Skip to content
PodSubs
ShowsPeopleGuestsJobsNewsEvents
More
BlogPodCred
Menu
ShowsPeopleGuestsJobsNewsEventsBlogPodCredNetworksSign in
Join nowSign in
  • Shows
  • People
  • Jobs
  • News
  • More
    • Guests
    • Events
    • Blog
    • PodCred
    • Networks
    Join nowSign in

Tik & Chris

Tik & Chris artwork

Nobody Knows What AGI Is

Sep 12, 202644 minAudio
ShareXLinkedInFacebookWhatsAppEmail

No guest this week, which means Tik and Chris had all week to argue about whether this episode should exist at all. Here's the problem. Nvidia's CEO says AGI has arrived. A researcher walked out of Anthropic the same week saying we're not ready for it. And nobody - not one person in this entire industry - can tell you what AGI actually is. There's no definition. There's no test. There's no committee. The closest thing to an official one is buried in a contract between OpenAI and Microsoft, where AGI gets declared by hitting a profit number. That's it. That's the bar. Pratik does better in about four minutes: it's AGI when a model can go get the skills it doesn't have and hand back work the experts in that field would actually accept. He runs it against mathematics, where a model just went at an unsolved problem with 10,000 agents, and sits honestly with what that means for the people who gave their whole careers to it. Chris comes at it from the other side - what we allow, and how - by way of nuclear power, the printing press, and railroad doctors who thought going that fast would cause mass hysteria. It gets philosophical. It gets unexpectedly personal. Chris clarifies twice, unprompted, that he is not a madman. P.S. - Chris also walks through how they use AI Studio to power the website for the podcast. Chapters 0:00 Intro 4:33 Building the show's website by talking to Google AI Studio 13:21 Shopify drops React Native, and the case for disposable software 16:43 Nvidia says AGI is here 19:37 Nobody can define AGI, so a contract defined it with money 25:52 Navier-Stokes and what it cost the mathematicians 31:24 Math put us on the moon, and might get us cancer treatment 36:52 Nuclear power, the printing press, and every panic before this one Send us Fan Mail

Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.

Transcript

From the show. Tap a time to play from there.

SPEAKER_00What is AGI? What's the definition of AGI? And how do we know if it's here? Who decides if it's here? Who decides what the definition is? And that's up for debate right now.

SPEAKER_02Please welcome your starting lineup for the Teak and Chris podcast. It's Christopher Long and Pratique Desai.

SPEAKER_03Hello and welcome everyone to episode 13 of the Teak and Chris Podcast. My name is Christopher Long, and on the other side of the internet is Pratique de Sai.

SPEAKER_00Hello, everyone.

SPEAKER_03Hey Pratique. How you doing, man?

SPEAKER_00I'm good. I'm good. I'm uh I'm actually getting ready for um, or I have been working on Aria's first week of school this week. So that has been um interesting to say the least. I I'm a new father, so first time, and uh I don't think anyone warned me pro appropriately, I will say. Uh what's everyone surprising? Uh just the drop-offs. The drop-offs are rough. Yeah, they are rough, and then you spend the next hour just like wondering if you're doing the right thing and trying to figure out if uh you just need to quit everything in life and drop it and just to be with your daughter the entire time.

SPEAKER_01Which yeah.

SPEAKER_03Have you ever seen it's an old movie? I think it's in the 80s, Mr. Mom.

SPEAKER_00Uh I've heard of the movie, I've never watched it. So Yeah, yeah. Go watch it. Okay. I will.

SPEAKER_03You might relate now.

SPEAKER_00Probably. Probably. It's uh what from what I've heard about the movie, it's it's exactly it's exactly right. And um, but anyway, I'll get through it. Okay, I get through it, Chris. I think she's probably getting through it much faster than I am.

SPEAKER_03Oh yeah, she'll be fine. You're the one who needs help. That's right. So for everyone out there, help. Um so before we get into uh, you know, the real uh the real meat, if you will, of the podcast. Is there anything you want to like what anything top of mind? Just fun.

SPEAKER_00Um you know, I think uh a few things that I've seen lately out on the the the internet of folks that were really not using AI and all of a sudden have started using AI. Some of the use cases that they've been sharing have been fascinating, things I haven't even thought of. So um just just one anecdote. Um someone came to me recently who I never thought would have used AI, and they're like, hey, you know, I I've had this this struggle of managing my calendar with my husband.

SPEAKER_00And um, you know, did you know you can connect it with Gemini? Yeah, no, I knew that. Um, okay, well, then now I create a scheduled agent that not only manages and tells me what I've got going on for the week, but then it makes recommendations on what we can fill in things with. So um, you know, we live here in Syracuse. What's going on in Syracuse? Go go look for events and prioritize family time and prioritize like what's important. And here's a location where they're throwing a festival.

SPEAKER_00You can prioritize family time, it's within driving distance, and yada yada. And she's essentially using it to maximize her time with her family. And I like obviously that seems like an obvious use case, and I've I've thought about doing that, and yada yada yada, but she's essentially offloaded that task, and so on a weekend they're like, what's going on? We're just gonna do the thing that Chat GPT says to do, and they're just focused on family, and I just think that's that's a lovely, lovely little use case there.

SPEAKER_03That's great.

SPEAKER_00Like it's becoming like a social organizer, right? Yeah, I mean, social organizer and uh you know, back to my point, Arya's in school now, like optimizing the time you have with your kids. Like, what what could be more important, right?

SPEAKER_03So yeah, I think that is gonna be one of the largest use cases based on where the harnesses are headed. Um, yeah, and I think really I think we've already mentioned this, like harnesses are gonna be a big deal, they are a big deal. So, like, we need to have an episode about harnesses, but that is not this episode. Lucky number episode 13 is uh going to be about AGI, not because we were planning on talking about AGI, but because someone said, hey, maybe AGI is here. Um, but that's kind of a big topic.

SPEAKER_03So we're gonna warm up a little bit. And before we talk about AGI, we're gonna just talk about um how we use AI a little bit for the podcast. Okay, we're just gonna warm up. So I'm gonna share my screen. I'm gonna talk us through it.

SPEAKER_00Yeah, and uh as you're doing that, by the way, I wanna mention that when we first started this, you and I had this hot debate about Claude Artifacts versus Gemini Studio. And uh you you were pushing back on Gemini Studio being being like this tool that could be really producized and put into production. And uh I say that because what is this built on?

SPEAKER_03This is built on Google AI Studio. Um, or I think I don't Gemini Google interchangeable, but AI Studio. If you're Googling it, type in AI Studio and you'll find it. Um it I mean I come from a heavy development career, right? Like that's that's how I learned tech was making things in tech, um, which meant code, meant building. And there's still a part of me that really likes to uh know what's happening behind the scenes.

SPEAKER_03Um and so I started off using um uh what was the old brand? It was Fire Firebase. Firebase. Yeah. So I started off using Firebase on your recommendation. Um, so I could learn Google Cloud, and just because like they offer all of these integrations that are easy to snap in. Um, but you still were hands-on, you could code whatever you wanted. You had a lot of you had you actually had to uh make a lot of choices and configure a lot of things.

SPEAKER_03Okay. Well, I got the message hey, uh, that's going away. We're getting rid of that, or it's evolving into Google AI Studio. So if you want to keep using our stuff, you're gonna have to learn the new stuff. So that's what we did. So for the podcast, we have a website, teac and Chris Podcast.com, and we built a site on Google AI Studio. So this is the site, it's got our logo real big up front. There's some Easter eggs on the side if you scroll over things. We've got this box where people can enter in things they want us to talk about on the podcast.

SPEAKER_03Down towards the bottom, the latest episode of uh the podcast on YouTube shows up, and then all of our other channels you can get to down here. So that's one thing, one asset or aspect of the website. The other one is uh we have a dashboard. So when people do type in the box and let us know what they want us to talk about, we have a dashboard we can look at, and that dashboard is protected by um Google Auth or Firebase Auth. So this is a graphic kind of talking about the architecture.

SPEAKER_03Um so you start with building in Google AI Studio, and like it's using React and Tailwind CSS and some things, it's using Framer Motion for animation. Uh, but I didn't make any of those decisions or choices to use those things. All I did was I went into AI Studio, I said, this is what I want to build. Um, and you know, go for it. You build it. And these are the choices that AI Studio made. It built it. Um, and it's actually done a nice job and it's been super easy.

SPEAKER_03The one kind of connective tissue is uh GitHub. So we have a repo on GitHub for the site, and Google AI Studio, whenever you know I make a change, it will publish to GitHub. The site is not hosted by Google AI Studio, it's hosted on Google Cloud Run, okay, another Google product. But um Google Cloud Run, you can connect also to GitHub, and it will automatically deploy your site if you want it to. You can set that up to the web whenever something gets pushed to GitHub.

SPEAKER_03It's super awesome. So the result is I just talk literally to Google AI Studio, like there's a mic if you want to. I can say, hey, change this. Um, for example, for the weekend I added our Buzz Sprout um token or button or you know, call out on our page. And I just said, here's our URL, add this to the bottom, make it look like the other buttons, and it did it. And then I said, you know, publish to GitHub, and then I don't know, a few minutes later it was live out on the site.

SPEAKER_03I didn't have to worry about it. Um, the other thing I like about Google AI Studio is they still have the Firebase integrations among some other integrations. So, like Firestore and authentication. So if you need a database essentially, and you need someone to be able to log in, that's just kind of covered. So I just say what I need and it will use those integrations and make that happen.

SPEAKER_00Um well, and and I I think I think that's the point there, right? Is that while we're showing an architecture and for the folks listening, it's quite a pretty sound architecture for the technical folks. Um you didn't need to dream this up. You didn't need to tell it to do this, you didn't need it to say, do it this way or use these integrations or particular patterns. You just said what you needed, which which was the actual business need. Um you know, a landing page that demonstrates our podcast, puts an episode, does these icons for these links, and then a dashboard where we can you know see the the results.

SPEAKER_00It decided the architecture. This is what it decided, which makes a lot of sense for us. And and not only did it decide this, it then tapped into tools in the greater Google ecosphere, uh, which is Firebase, uh Firestore, and authentication, to be able to do the additional things that you're asking for, which is security and databasing. Um which you didn't you didn't say do security and do databasing. You just said I need a place to access the results, which obviously then requires a database and it knew that.

SPEAKER_00I think I did needed security.

SPEAKER_03I was gonna say, I think I did, like, I'm not, you know, insane. I think I did use the word securely in my request. I'd like to be able to securely log in, but I didn't specify like, you know, for it to test the website for you know security flaws and things like that. But you know, I'm not a madman. Like I do want to keep people out of things where I that I want to keep private. But to your point, it just decides. It just does it. Um it just does it. Just does it. And you know what?

SPEAKER_00I think and I think that that like we we you know, I kind of gave you a little bit of grief there. Um, but we're we're splitting the personas, right? Like you want to see what's behind the hood. And I think that's largely driven by the fact that you know what you're looking for when you look behind it. Like, you know, if if I asked my my wife, you want to look what's behind the hood on our car, she would be like, No, that seems pointless because I don't know what the hell I'm looking at at that point, right? Um, like you if you know what you're looking at, then yeah, everyone wants to look behind the hood.

SPEAKER_00That makes sense because you know what you're looking for, and and you have this probably lack of trust that you can then exercise. Well, if you don't have the lack of trust, well, if you have the lack of trust, but you can't exercise it, there's no point in looking behind it. If you don't have it, then who cares? Um, in this circumstance with AI Studio, I think the reason why I like it so much is that you can, if you really want to, you can look behind the hood, but it's not the point. They have built an experience that doesn't necessarily center around looking behind the hood.

SPEAKER_00They've built an experience that centers around the lack of ability to know what you're looking at if you did. And that that's a different persona and that's a different thing they're building for, but it democratizes this pattern into a mass population that needs to be able to do this too.

SPEAKER_03The I'm starting to feel a little bit of disconnection um between my desire to know what it's doing and the output. Like I'm already starting to adjust to this paradigm because I didn't care or even look to change or to see what it was using, the model that it was using to do all this. Like I did, like it was just too easy for me just to get in there and start. And I started and I started seeing results, and I didn't like I was just like, oh, all right, this is working.

SPEAKER_03I don't I don't care. I don't care what model it's using, which is way different from the way that I build things um with ChatGPT or with Claude.

SPEAKER_00So I I I love that, right? Because I think it shows the the level of trust that we've moved the needle on in the last eight months. Um there was a post actually today from a you know, presumably director of engineering, head of mobile at Shopify, where he said the world that required React Native, and for the folks that don't know mobile technology, uh React Native was a way to write the code once for mobile apps and then deploy it to Apple and Android.

SPEAKER_00Now, Apple and Android have different architectures, different infrastructures, and required different sets of code. So React Native was this Facebook answer, so you only have to write it once and then deploy it everywhere. It diff it definitely didn't have all the feature sets and all that. Like there was a diminishing return on writing it once and deploying it everywhere. Shopify just announced today like that need for consolidation to write code once is gone. Why? Because they can literally write it once in Apple and do all the Apple features and then hand it to an AI and say write the Android version, and it will do it with such efficacy that they called React Native a 2010 sorry, 2020 type project or try type solution, and in 2026, that need is gone.

SPEAKER_00They're still writing it once, and then they hit a button, AI writes it the second time for Android, and they get the same efficacy. And so now you have to presume that the the level of trust that would be required to not have to review the code for Android, would that have been a solution in 2025? I don't think so. I think the level of trust you just told us about and the level of trust he's demonstrating by making that public announcement is massive. We've come so far. And so I I think, yeah, and and so I I think this, you know, this concept of vibe coding um is move is really moving from uh vibe coding as a as a as a fun thing, fun projects, to I think what we're gonna find is these solutions that are gonna pop up.

SPEAKER_00And I was just talking to someone today about this concept of, you know, historically we've had general purpose applications and general purpose SaaS platforms. I think we're gonna end up in situational platforms where a team somewhere just has a need and they just go, this need is never gonna exist in three months, but we'll just we'll just do it, we'll just build it because it's just easier than trying to solve this need operationally. Let's just manifest an application, we'll use it once, we'll throw it away, and we'll move on.

SPEAKER_00Um so that situational software is gonna start to become the way that we operate business. Um, and and it was never possible before.

SPEAKER_03I mean, it's a new window already, which is crazy. Um, and I don't think that window will last like too long. Like I think we'll go from situational uh to just a singular AI running your business um in the not too distant future. And it's it's gonna figure it out. And that actually is a great segue into what we need to do. Perfect segue, which is AGI. Um, would you like to kind of set the stage for that for that conversation?

SPEAKER_00Yeah, so we've talked a lot about um you know the CEO of Nvidia, um Jensen Wong, for the folks that don't don't know don't know the name, right? The the CEO um has been has said this a couple times and and then he makes waves every time he says, and look, his his job is to make waves, right? As the CEO of a publicly traded company that um you know builds these chips that have to continue to be you know defensed um against competitors and have to stay at the forefront, like he's going to make waves, he's going to say things that generate hype that then cause you know little podcasts like us to talk about it.

SPEAKER_00So we know that we're falling into that trap. Now, knowing that we're falling into that trap and actually dissecting is two different things, let's dissect it. So he he you know he said something along the lines at um at a conference that AGI is here and he was referencing you know Astra, the the new the new model out of ChatGPT. By the way, if you have access to it, try it. It is phenomenal. Um I have now moved all of my uh usage, daily usage from Claude to Chat GPT as a result. So I will go out on a limb and say it is it is that good.

SPEAKER_00I've moved so and I probably in six months I'll move somewhere else because that's just how life goes, right? Um so so I get why he's saying that in terms of efficacy and consistency and and feel of the model. Um, but then it caused waves, right? Because at the same time, there's other things happening where there's fear around AGI, right? Um there's a there's a moment that happened this week where a researcher from Anthropic actually left uh after three years of being there.

SPEAKER_00He's a junior researcher, and he said, We're not properly deploying this type of technology as we experiment and research it um and and it's going to cause our demise, right? Something along that line. That was a public comment that we're gonna do. We being anthropic and subsequently humanity. Yeah. Right? He's referencing the work that he did at Enthropic. Right, but like And what was weird is they didn't come out and say, Eliminate humanity.

SPEAKER_00Elimin eliminate humanity. Okay. And what was weird is there was no public statement that Anthropic is famous for against this. Um in fact, what happened was his senior researcher actually came onto the thread and responded and said, Ha, yeah, kind of. And also we put the statistics around 10% probability. Right. And so, you know, obviously this is causing waves, and while while Jensen Wong is saying AGI is here, and then you know, we're being told it's gonna, you know, 10% probability it's gonna cause the human demise, like we're talking bronze age level collapse type stuff.

SPEAKER_00Um, there's a lot of fear around this, right? And and I think I think what's interesting is um the first thing is what is AGI, right? Like what's the definition of AGI? Um, and how do we know if it's here? Who decides if it's here, who decides what the definition is? And and that that's that's it up for debate right now. Like we we can't decide that. We don't know who decides that. We don't even have a definite definition. Um I'll pause there, Chris. Like I one, I want your take on what I just said, but two, what's your take on the fact that there isn't even a definition of this thing that we're saying is here?

SPEAKER_03I think my take is about the definition. Uh the up to now, when anybody says AGI, at least in my head, it's always been like a sentient entity. Like something that recognizes that exists, um it has it displays somehow um feelings and emotions over things. Um it can make choices, it can, it can execute on those choices, but those choices are you know driven by you know something a little more complex than just instructions.

SPEAKER_03Um so like there's this idea of really like I'm trying to think of of a good example, but which I can't, right? Because I don't think we've ever seen it. Um but of something that you really feel is alive. So I guess that's the thing is up to now AGI equal sign life. Um but uh that's not gonna be it. Like playing around with I I did get a chance to play some with Astra and really like looking at what other people are doing with it and how Astra works.

SPEAKER_03And I think I mean, really, I think Fable um from Anthropic is is similar. Um but I you you feel it a little more, it seems like with with Astra, is that it really feels like like I I can give no instructions, I just give a goal, I give an outcome to it, and it just goes and it lives its life and figures out how to reach that goal.

SPEAKER_03And when you see like everything that it's thinking about and and like how long it can operate. Operate. Like this isn't a oh, if it can't get it done in five minutes, it quits. It will just run and run and run and do things. Um and just like it feels like I gave a task to a person. Right? Um, so I don't want to say like it is AGI, but like that feeling of just like it's its own thing, and I don't have to uh give it all of the context and information uh and prompting that I used to give it, like it can now just figure it out on its own.

SPEAKER_03It's kind of incredible. It it did tear through my usage limits. Like, um, so it's not like something, at least right now, that like most of us can just take advantage of. But like I could feel it a little bit, like, oh wow, this thing is kind of amazing.

SPEAKER_00Yeah, let's let's pause on the commercial viability of these highly, highly intelligent models, right? We've talked we've talked about that ad nauseum. And I think um I think the fact that there is no definition is is the key point there, right? And I I th I like I like where you're going, or you know, life. And don't forget, life is a contested defined term, right? Like we we we define life on earth as one thing, and and then all of a sudden we we find some other form of life in the deep sea, and then all of a sudden our definition of life changes.

SPEAKER_00So um, and I'm sure that'll change if we start to explore the stars and and yada yada yada. So life is a di is a contested term. Um I think I think you touched on something that I want to pull out because if you give it a goal and it doesn't necessarily have the skill sets to do those things that require it to meet the goal, and then it somehow figures out a way to acquire the skill sets. To me, I think that's the line. Right? That that like with zero incentive on making this definition, and I'm gonna touch on why that's important in a second, with zero definition, I think that's the line.

SPEAKER_00For me, that's like that's as good as we're gonna get in terms of finding the line between not AGI and AGI. Um, because if it can go out and get those skill sets, I think we're we're generally intelligent, right? Generally intelligent in that I've got a general model that has general intelligence that I can give any goal to, and it will go find a way to get the skill sets. Might not be commercially viable, that's a different thing, but it will find a way. Um to me, that's the line, and and produce an output that would be accepted by a population trained in that domain, right?

SPEAKER_00That like that's a key point as well. It can't just get skill sets, produce an outcome, and no one in that domain would accept that as an outcome. Right. Right. So so I think that like to me, that's the line. And the reason I stressed on the fact that I have zero incentive on defining that is that incentives is what's driving people on their definition of AGI, because the the first definition of AGI that I've seen structurally and contractually was between OpenAI and Microsoft in their arrangement where they famously tied AGI to profit thresholds and a board declaration.

SPEAKER_00Like it's an entirely commercial trigger versus a scientific definition, right? Um, and so like who's incentivized to define AGI and why and what's their incentive to define it? I think that's it, that's a big part here. And we know what N Jensen Wong's def you know incentives are. He wants to sell more chips. Uh, if he wants to sell more chips, what does he need to say about any model and all models so that he can continue to sell that shovel, right? Um, I think I think there's a certain thing that drives him on making those definitions.

SPEAKER_00Um, one one thing that's been really interesting, Chris, is recently there's been all this talk about um so as we talk about solving problems in a area where the domain will accept the the outcome. Let's talk about mathematics, right? There's all this talk about models solving math problems. I think the the most recent one that's being talked about is the Navier-Stokes problem. Um and for the for the audience, like that's just a problem that mathematicians have put, I think like a million-dollar prize behind solving this problem, because if you solve it, you've cracked a problem that hasn't been solved ever.

SPEAKER_00Um and all of a sudden, models are starting to solve problems like this. And so the question becomes you know, is that AGI? Based on possibly the definition I just put forward, which is a generally intelligent model being able to gather skill sets to take a problem, solve it, with an outcome that the domain would generally accept. It's happening in math, right? So is that AGI? I don't know. Could it could it be that you can't only do it in one domain?

SPEAKER_00Because then that's just a domain-specific LLM, and that's just training, and that's where it doesn't meet the definition of AGI because it needs to be able to do it in math and in biology and in here and in here, regardless of being trained in all those things. Like maybe that's the difference between what we're seeing in mathematics and AGI. Um, but it's still solving math problems where the domain is accepting it, and that's that's a huge thing. So, Chris, I just want to get your take on that.

SPEAKER_03So I think the measurement of of how to consider something not alive as AGI. I we just tend to come up with those judgments as comparisons to what humans do. Um, and we don't have a human who is um all good in all areas. So I I don't think that's gonna be the requirement uh for um the majority of people to say something is H E I or not.

SPEAKER_00Um I I think that's fair. Yeah.

SPEAKER_03Yeah. I think it's gonna be more about um how it arrives and behaves at its goals.

unknownYeah.

SPEAKER_03Um which like it's very rapidly starting to feel like uh it's on a slope that we as humans we're not gonna be able to catch up on. Like that's gonna be a problem too. And uh that's a I don't want to open up another topic, but like um things are gonna be so smart, so smart, um, we won't need to solve complicated problems. Um so then what do we do?

SPEAKER_03Um, but again, that's that's a topic for for another time. So um considering that these models are getting to a place where it's solving problems we are taking forever to solve ourselves, and it's doing it quickly, um, with the assumption that it's going to be able to do that in other fields over time.

SPEAKER_01Yep.

SPEAKER_03That there were some concerning aspects of that. Talk us through like why we were debating on even talking about AGI this week.

SPEAKER_00You you and I have had a had a few series of debates this week on like should we talk about AGI? Because it's not it's not really a thing that we can define. So if we can't define it and we're just feeling it, and then there's just a bunch of fear, like what fruitful conversation is gonna come out of this? Um, so we've had a bunch of debates. And then we talked about like, do we even talk about the Navier Stokes problem? Because it's controversial. And well, I think we have to. And why is it controversial? Let's address that. It's controversial because there's mathematicians that have spent their entire career, entire careers, trying to solve this problem to win that contest, right?

SPEAKER_00Um and then and then a model that's trained on the very thing that they've been putting out as proofs and as learnings as they have failed to solve the problem and tried to solve the problem, like the model took those things and all the learnings from math prior and then was was just able to outwork them. Was able to outwork them, was able to outthink them, was able to out to like just you know, I think there was one instance where they spun up like 10,000 agents.

SPEAKER_00I think that was the Navier Stokes problem, actually. Like, how does a mathematician compete with that? And and it's controversial because like I feel for that mathematician. I really do. I really do, I really do, and I and I don't want to undersell how much was taken away by a model just being able to move forward with this, right? I I don't know what to say to them. I I really don't. You know, and and this is this is something that we've addressed time and time again. It's it's it's also what you just punted there. Like, what is left for humans in a space where these models can just do it and and can just solve these problems and go out there and move science and move mathematics.

SPEAKER_00I don't know. I think we're gonna reserve that for another episode. Um, and I feel for that mathematician, I don't know what to say to them. But there is a little bit of excitement here, and I want to I want to acknowledge and I want to be transparent on that. The excitement comes from like, these are mathematics problems that haven't been solved. Remember, math put us on the moon. Right? Math is what allows us to fly every day. It's what allows us to drive to the next place every day. Math is what allows us to build houses that don't crumble.

SPEAKER_00Math is what allows us to do the things we do on a daily basis that we quite frankly sometimes take for granted because we forget it's math. And and the problems that were solved in math to allow us to have the house, to allow us to fly, to allow us to be on the moon. The next problems are gonna allow us to do the next things. And what excites me is well again, while I feel for the mathematicians, what excites me is where are we gonna be able to go and how accelerated are we gonna be able to get there if we continue to solve problems?

SPEAKER_00And look, then one near and dear to me is cancer, right? Now I I realize I'm conflating mathematics with biology a little bit here, but don't forget, we're solving biology problems too. And a lot of chemistry and a lot of math, like this all comes together. It's all science, right? Um if if I told a cancer patient that yes, there's a ton of researchers at a pharmaceutical company working on a problem, so we're not gonna use a model to try to expedite finding remedies, like that cancer patient would be pissed, would be furious, right?

SPEAKER_00And remember, this is about patient-centric care. This is about quality of life, whether it's flying and getting to the moon or the next thing that we're gonna do or solving remedies in cancer, like think about the problems we're gonna be able to solve. Not Navier Stokes problems, not mathematics problems, but life problems that we're gonna be able to bring to the table and actually move the needle on humanity because we're solving these problems that we have not had enough resources, time, materials, money, whatever, to get over.

SPEAKER_00And AI is doing it. Now, Anthropic is probably spending about a trillion dollars trying to do it, but let them spend the money. Let's solve cancer, right? Like, let's go, let's do that. And so, look, I feel, I feel for the mathematicians who've dedicated their career. I feel for the college students who are like, what the hell am I gonna do? Um, like what is left for humans? I feel for all of them. Um, but I I I wanna I want to counter out counterbalance that with excitement that we're gonna solve problems that we like systematically and structurally put these blockers in place to be able to solve.

SPEAKER_00Um and and and there's there's trade-offs, we have to address them. Um we only have so much time on these podcasts, so we're gonna address all of them. But right now, I'm I'm gonna stick to excitement. I'm I'm excited because I I'm excited for what comes out of us solving these problems practically.

SPEAKER_03I I'm on the same page. Um I've always been um pro-tech. Like I started doing things when I was very young. I think I was 10 years old when I and I'm I'm older than 10 now. Um, but I was 10 years old when I started coding. So like I've just I've always had tech around me. Um one of my grandfathers uh was a codebreaker um for the US military, and just that's something we need to talk about later.

SPEAKER_00That's pretty cool.

SPEAKER_03Yeah, it's pretty cool. But like because of that, he he was just a tech guru. Like he always had everything um before anybody else. Um so I just grew up around all this stuff, and it was always interesting, but more importantly, I always found ways to make it useful and helpful. Um and there is a lot of talk around like, do we need to stop with all of the AI shenanigans? Um, do we need guardrails?

SPEAKER_03Of course we need guardrails. Um, but the well, and again, like especially this week, like someone coming from one of these giant companies working on this is afraid. Like they left to like let everyone know, hey, I'm not, I'm not sure we're doing it right. Um that's all super valid. Um, but what seems to be happening is people are choosing a side, and they're either for AI or they're against AI.

SPEAKER_03Um, and I'd like to put it on the record, and I said something very similar um on my friend's podcast, shout out Chasing the Signal, which will be released soon. I'll I'll promote it more later. Um, but this idea that maybe you can be pro-human and pro-tech and pro-AI. Yeah, maybe those two things can live together. Um and maybe if we had more folks with that same alignment, having these conversations um and putting pressure on the forces that be.

SPEAKER_03Um, and a lot of these folks work at the forces that be. Right? Um, like the only way it gets better is if we acknowledge what's happening and we talk about it. Um we have accepted so many things as humanity over the you know centuries that we've existed. Like we have all these nuclear things, nuclear power, nuclear weapons, um, and we just live with them. We all know how dangerous they are. We live with them. A lot of us are not happy about them, but they exist.

SPEAKER_03Um and like we've had to figure out how to coexist with these things. I uh it feels the same with AI. Like it's here. The only way to stop it would be um, you know, some kind of technological reset of the planet. Um, and I'm on the side of that not happening.

SPEAKER_00Yeah, I I think we're all on the side of that not happening. Okay. One thing we can we can all agree to, right? No, no Dune level um extinction events.

SPEAKER_03Yeah, I don't want to be mining spice. Um we're full of nerd references all of a sudden. Um just all of that to say. Like, I think we can't be empathetic and realistic at the same time. And I think we can have a vision or a hope that we as humans want what's best for everyone. And some of that is going to come from AI. So, how do we do our best to treat it with the kid gloves it deserves, but still use it to do things like cure cancer.

SPEAKER_00I I I love the I love that, and I love the nuclear weapons example because I think we can all agree that nuclear weapons could cause mass destruction of the human population. Right? And and so I I think it's it's it's it's an interesting thing because we live with it every day. We we live with that threat, we live with that fear, but we also live with the benefits that nuclear uh power has given us. And so, and I think I think about history, right? Um off the top of my head, like the printing press in the 15th century, religious leaders were worried about you know the ability to pass ideas through mass-produced books, right?

SPEAKER_00Right, railroad travel, like doctors and scientists thought moving that fast would cause like mass hysteria, basically. Um photography painters were worried about you know, our diminishing you know, love towards true art. Photography is true art now. So is art. That is historical. Like that, those things are both living in tandem. The telephone, television calculators, like there's these examples pop up, and I realize like AI feels so much more um, it feels more like an existential threat.

SPEAKER_00But nuclear actually then says, like, okay, well, that is also potentially an existential threat. So how do we live with it every day? And and then there's drones, and you know, we can we can go on and on, but the the point is like I love I love the the analogy because we live with it, we also benefit from it. We're excited about the benefits, despite the fact that no one actually attributes the benefits we get to nuclear power because we just don't know. We don't think about it, right? I think we're gonna find the same with AI. AI is going to solve things and has already solved things that has made our life better.

SPEAKER_00If we don't know it's AI, we accept those things every day with an open arm. Yeah. And I'm I'm sure those examples are gonna be more and more. Now, I feel for that mathematician. I feel for those college kids, and I think we have we have to find a solution for them. There needs to be, you know, just and ethical job transitions. Just like with uh with everything that we've done historically, there needs to be those ethical job transitions. Um, that needs to be in AI as well. We need to find that. We need to harp on that, we need to, we need to have focus on that.

SPEAKER_00And you and I are gonna talk about that extensively in this podcast because we want to focus on it. Um, I feel for those uh companies that are gonna have to reorient. I feel for those people that are gonna have to re-orient in their career that have spent their career trying to solve something AI just did in seconds. I'm also excited about what it means to finally solve that problem. I'm excited about what comes out of it. So anyway, I've I feel like I'm uh I'm beating a dead horse because I'm just so passionate about this. Um, but I think that takes us towards like with AGI.

SPEAKER_00Does the definition matter? Does that term matter? Or just what we do with this technology matters? Is that kind of what we're getting at here?

SPEAKER_03I think ultimately that's always what matters, right? Is what is the outcome? What happens? Was it good? Was it helpful? Was it not? Um, so no, we don't uh well, we being maybe you and I, don't personally care about a proper definition and someone putting a flag in the sand saying this is the moment. Um and in reality, if you look at history, all of those kinds of moments are defined by generations later, right? So like I don't even think that's for our generation to decide.

SPEAKER_03What we have to decide is what are we gonna allow and how are we going to allow it? Um and that'll be our mission. That'll be our mission moving forward.

SPEAKER_00It I love that, and so let's define the mission. Just like with nuclear, we as a humanity know the threats that AI could pose on our society. It's up to us to decide what we do with that knowledge.

SPEAKER_03So uh But hey, you and I are just nerds. I mean, we're very cool nerds.

SPEAKER_00Well, we're cool nerds, yeah. Yeah, there's a spectrum.

SPEAKER_03Yeah, there's a spectrum.

SPEAKER_00Um nerds to really cool nerds.

SPEAKER_03Really cool nerds. We want to know what all the really cool nerds think. So if you are listening or watching this, uh let us know. Drop us a comment, uh, hit us up on our website. This is episode 13. We have other episodes. You can check them out on YouTube, you can check them out on Spotify, you can check them out on Apple Music. We also are reaching out now on Buzz Sprouts, some other podcasting platforms. If you have any ideas, please let us know. We would love to hear it. Like and subscribe. Such a fun thing to say.

SPEAKER_03And we hope uh this has been interesting. And again, we want to hear from you on this topic and other topics. So take care. Think about these things, um, or not. Just go have a fun weekend.

SPEAKER_00Have a good weekend. We'll see you next time. And uh I'm really looking forward to the comments on this one.

SPEAKER_03All right, take care, everyone. Bye.

SPEAKER_00Thank you.

SPEAKER_02That's the buzzer, folks. Thanks for listening to the Teak and Chris podcast. Do us a favor, like and subscribe.

Open the transcript file

Hosts and Guests

Nobody is credited on this episode yet.

Credit the people on this episode

Be the first to rate this episode

More episodes

All episodes
  1. Meet Meta Muse: an AI assistant with a soul

    Nov 8, 202634 minAudio

    Chris and Pratik compare notes after weeks with Meta’s Muse, from its single ongoing chat to memory and soul files. They share practical uses for calendars, email and ongoing goals, then discuss payments, credential...

    Credit the people on this episode

  2. AI Fatigue and the Case for an Assistant That Helps

    Oct 16, 202626 minAudio

    Is AI exhausting you as much as it’s exhausting us? Chris and Pratik are tired of keeping up with AI. Their solution to dealing with AI fatigue involves more AI, which they acknowledge is a snake eating its own tail....

    Credit the people on this episode

  3. AI Harnesses: Your Phone, Siri, and Meta's Muse

    Sep 22, 202658 minAudio

    Pratik polled an entire airport bar on his way to give a talk on AI ethics, and every single person told him they don't use AI. Chris and Pratik spend this one on harnesses: the thing wrapped around the model that...

    Credit the people on this episode

  4. AI Is Thirsty (Like, Really Thirsty)

    Aug 9, 202656 minAudio

    Tik and Chris finally do the thing they've been teasing - they bring on a guest! Craig Martin, a water treatment project engineer from central New York, joins the show to pull back the curtain on something the AI world...

    Credit the people on this episode

  5. What's That Worth? AI, Pricing, and Your Health

    Jul 25, 202636 minAudio

    Chris kicks things off with a story that'll either make you laugh or cringe: someone paid $100+ for an AI-generated song that you could make yourself for much less. From there, Tik and Chris pull that thread all the way...

    Credit the people on this episode