no, I would encourage folks to read up a bit more on it and really keep tabs on it because it unlocks a whole lot of future economies in the industry.

  Mike Collins:
  Yeah, for sure. Well, Mike, you’ve done your tour of duty. Congratulations.

  Michael Peri:
  Yeah, it’s fun.

  Mike Collins:
  We’re moving on to another AV partner next week, and you’ve been great. It’s been a lot of fun. I’ve enjoyed talking to you.

  Michael Peri:
  Yeah, likewise.

  Mike Collins:
  We’ll put you back on the circuit. So I’m looking forward to talking again down the road, and who knows what kind of crazy future we’ll be talking about at that point.

  Michael Peri:
  Yeah, exactly. Thanks for having me, Mike. It was a blast. I look forward to the next tour of duty.

  Mike Collins:
  Great. Excellent. Have a good one, Mike.

  Michael Peri:
  All right, bye.

  Sam Herrick:
  Thanks again for tuning into The Tech Optimist. If you enjoyed this episode, we’d really appreciate it if you’d give us a rating on whichever podcast app you’re using, and remember to subscribe to keep up with each episode. The Tech Optimist welcomes any questions, comments, or segment suggestions. So please email us at [email protected] with any of those, and be sure to visit our website at av.vc. As always, keep building.