EP8: Waymo’s Expansion, Meta’s NotebookLLama, OpenAI's Chip Plans & Tesla’s Autopilot Race

In this lively episode of "AI with Friends," hosts Marlon Avery, Sekou Doumbouya, and Adrian Green dive into a range of tech topics, from the challenges of autonomous driving to the latest in AI advancements. The trio discusses the expansion of Waymo into new markets, the potential of Tesla's autonomous vehicles, and the intriguing future of AI-driven devices. They also explore OpenAI's venture into chip-making and Meta's new AI tool, Notebook Llama, which competes with Google's Notebook LLM. With humor and insight, the hosts ponder the implications of these technologies on everyday life, including the rise of robotics and the potential for new business opportunities. Tune in for a thought-provoking conversation that blends tech insights with a touch of humor.
----------------------------------------------------
Welcome to AI With Friends, your weekly launchpad into the world of Artificial Intelligence. Hosted by Marlon Avery, a pioneer in GenAI innovation, alongside Adrian Green, VP of Engineering at LiveNation, and Sekou Doumbouya, Senior Staff Cloud Systems Engineer at Pinterest, this show is your go-to source for all things AI.

Our hosts bring diverse expertise—from AI strategy and tech innovation to industry leadership. Every week, they break down the latest AI trends, interview top experts, and simplify complex concepts for AI enthusiasts, entrepreneurs, and tech professionals alike.

Marlon, Adrian, and Sekou combine their unique perspectives, whether it’s Marlon’s collaborations with tech giants, Adrian’s leadership in global entertainment engineering, or Sekou’s cloud systems expertise. Together, they make AI insights accessible, actionable, and exciting.

Tune in live on Twitch & YouTube every Wednesday at 9:00 PM ET, or catch us on all major podcast platforms.

Here’s what you’ll get:

• Cutting-edge insights from AI leaders
• Real-world applications of AI technology
• A vibrant community of forward-thinkers

If you're ready to stay ahead of AI trends or spark your next big idea, join us each week for an hour of engaging, thought-provoking content.

Subscribe now and become part of the future of AI with AI With Friends!

Chapters

• 0:00 Intro
• 5:40 Computer Vision Tools
• 11:08 Claude's Unique Responses
• 15:34 Tesla Ownership and Autonomous Experience
• 20:27 Waymo's Market Expansion
• 23:46 Waymo vs. Tesla in Robo Taxi Market
• 42:45 XAI Colossus
• 46:38 Energy Consumption and SMR
• 56:26 AI Adoption and Consumer Accessibility
• 1:04:41 Meta's Notebook Llama
• 1:18:18 OpenAI's AI Chip Development
• 1:31:49 AI Developments and Merchandise Plans
• 1:34:51 Conversation Ending

Transcript

Marlon Avery: Yo, yo, yo. What's going on? What's going on, fellas? 

Sekou Doumbouya: What's up? What's up? 

Adrian Green: What's up? Happy Halloween. 

Marlon Avery: Oh, yeah. It is Halloween. 

Sekou Doumbouya: That's. 

Marlon Avery: Wow. 

Adrian Green: Yeah. 

Marlon Avery: Yeah, man, it feels like a. It feels like a. A trick over here. We're. We're having a morning session and stuff. This is a first time, you know, for the pod. We was having some extreme difficult technical issues and stuff last night, and so we decided to develop. Deliver morning podcasts and stuff here, man. So it's been. It's been good so far. Adrian, man, how was your week? 

Adrian Green: Not bad, not bad. Doing a lot of work in. On the AI side. I made a country song last night from a joke in our group chat that I got with friends of mine and was able to actually cook this up in maybe a minute and a half on Suno, and it was a lot of fun. Would you like to hear it? 

Marlon Avery: Actually, we would. Everything. Are you. Are you. Are you doing the chorus? Are you doing the hook? Or is the. Is the voice itself aging rate, too? 

Adrian Green: This is my voice. I actually put it into. Into the, you know, uploads there and then gave it a prompt to write a country song about driving down a dirt road drinking whiskey with your dad. You know, the joke was wearing MAGA hats. I don't want to make this, like, a political, you know, a political thing, but like, that, it was like, you know, I'm hanging out with my dad, we got our mega hats on, and then it turned into a whole thing in the. 

Marlon Avery: In the. 

Adrian Green: In the group chat, just as a joke. But I put that into the prompt and it wrote, let me see here. That goes on. That goes on for a full song, full hook, everything. Yeah. 

Marlon Avery: If you're about to release a country album, you know, we would like to ask you to sponsor this podcast, you know, before you blow up and stuff. 

Adrian Green: Yeah. Before I change up. Up on you. You know what I mean? 

Marlon Avery: Right? 

Adrian Green: Yeah. 

Marlon Avery: Oh, man, that's great. Seu, how was your. How was your week? 

Sekou Doumbouya: Oh, man, busy. It's always busy in my world, you know, But I would do a little bit of a little professional development using my using Notebook lm, which I. I've been finding very, very interesting. Like, I've been just tossing information in there, just, like, things that I know that are, like, already publicly available about me and trying to, like, get, like, having. Having it create a podcast based off of me as a subject I found super fascinating. And I think the AI is like, other LM is like, tuned. Tuned to Be super positive. So it ends up like if you're, if you're, you're having a bad day, you know, put some information inside of a notebook, LM about you and have it do a podcast and nothing will cheer you up better than that. 

Marlon Avery: And look this, you, you should, you should, you should have a web hook attached to it that sends it over to Alexa so you can get an alarm clock that wakes you up with this new generated podcast every single day. You know, with your, all your information stuff with it. Oh yeah. 

Sekou Doumbouya: Oh, that would be, that would be wonderful. That would be wonderful, I tell you. 

Marlon Avery: Oh man. Yeah, man. On my end building stuff too as well. Have a, have a client and working, working with and building using computer vision tools kind of on the front end and stuff. There is use computer vision tools to kind of scrape yourself the information and stuff that they gave us from some of the, we'll call it, we call it aspects of stuff that they're using. And so use some computer vision, descriptive information, take information, send it over to OpenAI, everything to do everything like summaries, extractions, update, character recognition when it comes to people, a few other things. And then sending all that information and storing it into a video vector database too as well. And so I ran into some challenges and stuff there lately, were they. But I was able to figure it out and stuff. And so yeah, man, it's been my week and everything so far. 

Adrian Green: Yeah, dealing with the embeddings is a pain. It's a lot of like just, you know, hope, hoping that they're going to go through and really map and index correctly. But you get used to it. Like the more that you work with it, for sure I've learned more about how to be effective with it. 

Marlon Avery: Yeah, I'm actually, actually I'm probably going to ping you because there's some questions I have over there. And so I was trying to give OpenAI a chance to use their database, but I might have to switch to pine cones from my titan. I'll definitely hit you up and everything because that's annoying. 

Adrian Green: I hear you. 

Sekou Doumbouya: Should always give before you leave open AI's ecosystem system and using it using all of its tools. Just try it once in claw. Just try it once, try it once, use that, use artifacts once and see how you feel about that. I feel like it's a, it's a different experience. 

Marlon Avery: I, I see so many people that have become claw prostitutes and I, I, I'm seeing so many people on my end. I haven't seen that big of a Change Like I haven't got that big of a like difference in response or anything. And, and it's again it's a lot of still like in the prompting and stuff everything as well. But I haven't seen that big of a change stuff on my end and I've tried. I've taken times where I've like I wrote the same prompt and I'll put it in open AI and I put it in cloud and everything and it's just like. And then maybe maybe also too because I'm still in a free tier but you still get access to 3.5. Yeah. So who's. But but it. For me I just haven't seen that big of a change. I haven't seen a bigger change to Switch. I will say that's just been on the, on the website. I haven't tried the API side and everything. But yeah, and so yeah I have. I just haven't entertained that big of a, that big of a difference. It for me is actually kind of added some steps and frustrations and everything. 

Sekou Doumbouya: So yeah, API is just a tad bit different too if you're using that. So so but yeah, you know, it is. It's really. The LMS are super, super close to each other but the, the tooling that they have around it. I think that is what the differentiator is around with, with Claude. They both have like their you know, pros and cons, things like that. But when it comes to like you know, memory and how that's used, I feel like that is the area that OpenAI is kind of catching up right now to cloud on. Somehow they leapfrog them on some of these like components. But OpenAI has come back with their agency workflow that they're promoting. We'll see. 

Adrian Green: One question about that, Seku. So when you say how they're the memory utilization and you know, how their system is structured, how does that show itself to you using the application? Like what are the key. Is it speed? 

Sekou Doumbouya: It's. It's speed. You know, the chance of it like veering off from what you've put into the actual and their version of embeddings which is the artifacts. And also they do some little tricks so that you don't get chosen charge by interacting with the, with artifacts as much as you do with using it on the open AI side. But I think that's kind of the main difference is like you. There's a, there's a higher chance you put something like in artifacts for it to actually be relevant to the your session than it is in open. Open AI. So. 

Marlon Avery: Yeah, I'm have to keep. I'm have to keep playing with it and stuff for sure. But yeah, I've. I've tried it several different times. I also think here's a question on the. I know I remember around the initial time where Claw was definitely starting to get popular and I was asking people how they felt about it. It also felt like there was a component of Claude was giving you slightly different responses than ChatGPT, which made it feel like an entirely new experience because we've been using ChatGPT for, you know, consistently for so long. So a lot of times we were kind of getting some of the same, you know, same window of responses and it's like, you know, claw camera and give you like a 10 difference and you're like, oh my God, this is the best thing in the world. You know, I've, I. What do you guys think about that? Because again, I think it's also to coming back down to high detail and how to find it so you can write your prompt. That may be why I just haven't seen that much of a difference or anything per platform on which each platform. But what you guys thought. 

Adrian Green: It sounds like what I'm hearing is like Coke versus Pepsi, MTV versus VH1 kind of, you know, whatever kind of format best better suits you would be the one that you would choose in that. In that case, if the answers are basically the same, but it's the formatting, it's a style that you, that you like better. It's like, you know, that's the differentiator. 

Sekou Doumbouya: That feels like probably the best analogy right now. Oxyrunner so OpenAI's competitor to artifacts is Canvas, which has the integration of, you know, adding different components into it. So that's. That is. That is the compare comparable piece. So if you are using Canvas. Yeah, I am to s the one that you're using. Yeah, Artifacts is direct competitor to that. And there's just like I said, there's. When you go down, I'll send you a. Just like a. Maybe you can actually post this as like a artifact after the show. Showing the comparison between Canvas and Artifacts. 

Marlon Avery: There you go. Like a little work subsection. Yeah. Yeah, it's. Yeah, I would definitely love to. Because I would love to see like you guys experience too as well of what's the big difference, you know, stuff there. And so, yeah, I just haven't seen that. That huge difference. And I, I've used Canva too as well. I mean I know it's in beta, but I've definitely found some. Some issues and some bugs and stuff with it. You know, just wait. Does opening. I have a bounty program. 

Sekou Doumbouya: A bug bounty program? Yeah, that's a good question. 

Adrian Green: That's a really good question. 

Marlon Avery: Maybe my new side hustle. Yeah, I know. 

Sekou Doumbouya: So. 

Marlon Avery: Because, yeah, I found. I found a few issues and stuff with it and how it's taken. Please send it to me because. Yes, guess who's about to get sidetracked. This guy. But yeah, so yeah, I definitely found some issues and everything. I quit it too, as well. I haven't. I haven't tried. Actually. Yes, I did. I did try claws Artifact and I tried Canva into as well. Camera definitely sim beta, but it still has some issues. 

Sekou Doumbouya: Make sure I'm saying the wrong thing. I said. So Open AI canvas is what I was talking about. 

Marlon Avery: Yeah. 

Sekou Doumbouya: Oh, but you're talking about your use of CANVA itself, the new integrated GPT. 

Marlon Avery: That they have no. 

Adrian Green: OpenAI canvas. 

Marlon Avery: Yeah. And so, yeah, they got some issues and stuff. I found. 

Sekou Doumbouya: Bug bounty. There you go. 

Marlon Avery: Yeah, there you go. I know somebody. I'm not. I know someone made a significant amount of money when Google runs on Google, open their bug bounty program. Like some, like almost like a decade ago or whatever. Quite some time ago. And so, yeah, so, yeah, yeah, we definitely figured it out. You know what? Something here. Seku, you and I have something in common. Adrian, I know you're gonna join the family here soon. You know, we both. We both have Teslas and, you know, we. We definitely have our, you know, experience and, you know, I think also too, the autonomous experience is all based off your understanding and expectation and stuff of it. And so, you know, you can be. You can kind of get into it with an awe if you have like a low or just zero amount of expectation. But, you know, if you are expecting it to be, oh, man, oh, Night Rider, you know, and stuff, and then just kind of like. 

Sekou Doumbouya: Although the summons does make me feel. I started using the summon when I go when I get off the train. And I do kind of feel like, you know, Michael Knight a little bit, summoning the car to come to me. It's. It is a pretty cool feeling. 

Adrian Green: I. That's my favorite part. That's. That's my favorite feature is the summoning. Because, you know, for instance, where I live, the parking is really, really tight. If I could summon it out of that space just to, you know, just. That would be huge. 

Sekou Doumbouya: But also, this is where the expectations come in. Like, make sure I always make sure there's no obstacles, there's no cars around me. I'm getting there. Like the parking lots empty. There's no pillars next to the car. 

Adrian Green: Okay. 

Sekou Doumbouya: Be able to see because I've, I've seen some folks on Reddit who have had a bad time with their. What they're summoning. 

Adrian Green: That's good to know then. That's good to know. 

Sekou Doumbouya: But expectations as you say, Marlon. 

Marlon Avery: Yeah, yeah. You know what I would like to see? I would love. And this is not a promoting to this area. I just know, I just know there's one individual, there's one case that beat a DUI because he or she was on fsd. 

Sekou Doumbouya: Really. 

Marlon Avery: I just, I know it's out there somewhere. I just know, you know, because it's like, is it, is it the law? Is it the law that I'm behind the wheel or is it the law that I'm driving, you know, the car and everything. I'm behind the wheel and I've had a few, but I'm nowhere near driving this week. And I just know there's at least one person, especially when it first kicked off, that has, that has beat a DUI case because they were in FSD for self driving. 

Adrian Green: Wow. The person got in there and represented himself, probably had all the evidence. 

Sekou Doumbouya: I think that's the future. That's Elon's future. You know, you could be sleep and have the car drive you around. That's like the, that's what we're being sold right now. What the future he's envisioning. We can sleep and. 

Marlon Avery: Yeah, sleep. 

Sekou Doumbouya: You can be inebriated. 

Marlon Avery: Yeah. I, I do believe there's a potential to get to that future. I, it's, I, I would say some years away. You know, I mean just even take out regulations, let's say like regulation and insurance is approved orbit today. You know, the safety of it to take a nap or just completely be, you know, hands off. You're in the back seat like where it may be. I've seen, you know, my FSD freak out so many different times and it's not, it's not consistent, you know, but it only takes one time, you know, for it to something to happen. And it kind of just, you know, go off the road or, you know, freak out or just hit the brakes or whatever it be. And so I think we're, I think we're some time away, you know, from. Yeah, from that being a safely reality versus a reality. 

Sekou Doumbouya: Yeah. Well, so see, we'll see what Elimina says Was Xai. Maybe he'll speed this up a little bit with his, you know, the GPUs he's building behind the scenes here. 

Adrian Green: So yeah, that's one of our topics today, I believe. 

Marlon Avery: You know and, and speaking up too. I mean he definitely has some competition and stuff now Waymo, you know, Waymo has recently raised $5.6 billion and stuff in funding which is the key area is they said and they have, they're going to now expand into new markets and those, some of those markets now are Austin in Atlanta. I've actually got to see a tick tock or somebody hopping in and stuff in Atlanta and stuff. So. So that was cool to see fellas what you guys thought. Waymo is starting to expand here. I also kind of find it funny that they're expanding days or a couple weeks after Tesla announced the Tesla, the Cyber cap as well. And so you know, look a little fun kind of competition and poking yourself there. Adrian, man, we'll start with you. What's your thoughts? 

Adrian Green: I just think it's interesting we were talking about this a little bit earlier. The Waymo presence out there is kind of, they're just gaining more traction, covering more area. It's, you know, when we have this alongside of the Robocab promises it's, you know, I start to question how much can the robo taxi area of the Tesla business be driven by just the appeal of Tesla Elon Musk and is is involved in that as well. Driving the driving that. How can they compete with someone who is handling, you know, really going through all the work of clearing the regulations, normalizing the self driving cabs. You know, maybe it is tactical at the end of having them maybe pave the way for you know, Tesla to maybe expand into those markets too. 

Sekou Doumbouya: Yeah, I think we also, we were chatting a little bit about this and before the show and like there's, there's such huge value being the first to actually do something like this at scale. Right. You know. 

Adrian Green: Uber is still riding on that. 

Sekou Doumbouya: Yeah, Uber still, they're still trying to figure out, figure it out. They have deals that they're starting. You know, you have, there was another company that got Cruise. They got kicked out of the Bay Area because they were like crashing into things. So they're, you know, their trust is, is completely gone. But it's so interesting. I had a, I had an on site last week at my job and the number of folks after the onsite who were like, hey, I want to get away mo. It became like almost like this Novel thing. I'm like, okay, Waymo's winning right now. And once you start that lead, it is so hard to catch up. 

Marlon Avery: You have to do something once it becomes a verb. 

Sekou Doumbouya: Yeah, almost like, yeah becomes like Xerox. Almost like I'm gonna go to Wayo. Like I'm gonna Google something. They're. They already, like, I feel like they already have this in the bag now. I like the Tesla folks have something interesting because they have these new fancy cars that they're going to introduce. But like the fact that they added a whole another step in front of them, releasing their, their robotaxi and the lead that Waymo is going to have is going to be, it's going to be like Uber Lyft in some way right now. 

Adrian Green: Describe that step that they have. 

Sekou Doumbouya: Oh, the. Which one? The step two. 

Adrian Green: You said there's an extra step with the robo taxi. 

Sekou Doumbouya: Oh, with the robo taxi. Because they have to build a car. Yeah, have to build a car in order to do this. There are cars on the road right now that Waymo has. There's no. Right now they just need to acquire customers, inquire in, increase their brand. Tesla, like it's going to have to go through that whole process itself, like once you're first in a market. 

Marlon Avery: I disagree. 

Sekou Doumbouya: Well, Tesla's still early enough. While there's lots of Teslas out there, like, it still is associated with more of, you know, unless she's, it's like, it's still like a wealth component. Even though Teslas have like dropped in prices to it. The everyday person getting a Tesla, interacting with the Tesla is still a novelty. Right? A lot of folks. 

Adrian Green: A lot of folks. 

Sekou Doumbouya: You got new one. There are folks in San Francisco of all walks of life that have gotten a Wayo because Wayo's cheaper. Like, it's, it's, they're, they're, they're building their brand around what automated fleet looks like right now. And I think that's, you know, there's ground that they can make up being Tesla and having all the personality that's connected to some of that, but I don't know, I, I expect way more to expand their lead that they have right now. 

Marlon Avery: Yeah, I must, I disagree. I'm gonna tell you why. It's the first understanding and the first acceptance and the first awe of self. Driving is tied to the brain of Tesla. It's, it's tied to Tesla, you know. You know, even, even when you know you're, you're driving and everything and people come ask you questions, you know, they want to first question they ask you is either the self driving or the battery, you know, and stuff of it. And so that, that is a conception that's already planted inside of the human mind, even if it is a seed. And so now when it's time to make a decision and make the switch to self driving without a driver, it becomes, it's like steps are already been made. It's like steps are made versus when you are in Arizona and you made a herd or you might have seen Waymo, there's still elements of trust that you have to go through, through that process as well. And so it's the conception of Tesla's already been doing this, I've heard about it, you know, things like that. Okay, cool. Yeah, I'll just hop into it. You know, like you said, their, their biggest problem is either they're gonna have to go and take those leases, I mean if they do want to catch up or if this isn't their plan, whatever, either you're gonna take those leases that they said they was going to use and go ahead and start to kind of like turn those on, you know, for you know, Uber, you know, taxidermy is okay everybody call it because as you know, there's no such thing as quick reductions, you know, on the manufacturing side with Tesla. And so I think Elon said 2026, which he, which means two in 20, 2028 stuff and everything, you know, but with also way more, you know, they have to build the cars and stuff as well. And so I, I think it's the, it's the, it's the, it's the entry of autonomous driving and it's like that entry, a lot of seeds have been planted and it's a tie to Tesla, you know, with that way more still has to like go build that, particularly outside of the west coast, you know, so from the rest of the country as well. And I would say this too as well. While we're looking at Waymo and we understand that Tesla is about to head in that direction, my concern for 7 million people and I, I just know they're waiting to hit this on button is on Twitch is Uber. 

Sekou Doumbouya: Oh yeah, Uber. Uber's not just United States, right. Uber has a international presence. You go to another country, you're not going to see left. You'll see Uber wherever you're going to be as a default system for, for taxis and things like that. You know, oddly enough, I feel like that's a, you know, being able to get to those locations to bring your technology out there. I feel like, yeah, Uber is pretty well equipped for it, but trust in public is a big thing. Thing. I think wh mo seems to be gaining a lot of steam for it, at least from what I'm seeing in like in the Bay Area. Like yeah, it is a, it has. 

Marlon Avery: Become more, it's been there for some years too. 

Sekou Doumbouya: It's been for years, but like now like they're everywhere. They're everywhere. You see them all over the place and people are opting for the wayo. 

Marlon Avery: They're efficient too. Like they, they are efficient. Like I don't, I haven't seen a lot of bad experiences, you know, within the wayo and so like they're, they're, they're efficient as well. I can't, I can't remember the autonomous driving one that Domino's is using. I don't know if it's way more or not, but they have, they have cars. I think it's like the, it's like the Ford was like the Ford Taurus or something like that. And it's like the, it's like the back seat is a warmer kind of like not a oven, it's like a warmer that keeps the pizza and stuff warm. And then you go put it, you put it, you put in a code once the car arrives and the back right window goes down. And then like you can grab the pizza and stuff from there. 

Sekou Doumbouya: Order some dominoes just for this. 

Marlon Avery: Like, like, hey, is it going to, it's going to be self driving? No sir, we're not. Okay, cancel the order. Like, oh man. 

Adrian Green: I have a question about the, the, Is it, is it still a chance that Tesla can. One of the early promises was that you're going to be able to actually this car is going to be able to drive itself and it can be a taxi while you're sleeping. Is that, you know, still a potentially something that can potentially happen, you guys? 

Marlon Avery: I think, I think, I think the, I think it's going to happen in steps. This is what I mean going back to the Focus right now across the board with artificial intelligence right now is the simp, the, the connection of tools and the integration of tools and everything. And so like okay, we have the foundation models and stuff, we built everything. These things are working now. How do we connect it to the rest of the world simply and stuff. That's what the Focus is. Same thing right now. In my opinion it's going to happen with the autonomous driving and some aspect because I believe that it's not it, you're gonna, you're not. You, you're gonna. It's one major step that they're not talking about, which I believe is going to happen first because it's safer, you know, to happen first is partnerships and connectivity with other organizations. Example, the ability would that your Tesla will be able to sit in the car, you partner it with, with Target and stuff when you go make an order and the cargo picks up the order on your behalf and then comes back. Yes, you know, I think that's where we're headed. Or the ability that you can connect it to, you know, the Atlanta airport. The car goes pick up, you know, your, your cousin stuff from the airport and everything and then comes back and everything. You're also going to have to figure out how to do connectivity and charging in between that too as well. That can be figured out. Another thing too, they were talking about this wireless charging with it with the, with the, with the Tesla's cap. I don't believe that whatsoever. Anyway, so, you know, you're gonna have to like figure that out. Well, you're gonna figure that out too as well. Once those connectivities and stuff would happen and specifically for the consumer who, you know, who has textiles, I think then after that they will start to build where it's like, oh, by the way, connect this to Uber, you know, connect this to Uber to go make, you know, X amount of dollars or anything, you know, in these X amount of hour stuff, you know, to as well. I think that would happen. Yeah. I don't think you'll jump to the next step. I think you focus on the first because the first one still serves the community. And also too, it gives you understanding of infrastructure, you know, within. You know, how to do, you know, things like time estimations, you know, how to do allowing, you know, somebody's employee into your car, like all different things like that. You know, I think you go there first for the consumer and then you go to the next one. 

Adrian Green: Yeah. And because there's a roadmap for that is why I say Tesla's still competitive in the autonomous driving space because they're. That that could potentially happen. 

Marlon Avery: I mean that, that, that is it. I mean we're talking about now. I mean a couple years ago it became, it became mostly like wildly accepted for Ubers, you know, for like Ubers to take your kids during the. Into school and you know, things like that. I mean now we're talking about your car is going to go take them to school, drop them off of school, you know, waiting that line or whatever it may be and then, you know, go pick them up, you know, as well. And so, and then you're not, you're not paying an extra, you know, 25, 30, $40 per day or what it may be, you know, you're just using, you know, your own vehicle, you know, to do so. And by the way, you'll have it scheduled. You, you, you'll see scheduled charging become over focused or something too as well, because then you'll be able to, you know, integrate it with your calendar to understand like what, what where it needs to go and stuff, like in the next, in the week. So it'll tell you like, hey, to make sure that I get to daycare, you know, we need to charge. Just like your phone says, hey, charge before you go to bed or your Apple watch, charge before you go to bed. It'd be the same thing. 

Adrian Green: Yeah. Now I'm going to go even more futuristic. Is it possible that we see it. 

Marlon Avery: Bring it on. 

Adrian Green: Is it possible that we see a future where people will actually, you know, make money off of that system? So say there's an event happening. Hey, look, we're gonna drive to Atlanta and park the Tesla because the Tesla is going to, you know, do act as a cab around this stadium, around this game, during this game. 

Sekou Doumbouya: Oh, you're doing like a. Yeah, like self driving Toro. 

Adrian Green: Yeah, Right. Yeah, basically. 

Sekou Doumbouya: Yeah. Oh, yeah, I'm sure. I feel like that's definitely on their minds especially. That's kind of like the dream, right. When they were building the Teslas. One of the visions was for you to not only have the car that you can use, but the time when you're not using your car, your car be doing something for you, making money for you. 

Adrian Green: In that case, it's like a capital good where, you know, a normal car depreciates the, you know, when you drive it off, off the lot, it definitely affects that because if you're going to be able to earn money back from. 

Marlon Avery: The car, then that's, that's what, what. 

Sekou Doumbouya: Uber is kind of built off of. Airbnb is kind of built off of that same thing. Utilize time. 

Adrian Green: Yeah. 

Sekou Doumbouya: Or space. Right. 

Marlon Avery: Yep. 

Sekou Doumbouya: So, yeah, it's, it's, it's definitely inevitable. 

Marlon Avery: I'll take it a step further. You know, as of lately, it seems like Elon's number one enemy, his arch nemesis is the government or himself. 

Adrian Green: Little column, A little combie. 

Marlon Avery: Yeah, I'll take a step further. Anyone remember one of the main reasons, or probably the main reason why Amazon, Amazon Prime Got got created. 

Sekou Doumbouya: Undercutting FedEx or something. Or UPS. 

Marlon Avery: Yep. USPS. 

Adrian Green: It was USPS. Okay. 

Marlon Avery: All right. It was USPS everything. And you know, Amazon, they got sick of it and they created this whole initiative and everything around Amazon prime and you know, deliver home packages and things like that. For the ability for your Tesla to go deliver packages, you know, as a, you know, delivery driver and stuff where you can make money and stuff too as well. 100 see that happening, you know. Yeah, I can. 100 I can, I can see him doing it as an ego, like an ego thing just to stick it to the USPS and also to piss off Amazon, you know, as well. And like, you know, now you're talking about you know, peer to peer packages that can be exchanged with especially particularly within the same city. And so you know, you're just ma. Now you're making orders. You know, you're doing orders, you're doing, you know, in some ways you're like, you're doing doordash, you're, you're doing package handling, you know, and stuff as well. And it's literally just like your Tesla is kind of going or something as well. And so I think there'd be some time and stuff off. But I can definitely see, you know, that happening just for him as a, just, just to kind of stick it to some people. 

Sekou Doumbouya: I can only imagine the amount of insurance that has to pay. Yeah like the soul brain person at all these organizations because you have to. 

Marlon Avery: Law, legal, you're required to do right. 

Sekou Doumbouya: As a fiduciary is to have some type of insurance to make sure that you know, the company can continue on without you. Like I've never seen any this, this many businesses that are hinging on one person. I feel like SpaceX is a little bit more autonomous now but he's definitely like, you know, he's heavily involved that catches a giant, you know, rocket. Yeah, I feel like that came from the mind of like you, Elon Musk or someone that like he's brought into that orbit. So I think that's going to be. I think that's interesting. Like all these things are hopes and dreams are really attached to this one person if he gets bored or changes his mind like. 

Adrian Green: But can we just talk a little bit, can we just mention a little bit about that, you know, the catching of the rocket. You know me of me. You know, just looking at the progression here, I saw them try to land the cylinder back down which is basically the flip the bottle game. You know what I mean? It was falling and it was, it was difficult to get that to stick. That landing particularly you got, you know, exhaust and you know, acceleration to decelerate perfectly to get this thing to land flat. It makes so much more sense to catch the rocket, you know, like how. 

Sekou Doumbouya: It makes, it makes sense but like at the same time it's crazy because. 

Marlon Avery: It'S like, it's cool. 

Sekou Doumbouya: Skyscraper. Yeah, yeah, really, it's really big. It's like you're catching a skyscraper with tweezers in midair. But yeah, it's, it's a me idea. And the way that they pulled it off is like, oh, all right. Well we just like made a huge step for humanity just now for sure. Quick idea here. 

Adrian Green: Yeah. 

Marlon Avery: Also, also, also think this is like, this is one of the main reasons that kind of almost what you said, Seku. But from a company standpoint, it's also one of the main reasons why, why they created the Tesla car insurance. It's to hinge their bets, you know, as well. And so the ability that your car would be to do autonomous things and I would imagine you probably have to have to have test insurance because I just know I can, I can. 

Sekou Doumbouya: No one's gonna. 

Marlon Avery: I can see. Yeah, yeah, yeah. 

Adrian Green: And so, you know, guys not showing up. 

Marlon Avery: Yeah, yeah, yeah. So Tesla like, yeah, we got it. Switch over to Tesla insurance and everything and we'll take care of it, you know, everything at a low rate or so what it may be. And so I think this is why one of the early reasons why they switched over, you know, to announcing Tesla insurance because of things like this where they get it ready for which is just. It's, it's a, it's, it's the, it's the right hook you don't see coming. And it, that is going to, that's going to hit a lot of like a lot. I mean insurance itself, that's his, that's his own discussion everything and so but that's gonna, that's the right hook you don't see coming because that's going to impact a lot of businesses. 

Sekou Doumbouya: Yeah. Yeah. Wow. 

Adrian Green: Yep. 

Marlon Avery: Well, also too speaking of Elon, you know, he has announced this thing, the Xai Colossus, which is a supercomputer. It says environment considerations of such massive compute is going to power and the SMR role is sustainable for sustainable energy. Says the role of the super computer, super micro and a specialized SMR power source in building, you know, this infrastructure. Seku, you know, this is one of your, you know, the, one of your expertises and stuff in the area, man. What's your thoughts there on XAI turning on this? 

Sekou Doumbouya: Yeah, yeah. You know, with. It is. It's a pretty big feat. So it looks like they partner with Super Micro to get this environment built out. It says that took 19 days to rack, stack, plug in, put OSS on these systems, you know, and like that. That alone is crazy. Like how many engineers it takes to do that. The level automation that it takes to make sure you have to have in place, you know, there. The thing that I think the thing that like, you know, you have to remember when, I guess when they even give you the numbers of 19 days is like that's really like stacking, racking and stacking machines and getting them. Getting them going, which is a big feat. 

Adrian Green: It's a physical feat too, when you think about it. Like. 

Sekou Doumbouya: Yeah. 

Adrian Green: Along with technical feet. 

Sekou Doumbouya: Yeah. And I question, like, were there robots involved here? Because like, like, you know how like I worked and like your back will be shot just plugging in cables. Right. You ever seen the, like the. Was it the videos on YouTube or Tik Tok of people who used to do those old, those dances, stuff like that from like the, from like the, the late 90s, you know, and them. These same, same folks in their 40s with their bad knees and stuff like that. Yeah, like, I like, I can imagine that's a lot. 

Marlon Avery: That's a lot. 

Sekou Doumbouya: That's a lot body. But the thing that kind of comes up is it kind of connects to this topic we're talking about of powering these things using H1 hundreds. So for H100, it pulls about at least at peak, maybe 700 watts per H. 100. Right. Times that. Times. 

Adrian Green: Well, it. 

Sekou Doumbouya: Right. 

Adrian Green: It's a little bit less. It's like 300. 300 watts per GPU. 

Sekou Doumbouya: 300 watts per GPU. But that's. Is that peak or is that like average? I guess that's the. 

Adrian Green: That's probably an average. 

Sekou Doumbouya: Very average. 

Adrian Green: Yeah. 

Sekou Doumbouya: Yeah. You're probably not running at peak at all times, but just this idea like the amount of power that is you times that by 100,000. And you start to understand a little bit more around why companies are seeking nuclear reactors compute for their data center. Small modular nuclear reactors for their data centers. And it all kind of comes together. All right. 

Adrian Green: Yep. 

Marlon Avery: Adrian. 

Adrian Green: Because it. Yeah, it's going to be us talking about how the. That was set up so quickly. Sounds like us discussing how the pyramids are built. You know, how did, how did people put this together? You know. But I guess when there's a will, there's a way, you know, so I agree with you that it's. The conversation does turn into power. You know, I'm sure all of you have maybe felt or just questioned a little bit when you put a query or a prompt into chat gbt just how much energy, especially when it takes a while to come back. It's like, man, is there like a whole coal plant burning to give me back this answer? Like, how much power is this? Is am I, am I using and you know, contributing to the overall consumption of this system per quarry? You know, it's interesting, but. So for the amount of energy that it takes to run this super cluster, it's, it's perfectly suited for the SMRs that are coming out, which cover power within that range. Talking about like roughly like 50,000 homes that it could power. So there's also, you know, on the top disruptive stock list, you start to see the appearance of companies that do that that are setting up these SMRs. And the more that I looked into it, it's, you know, it's, it's, it's so funny that when I'm curious about something, I look into it and Bill Gates is already involved in it. You know, he's already, already there. So he has a SMR company as well. So it's, you know, smaller footprint, smaller, smaller geographic footprint. And it's, you know, these small nuclear reactors could be perfect for these data centers, these big GPU clusters that are being set up by these companies. So it's an ad for that too, as well. 

Marlon Avery: Yeah. Here's a question. 

Adrian Green: We. 

Marlon Avery: The. Okay, in our cell phone life network, what was after cdma, CDMA switched to. Oh, the gsm. 

Sekou Doumbouya: CDMA and GSM were two competing ones. And then we went to lte. 

Marlon Avery: Yeah, that's what it was. 

Adrian Green: Yeah. 

Sekou Doumbouya: Yeah. 

Marlon Avery: So we saw the switch From CDMA to LTE underneath the kind of LTE umbrella, we got 2G, 3G, and then we start to see the expansion to 4G and now 5G and what it's looking like consume here at the infrastructure. If you, I guess you want to continue it, we call it 6G unless it switches to something else. But with that from, I would argue from 3G particularly to 5G, we saw an expansion of IoT devices. And that was the reason, One of the reasons to do so is because you needed more kind of like pathways, highways and stuff, if you will, you know, for it not to kind of like get a traffic jam here on the network and stuff. When it comes to like speed and access. Post that as well. In this 5G world now you have individuals who can, they're having drone companies. You know, they're doing drone things like drone photography. You know, they're doing things like drone was, it was like drone artwork, like fireworks and things like that, you know. So yeah, light shows, you know, stuff you're starting to see kind of like more even online RC cars and stuff where people are having races, competitions and things like that, you know, having fun which that itself can you know, become a business as well. Along with the things that we've already been experiencing from your Alexas to you, you know, to your, your Siri devices, Google, things like that too as well. Which is also aspects of business to as well. The question is we saw what that network did when increase opportunity for capacity, the increase of GPUs, data centers. It's going to increase the capacity for rendering, for clustering. I'm not for rendering for inference and all these things like that. The question is what is the output of that for your consumer, for your everyday person? Like what are some other opportunities you think we'll see once we kind of have a let's, let's call it more compute? Yeah. What's, what's, what's your guys thoughts there? 

Adrian Green: Well, I think that we're, you know, we're betting on the promises of the big players and how they're going to use it like meta. They continue to dump a lot of money into AI with you know, promises that this is going to enhance, you know, the metaverse and the other, you know, billion things that they do in order to make better products for the consumer. So really I would say firstly it's. 

Marlon Avery: Hold on, do me a favor, switch it. I want you to, I want you to talk to the listener, the consumer and everything and give me some, give me some ideas of how they're going to benefit. So for example, it was very hard to predict that from 3G to 5G that somebody may be able to start their own drone company, you know, and stuff and everything. And so you know, give me some ideas where you think, you know, the listeners up here will, may be able to how some things we'll be able to benefit from or how they'd be able to start their business. What type of things, stuff you'll come out of this. You know as well. 

Adrian Green: I would hope the businesses, number one, I would hope that it supports local municipalities so it would enable the local municipalities there to be able to analyze water consumption, energy consumption, all the data in order to optimize how those systems work and maybe predict, give some predictive analysis as well. I think that's the first thing that comes to mind. 

Marlon Avery: Singapore is working on something similar actually. They launched it like two years ago. Silver Surfing AI allows the, allows the, the, the civilians there to tap into their government led AI where they can ask, you know, questions, get understanding of laws, you know, hey, where do I go directly to, you know, get this, you know, get this license or whatever for my business and you know, things like that. So yeah, I, I like that. That definitely can be a benefit for you know, the, you know, the consumer sake. What's your thoughts? 

Sekou Doumbouya: Yeah, well, I think actually the, our government is like telling folks that they should be using, the government should be using AI, right? For, for doing things. But one caveat that they, they've thrown out is that we shouldn't be using AI to do decision making itself, but we should be incorporated in all parts of the actual government itself. So there's some big pushes happening in the US from there. When the government do stuff, you know, I'm sure there's like, you know, there's a tactical advantage that we're trying to actually build from that. But those things that they build eventually do come to consumers. They eventually do roll out broader than just, just for themselves. But you know, they're super slow. There's the slowest thing in the world. Don't know. Wait for them. I don't know. I, I think right now I so sorry. I wish. I want, I want the open. I want, I want to get more efficiency in the technology to get it closer to running on our devices, better chips that we can run locally because companies that only, they only have the interest to actually hoard, they don't have. There's no, there's no, like, there's nothing driving them to truly be open now. They're, you know, Facebook does some amazing things with Llama, but it's, yeah, it's for a different reason. It's for, you know, impeding progress to the rest of the industry. 

Marlon Avery: It's indirect competition. 

Adrian Green: Yeah, well, it's, it's, it's impeding their, their money. I mean the progress I feel like is being made. But if you release something free that does 80% of what you're charging a premium for. 

Sekou Doumbouya: We need to, I think in order for us to like realize a place where like the everyday person can really take advantage of this technology, it needs to get on devices. It needs to no longer, it needs to get out of the Nvidia you know, world Nvidia regime, I guess we can call it now because they're, they're that big, right? 

Adrian Green: Yep. 

Sekou Doumbouya: I think that's, that's the blocker right now. And with efficiency is coming so folks are making chips that can do more and that, that I think that'll get it. Get us close to being able to do some of these things on, on devices that are like local to us or like appliances at least that we can actually have in house and really start, you know, pushing further because there's still that level of trust. It's not. Everyday person is not. You would think most people are using these AI tools right now just for, you know, writing and things like that, but that is not the case. Actually. There's a study that was recently done that's actually saying that the adoption rate of AI tools is actually trending in the direction of males and females. It actually hasn't actually gotten as close, which is interesting gender gap in this area too. But I don't know, we need more, more of a Linux type offering for these, for these technologies. Things are closer to consumers before can really truly be useful just having something that companies are producing and giving. I. That only needs to go so far because the interest isn't truly the consumer. Interest is profits. Right. So helping you start your job, if you start your job and allows you to buy more products from Shopify or something like that, great. That's a good AI integration. But like if you're. There's nothing. There's not like general tools, they're going to come to you that don't like have like a steep. You know, you can see how like you look at how much we're paying for AI tools right now. It's kind of crazy. Like I'm looking at my account, all the different little things that I'm using and I'm like, wow, this is, this is, this is crazy. This is not sustainable. I know that I'm a hobbyist, so I can, you know, I'm investing in my hobby. But like this is not sustainable for a normal person. 

Marlon Avery: So yeah, two, three things. Marcus said something very interesting. He was just on Impact Theory podcast and he said that we're in a world right now where somebody can go build a $20 million AI solution and it is not more impactful than your $20 version from ChatGPT. 

Adrian Green: Yeah. 

Marlon Avery: And like your, your everyday consumer or stuff like have access to that also think too as well. Kind of like piggybacking what you just said. Sekou is the other side of compute. The other side of these GPU super clusters you mentioned is on device models and on device and self experience and everything typically on device sounds like you know, your cell phone. I'm gonna take it a step further and said this is where we really start to see robotics rise, you know, on the other side of that, you know. So if it's your, if you share what's, what's the Tesla robot call? 

Adrian Green: Optimus. 

Marlon Avery: Yeah, Optimus. If it's your Optimus. You know, if it's something, if it's something from Boston Dynamics. I know Google actually know DeepMind is there. They got a few, they have, they're a few things now in the office where like it, it'll show somebody whether someone like a new employee, it'll have them check in and the, the, the robot itself will like walk them or roll them, direct them. Geez. To their desk, you know and stuff, you know, for that doing for an employee. So yes, I do agree with the own device backs out of this. I think their own device will be insanely and beneficial for the consumer as well. Watch this. We, we're, I mean we're human beings are creative. You know, it's as well. I can guarantee you, I can guarantee you the minute those things get launched at a friendly consumer price, like whatever may be, I mean right now I think it's rumored to be like $30,000, you know, for Optimus and so forth. I can guarantee you you're gonna start to see entire companies pop up simply for accessories of your Optimus. 

Sekou Doumbouya: Yeah, yep. 

Marlon Avery: Clothing pads, yep. You know, jewelry, you know, other, other connectivity devices. So it'll be, you know, it'll have, you know, if somebody doesn't it basically have it like its own phone, you know and stuff that you can call it anything to request it and stuff. I can guarantee, you know, and stuff. And so that also too you know, introduces new jobs and new workforce and everything off of something that we thought was going to just kind of wipe us out or check us out where it may be. And so yeah, that's one of those, that's one of those things that will definitely, you know, see like somebody sometimes I'm gonna come a millionaire building Optimus accessories. 

Sekou Doumbouya: You know, I'm imagining like the NBA 2K outfit so you can buy a few characters, imagine those things. 

Marlon Avery: Yes, absolutely. 

Sekou Doumbouya: Optimist. 

Marlon Avery: Absolutely. Absolutely. 

Adrian Green: I still say, I still say it's going to be the robot dogs. If Boston Dynamics just decides to release a little one, this little toy poodle size dog that you can Train. And you can, you can, you can do that, man. 

Sekou Doumbouya: Okay. 

Adrian Green: People would love that. 

Marlon Avery: On that same podcast, Mark reasons Margarita said he has two robot dogs. Yeah, that's running around the house. Yeah, he said they have, they have. He said the one that these. The one have just have the legs or anything. They have one that's like has wheels and everything too as well. But he said yeah, he said runs around the house or they goes upstairs and stuff. He said it can outrun you. You know, it says. Well. So yeah, he says he has two. 

Adrian Green: That's great. That's great. And like maybe there'll be a. Once that gets launched, you know, when it's more publicly available, there'll be training data that people will like monetize and upgrades and add ins. 

Marlon Avery: Absolutely. I mean this is the entry of the Matrix when you download some type of program that you want to or dogs to start doing applying, you know, but also too there's personal benefits as well. So like now you know, sending your Boston dynamic dog in the car with your child who is getting Thomas he delivered to school just as a extra layer of protection. That's a lot of money anyway. 

Adrian Green: You know. Yeah, yeah. 

Marlon Avery: And so yeah yeah. This is about to be a wild world. So that means we need to make AI with friends Optimus T shirts specifically fit to size. You know, hoodies. You know, we'll go, we'll go to whole nine and stuff here. 

Adrian Green: Yeah, yeah. 

Sekou Doumbouya: Make it so you can maybe pick your. Your out outfit to put on top of your Optimus on your shirt for sure. 

Adrian Green: I think when like one of those big, those big fashion houses when they do the entire show, the entire Runway show with just Optimus robots. 

Marlon Avery: You're. You're so right. I can see it. I can so see it. 

Adrian Green: They gotta be working on that now. 

Marlon Avery: Yeah, you know it's funny. And we're going to see all this before GTA 6. All right, so here we go. All right. So Meta has announced a new tool, a new competitor stuff in this, in this C suite of tools Net. Meta has introduced Notebook Llama. You know, this is obviously a direct competitor to Google's Notebook lm. So the ability to develop ticket to do rag applications to have outputs of things like podcasts, summarizations, you know, things like this. And so face Meta has said, hey, you know we want to jump into this marketing tool as well. Yes, Google, we see your Notebook LLM and we raise you Notebook Llama. Adrian, back to you man Wish what's your thoughts and stuff here? And also too Wait, my bad. It also, it also is free too as well. And so we know right now no bucket limit is free, but for the time being, we know that is coming from Google. But Meta, say yes, we see you. We're going to go, we're gonna have it for you stuff here as well. Adrian, what's your, what's your thoughts on the, the notebook industry. 

Adrian Green: Yeah. 

Marlon Avery: And the competitive, the competitive landscape itself here? 

Adrian Green: Well, I think that I can't help but speculate that this is serving meta by decreasing the value of the other AI tools. So I think I mentioned previously too, if you release something for free, that does about 80 well or 80, you know, 80, 90 well, as the one that's released at a premium, you know, it decreases the value overall of the premium product. So while we're in the discussion about tooling really being the really way that AI companies are going to make money by having the appropriate tooling, Meta is not necessarily doing that. So they're not keeping up, as far as I know, with the tooling side of it and releasing the, those products that leverage the Llama. The Llama models. But what they're doing instead is just okay. Instead of developing. What we're going to do is kind of level the playing field out there until we get ourselves together on what we're going to do. 

Sekou Doumbouya: Yeah, yeah, yeah. 

Marlon Avery: It's like, that's a great point. 

Sekou Doumbouya: Race, right? It's like a nuclear arms race between. 

Adrian Green: Oh yeah. 

Sekou Doumbouya: All these different offerings from these companies that are, you know, pretty hostile against each other. So what meta did they, they pretty much, they released this product open source, like pretty much like all the. They've released like a couple of Juniper notebooks that not only describe how to do it yourself, but someone actually has put like something on hugging face that you can go and mess around with and has basically made it so that, you know the chances of this coming out of Google Labs and being a successful product and not having good competition, like it's definitely going to have competition. Someone's going to make something that is more well integrated because right now it's still in Google Labs. And the way that Google Labs work is actually like a, it's almost like a separate part of Google. You don't have the same type of like integration into product and things like that. 

Marlon Avery: They put employment benefits. 

Sekou Doumbouya: Yeah. 

Adrian Green: Oh, really? 

Sekou Doumbouya: Yeah. 

Marlon Avery: Can I explain real quick? 

Sekou Doumbouya: Yeah, go for it, please. 

Marlon Avery: So, yeah, so how Google Labs, I would, I would, I would argue they kind of adapt this from Apple and everything and then it makes me iteration around it. And so basically when a project goes into secret and stuff at Google or they want to kind of see or somebody has an idea where they want, they, they, they feel like this product or service would benefit Google. You know, it gets, puts this category and search, it puts into, you know, Google Apps. And so, so then you know, you either assign a team, you go to recruit your team, things like that. So say for example, you have 10 players, everything who are building, you know, this tool to service. And so for maybe. And so you get a Runway, you know, and you get resources to build this. And then, and then initially you launch in beta. And so you know, once you launch in beta and everything and then you start to make your steps and stuff, you know, for now. So at some point, you know that you present this back to, you know, leadership and everything and then make a decision on if they want to proceed or if they want to cancel the project. The unfortunate thing about it, if they cancel the project, you also can lose your job, you know, and stuff with it, you know. And so you know, which is a, this is if a, you know, risk and everything is risk reward type to it. But if they take on the project and everything now you potentially have the, you potentially have to be. You potentially are now like the VP, you know, everything or you know, the VPs of like this new area of Google, you know, they, that you help, you know, kind of like birth, you know, as well. And so, you know, that's like how that process worked. And they did the same thing too with the autonomous driving. They, they recruited Sebastian Thurm from Udacity, you know, to come do this. Sebastian was clever because the time he was still working at Udacity, he figured out a way where he was like, hey, I'll come help you guys, you know, come do this. Anything on the research side of the self driving, you know, car. But I'm still going to work at Udacity because you know, I know how you guys work. So. So just in case you decide to counsel this, guess who's not messed up. Yeah, and so, and so, yeah, so that's how, that's how that process works or anything. So it's a risk reward and stuff there, you know, but you kind of like hedging your bets on that. This could, this could work and it's gonna go through but you know, there's, there's definitely some, some risk and stuff there. Go ahead. 

Sekou Doumbouya: Yeah, so yeah, I think for the, for a notebook lm what they're doing right now is it's you know, this. The good part that's coming out of this is that, hey, I now have a new free tool. They provide a workflow for how to actually do this. If I'm interested in creating a product that can actually serve folks better, I can now do that. Now. Now we still, you're still facing, you know, Google, they can actually go and integrate this into a product and get, get the lab team everything they need to be able to kind of scale this and figure out how to monetize this, this notebook. Lm. But the chances of them actually investing in doing that is dramatically decreased now, which I don't know. This is. I if for the consumers, this is great. But like Meta's use of open source to like navigate and like depower all the. 

Marlon Avery: To weaponize. 

Sekou Doumbouya: Yeah, yeah, like it's, it's pretty genius. I got. 

Adrian Green: Yeah. 

Sekou Doumbouya: Like, gotta give them credit. And the fact they're using open source to do that, it kind of still feels, I don't know, it feels like some of the spirit that we had in the early 2000s when this, you know, tech was. This type of tech was like really building up. Like, I like it, I like it. So I'm super happy that Meta did this and you know, of course I'm gonna try to figure out what I can build this is. That's what's gonna happen. So. 

Adrian Green: Yep. Some of the other products that came out of Google Labs is maps, reader, docs, groups, trends, calendar, voice search. So there's a chance that it gets integrated into this suite of tools like those did in some way. 

Marlon Avery: Yeah, yeah, it, it definitely have it. There's, like you said, there's definitely a benefit and everything there and it definitely can work, but again, if it doesn't, you know, they don't, they don't see this as a, a useful, you know, process of their time. You know, it definitely can backfire and stuff on you. 

Adrian Green: I was surprised, you know how Google operates. I was surprised that they merge those teams. The Deep Mind and the Wasn't there merge. They merged a team. 

Sekou Doumbouya: Yeah, Deep Mind and the assistance team. 

Adrian Green: Yeah, yeah. Because Deep Mind was the one who beat AlphaGo or was that, was that AlphaGo when that one that won, that it was kind of a historic event when they beat a chess champion or something to that order. 

Marlon Avery: But yeah, I'm, I'm looking for. I can't, I can't because you just said something interesting so there. And I can't find it. This also too, with, with Meta making these moves, this kind of Goes back what you said, being first to market because I can't remember the name. It's, it's, it's like tip my tongue but being like first to market and stuff as well. You guys, you, you. And I'm pretty. Where we talked about this, you guys know, like Facebook at the time introduced chat BT a version chat like 90 days before open AI did. 

Adrian Green: Oh, interesting. 

Marlon Avery: It was focused on research and like the, the research team, like the research audience and stuff as well. And the research audience got, they really, they didn't feel like that this was ethical. They feel it was wrong. And they like, you know, they went off on Twitter and Facebook, closed it down and they shut it down. And then like 90 days later, opening eye and that's ch. And the world change and the world changed. 

Sekou Doumbouya: Yeah, I do feel like. So for me, I feel like Notebook LM has that has like that chatgpt potential. I think the more folks that use this tool are going to realize how powerful it is. We were talking earlier about like, you know, putting in like right now it's a really interesting learning tool, like being able to put in a research paper into Notebook LM having it create a podcast which goes through different details and types of questions that you think you would end up having to answer. Ask about that while giving a complete overview. Like, it's in for me, it's like, it's like a new way to learn in some ways. 

Adrian Green: It's the matrix. It's downloading it. And now you know kung fu. 

Sekou Doumbouya: Yeah, now I know kung Fu. 

Marlon Avery: And it's, you know, galactia. Galactia. 

Sekou Doumbouya: The, the way they go back and forth and they're kind of interrupting each other. That natural flow, it is, it is quite interesting. And I can see, I can actually really see folks using this for doing peer reviews. That's like a big thing. Like, you know, every year companies got to go through and pull in. Pulling. Every. Every year you have to like, do like these reviews of your peers and pull in all their documents and things that they've done, interactions that you've had and then write up a story of that, you know, that interaction. Like, what if they can provide that information to you and they provide you a notebook overview session going over that information. Like, imagine how rich that. Like if that can do it right? 

Marlon Avery: Watch this. You, you, you still have to have a human in the loop because are you going to trust human beings to keep up with all that information, you know, and stuff, as they go along with like, views and stuff, or, you know, do most of us at the last minute, you know, start to like try to gather against all your stuff? 

Sekou Doumbouya: Oh, you know, the worst of us is going to come out for that because you know, when it comes to reviews, it's like, oh man, I don't, I'm not gonna spend time just send. 

Adrian Green: Yep. 

Marlon Avery: Yeah, exactly. 

Sekou Doumbouya: What's gonna happen? 

Marlon Avery: So that's funny. The Facebook AI tool I was talking about is called Galactia. 

Adrian Green: Galactia, okay. 

Marlon Avery: Yeah, yeah, they shut it down like 90 days before ChatGPT. And yeah, ChatGPT got launched and then the whole universe got changed. Wow. Speaking of ChatGPT, OpenAI has announced they are going to step into the chip making game. Say they're going to look to partner with companies like Broad, BAM and TS TSMC to kind of develop its own AI chips. Shu, man. Going back to you, what's your, what's your thoughts and stuff here? You, you know, we, we know. Yes. What's your thoughts up here? We know that, that these things are needed. It's going to kind of, kind of come, you know, kind of going back into the same conversation with Elon and X the folk there. What's your thoughts on here with the AI? Decided to step into the chip game. 

Sekou Doumbouya: That makes sense. It makes sense. So chat GPT, the amount of money that no chat GB, the OpenAI and their products rely on these GPUs for training and inference right now. And they're optimized for using transformation performers. Right. And in order to like, get them to get things to like a level of efficiency that's required for them to actually get to some of their AGI goals and things like that, like the current paradigm of, you know, spinning up all these GPUs, even if Nvidia continues to progress and things like that, that is the, a tricky place to sit in. Spending time to get some custom chips created for you that are optimized specifically for your code base means one, you'll probably end up spending a lot less for amount of hardware that you need. The funding that they've gotten is going to stretch longer the chances of them getting taken over by Microsoft or another. Well, probably Microsoft is the only person that's going to, you know, gobble up open AI reduces slightly, you know, because they're going to have to ask for more money eventually. So the more that they can fit within what they have right now, I think that's going to be the best for them. And if chat GPT's model ever, not that it'll ever become open, if it became open this would be even a greater thing, right? Being able to actually have those models running outside of their world and access to these chips. Like, that's, that would be interesting. That would be an interesting thing. That would be something that would probably benefit more folks. But they're just trying to stretch their dollars and it makes a lot of sense that they're going and seeking out optimizations for their own, you know, software stack. 

Marlon Avery: Yeah, yeah, I think the, I think this was only a matter of time, you know, even more so with them also too. They're, I think we also, we, a lot of people forget they're, they're also behind the scenes working on the hardware side of things as well, you know, so we still haven't seen this device that they're partnering with, with Johnny Ivy, with Johnny Ivy, you know, former head of design over at Apple, you know, so we don't know if that's going to be an underby, on demand device, if it's gonna be a phone, if it's gonna be a wearable. You know, I would love for it to be a competitor or a wipeout of the AI pin. Yeah, yeah, I would love that. You know, but it also could, you know, could be a VR device, you know too as well. And so. Well, you know, we don't know yet. But with all that being said, you know, we're still going to need, we're still going to need, you know, they still need their, I can see why they're doing this, but they're gonna need their own chips, their own computer and you know, to be able to serve kind of self serve and everything size with their, with their inference and stuff. And so. Yeah, go ahead. 

Sekou Doumbouya: Can I add one thing? I think there's one thing that's also a little strange about this. OpenAI doesn't have their own data centers getting chips. So this has to be for devices that are going to be consumer based. Like it doesn't, it doesn't make any sense to have custom chips which. Custom chips require custom motherboards, custom motherboards require custom chassis like, and it's, it's a strange situation to do that when you don't actually own any actual hardware. 

Marlon Avery: It's interesting. 

Sekou Doumbouya: Right. 

Adrian Green: Well that changes my whole thoughts about them making chips because I thought that it was to go into their existing data centers. But are they backed by Amazon? Is it Amazon or. No, Microsoft, Microsoft on the back end. 

Sekou Doumbouya: Like why are they making, why, why is Microsoft not making chip. Well, Microsoft is, they're, they are doing investments in chips and things like that but the fact that OpenAI is trying to build their own like I am a little puzzled about where, where the ship is going to go and what form factor. 

Marlon Avery: Yeah, interesting. Yeah it definitely has to go made like an underbind device but also too that is a, I would imagine such a, such an expensive roadmap in this area of developing your chip with the having the understanding of the proper roi. It's going to have to be something for your everyday consumer. 

Sekou Doumbouya: Oh, I just saw that it is an inference chip. So yeah. 

Marlon Avery: Yeah it's going to have to be something for your everyday consumer. 

Sekou Doumbouya: Yeah. 

Adrian Green: Now the significance of it being an inference chip is what is that when. 

Sekou Doumbouya: You want to run chat GPT and ask a question, you're doing inference as opposed to training. Inference is like every, that's like, you know, you, that's CPUs are actually really good for that as opposed to GPUs really excel at the training aspect of it. So if they're making a custom chip that's for inference, more likely it's connected to a particular device. 

Adrian Green: Okay. 

Sekou Doumbouya: That they're trying to like that they're thinking of. We don't know what the device is but it's definitely something that may not go in a data center because inference isn't is know it's a challenge but it's not the same challenge as the GPU challenges out there. And power consumption is very different. 

Adrian Green: Okay. 

Marlon Avery: AKA Rabbit counter days anyway Adrian first. 

Adrian Green: Well that's what I mean. My, my mind went immediately to Rabbit because it's what they this is a device, something that's going to be a device that chips are going to go into. That's exactly what we would want Rabbit to be. You know, something that has an LLM that lives on the device and it's super smart and super fast. You know, when I initially heard it about them creating their own chips, I thought that it was, you know, really just gobbling up your supply chain. So you know, if you're a newspaper company, it would behoove you to own the forest, you know, that the paper comes from. So I thought it was just typical, you know, typical, typical business in America kind of methodology there. But now after talking, after hearing that, you know, it has, it really does have broader implications for sure. 

Marlon Avery: I'm going to make a prediction myself here. I, I, yeah I'm a prediction on demand device. Maybe this is just the, the wish in me but I'm going to make prediction that this, they will launch a AIPN type device that they they're going to, they're going to integrate it though. I mean they have, they're gonna, they're partnering with, you know, Johnny Ivy which knows, you know very well and deeply how the iPhone works. And they already have, you know, certain parts partnerships and you know, landscape stuff that I'm going to argue, I'm going to say they're going to make a device that will be a, will probably going to lead the market on an on demand device that you can have access to. I mean it's the, the reason you, you partner with the, you know, tonight. I mean his, his mind on design side is just you know, next level, you know. But everything, a lot of things they're designed was also included in removing steps for the consumer. You know, stuff. And so you know, the, the famous quote of a thousand songs in your pocket. You know, we remove steps from, you know, CD player to record player to tape player. They said that like what it might be to be able to go, you know, listen to songs and everything. And so I would say there the functionality and the design and stuff of it. What they build will remove the step of having to go to your computer or your phone to access chat GPT. It would be a wearable of some, some sort. It would be a device glasses, heads, something earphone like it'd be something around that, that remove the step of the need to log in Internet, you know, almost things like that to get access to chat GPT and you will just be able to button clip. Hey, you know, hey Doug, whatever you, you know name your AI assistant like whatever it be and it will remove that, that process. And I think that is why you develop the chips that's also too while you partner with Johnny Ivy. And you know, you when you do that because then you don't have to go build like Sickle said, you know, you don't have to necessarily go build your own, like your own cloud infrastructure, but your own data center. If you're going to do the chip, anything that will handle the infrastructure that will live inside of a local on demand device. 

Adrian Green: Just a fair warning too. I just want to put out to the viewers when this happens. Be sure to tell your, whatever you name it say Doug, your GPT AI on device intelligence, that's on you. Tell it not to snitch because there's going to be some ransoms. Hear me out. It's going to be some ransoms where people are going to be like hey, look what we got. Hey, this is Doug. They kidnapped me. Oh my God. What am I gonna do. You gotta. You don't want that to happen. If someone gets their hands on your. On your device, you got to tell it. Hey, look, don't you know, it's got your social. It's got everything in there, you know. 

Marlon Avery: You know, snitches get unplugged. 

Adrian Green: Get unplugged. Wow. Yeah. 

Sekou Doumbouya: Add that into my. My prompts for now. I just have that somewhere. 

Marlon Avery: Look, that's. That's gonna be. That's gonna be a T shirt for Optimus snitches getting a plug. 

Adrian Green: Yeah, we need to adapt like, maybe adapt like Biggie's commandments into just like a kind of GPT commandments. 

Marlon Avery: Oh, man, the. The AI rules here from the culture. Oh, man. 

Adrian Green: Give it the tablets. Yeah. 

Marlon Avery: That is hilarious. There we go. Really? Look. KK agrees. Snitches get unplugged. You know, coming to a shirt near you. We will be introducing our store account separated here soon. Snitches get unplugged. AI with friends. 

Sekou Doumbouya: I love it. 

Marlon Avery: Oh, man, that's hilarious. Okay. Well, guys, man, this has been unique, you know, so for us, you know, we have, you know, shifted and everything. We have some technical issues stuff. Typically we run our live Stream every Wednesday, 9pm and we had to shift everything. We did the morning schedule because also, man, we didn't want to miss you guys, you know, want to make sure we still deliver, you know, the content stuff here. You know, who snitches get unplugged would have never been birth, you know, if it wasn't, you know, you guys, you know, so. Yeah, but man, we had the privilege, you know, and stuff over here, you know, discussing, you know, everything from Elon's, you know, is coming in with the thousand GPUs per cluster, you know, OpenAI and developing their first custom chip, you know, from broadband. Waymo has now also to, you know, introduce their 5.6 billion dollar expansion particularly into new markets like Austin and Atlanta. Then also too meta. They've also to introduce a new free AI tool as they continuously to kind of like get a quick jabs to the industry with endosing, you know, make sure they put free tools stuff out there. And they introduce Notebook. No, no, Notebook Llama, which is competitor to. Nope. Google's Notebook LLM. And so yeah, man, I think it was. It was a great episode, you know, fellas, man, what'd you guys think? 

Adrian Green: I thought it was one. It's one of my favorites for sure. 

Sekou Doumbouya: Yeah, yeah, definitely, definitely. It was very quotable. I'm already making a shirt right now, so. 

Marlon Avery: Oh, no, I'm so serious about that, dude. 

Sekou Doumbouya: Like, it's happening. 

Marlon Avery: Yeah, the graphics. 

Adrian Green: Like, what's. What's going to be the graphic that's going to go with it? 

Marlon Avery: You know, I'm. I'm so serious and stuff with that, we're going to set up a printful account, you know, and stuff of it. Stitches. Get unplugged. We'll figure out the designs. Yeah, guys, stay tuned and stuff with us. And this is going to be the launch here of power, man. That's hilarious. And so, yeah, okay, stay tuned so through as well. You know, we'll get you. Get you a shirt and stuff there. And so, yeah, man, we're gonna. We're gonna start quoting, you know, because, you know, we're going to sort of have some of the industry people buy it out poses. So we're gonna start quoting a lot of our stuffs and. And branded and putting on shirts and stuff here, know, and so. So, yeah, man. Snitch is getting unplugged from AI With Friends. All right, well, guys, man, Adrian, give us a. Give us a. 

Sekou Doumbouya: You. 

Marlon Avery: You'll be closing out and stuff here. 

Adrian Green: Guys, I don't typically. Don't close out. You gonna throw the ball to me, Marlon? 

Marlon Avery: All right, man. 

Adrian Green: Guys, we had fun. It's been great. I'll see you next time. 

Marlon Avery: There it is. All right, guys. 

Adrian Green: All right, bye.