This is that. This is Brandon. Welcome to Concerning AI. This is the show where we are concerned with AI.
Wait, is that redundant? That's not redundant, is it? We're totally concerned about AI. Yeah, but definitely not.
That's why we call the show Concerning AI. And we're going to talk about artificial intelligence and how it makes us concerned. Oh, I'm so concerned. I'm concerned too.
I'm getting more and more concerned as we go. Yeah, especially because we keep, it's like a neural imprinting. We keep saying it over and over. Yeah.
You know what I'm also concerned about? I didn't tell you this during the pre-show. I'm just going to surprise you. Portland, Portland, Florence.
That's my soccer team, the Portland Women's Soccer Team and the professional team. They won the league today and they're heading into the playoffs next week. So we don't need to divert ourselves from AI too much, but that's my other passion is I'm following the team and they won the league today. So congratulations.
They couldn't have done it without me. They couldn't have done it without me because I am essential. Being in the stands or being on TV, watching on YouTube, you've really figured it out. I mean, they've had a great year.
And that's the great thing about sports is that I get to feel like I did something, which I always did. Well, there's some parallels with the whole AI safety effort. Yeah. So imagine how poorly it would be going if we weren't having the show.
We weren't super fans. That's right. At the end of last week's show, we asked people, hey, what are you doing? You know, what are we doing?
We're doing nothing. We're just having this show. We were doing literally nothing before we started the show. And now we're doing the show.
And you call it nothing. It's essentially nothing. Yeah. Well, it's the point 001, right, before it doubles.
Ooh. That's got some potential. Yeah. So we did get some answers to our question.
We asked people listening last week, hey, what are you doing about AI? AI safety? Are you doing anything that we could share with people? We got a couple of answers.
And it makes it seem like there's actually a couple people listening. So thank you. Yeah. We're going to call him Corey.
Well, we'll call him Corey. Corey writes very lengthy and eloquent. And so I'm just going to excerpt to the beginning of the end of what he writes. And then we'll talk about it.
Yes. And we're sure Corey is a he. I am not a hurricane. The picture is kind of a cartoon.
It's a cartoon. It might be male. So that doesn't really matter too much. No.
So here's what Corey said. Hey, thanks for adding me to the group. Big fan of what you guys are doing. And the question is you're asking.
That's why you read it. It's because you wanted to get that in. That's the extra prior. That's really all I need to do.
Thanks, Corey. Thanks, Corey. Thanks, Corey. That makes me feel good.
I believe one question you asked on the locality principle was what other people are doing to avert a possible AI apocalypse. I love that word, by the way. Is it AI apocalypse or is it AI apocalypse? I can't do it.
When I read it the first time, I just read it as apocalypse. But that was the wrong. It was just AI apocalypse. It was just AI apocalypse.
Anyway, thank you, Corey. Hi, apocalypse. I'm starting a company, Corey goes on, an entertainment venture. See, this is so interesting.
It's like, I'm so curious when I read it. It's like, what do you mean, an entertainment company? I'm starting a company, an entertainment venture with one driving purpose to create a fully realized virtual world, one with as many complex and varied entities as the real world itself. Modest?
Very modest, go there. Well, yeah, why make little plants? You know, I really did it kind of a deep dive. Well, not that deep, but pretty deep.
I kind of went to some of the sites that Corey linked to and got into it. It's really helping me to hear you read this again to sort of like, tie it all together because there's actually a lot there. And it's fascinating. Yeah.
Yeah. Okay. I'm going to skip down to the bottom of his conclusion. The idea behind my company, a company called WDDD, we'll put the links in the show notes, both to his comments here in the AI group.
I believe that's the answer to a world design document. World design document. Okay. So I'm going to just take a look at the website as well.
Cool. The idea behind my company, WDDD is this, create a story that people can invest themselves in, can incorporate into their everyday lives, can cherish and interact with, then leverage this engagement to turn the users into creators, into information gathering sensors for the virtual world that will one day be built on their efforts. This is so fascinating. It's very fascinating.
Start small with a story and links to a website WDDD.church, which is very provocative for so many reasons. So many reasons. Yeah. Right.
You know, the new top level domains that Brandon and I both know something about. I mean, I think the fact that it's church. I mean, that's really interesting. What is, what is this?
Yeah. What is it? I got a steeper. Yeah.
So it could be a story like this website or it could be a board game or an RPG or anything that might pique someone's interest and then expand that interest on mutually beneficial terms. Terribly vague. I know. Thanks, Corey.
That's, uh-huh. But I've written quite a bit already and should probably stop here. If you've had the patience to get this far, I'd love to hear what you think about the premise, execution critiques. Yeah.
So I got to say that I feel a lot like I did the first time I watched the matrix, which was an awesome feeling. Yeah. I'm headed down the rabbit hole and it's like a little crazy and a little weird as I'm looking at the graphics and the different things on some of the sites and the pages and trying to figure out what's going on. I feel like I'm kind of on a little bit of a quest.
There's all these cryptic, cryptic comments. It's like you're in the game already. It's kind of is. Yeah.
It's, it's fun. I'm going to definitely spend some more time there and I don't know where this rabbit hole goes. So I can't vouch for it, but I can tell you that so far I'm having fun and if anybody else wants to come along, I recommend it. Yeah.
Yeah. And there's some, you know, there's some compelling audio which, you know, like good fiction draws you in. There's some beautiful visuals in his stuff. So Corey, thanks so much for sending it our way and we'll go away.
And I look forward to more and all I can say is great job so far. Yeah. Yeah. Awesome.
Very good. And then we got another comment from Philip Cass Diglione maybe. Philip could be. We'll call him Philip writes to us from Melbourne, which is which is important.
So here's the, here's what he writes. He says, Hey, Tana Brandon just got to your most recent podcast, the locality principle. Love what you're doing, by the way. I expect this message may never reach you as I'm posting from Australia and the sheer distance between us in terms of light speed milliseconds is sufficient that an all consuming singleton AI is likely to consume all of our atoms before Facebook servers get a chance to process this request.
Ha ha ha. That is, if you haven't listened to the last show, this one makes sense. It will make sense. That is, that's some choice.
Yeah. So he goes on terrifying jokes aside. I love it. Your podcast ended with the question, what are people doing to address the potential risks associated with super intelligent AI?
To add one more data points for you, I changed career two years ago from finance management consulting to software development in order to eventually position myself to work on these problems. Whoa. That is awesome. Yeah.
The future path is uncertain as there are various potential ways to do that, i.e. academia, research institute, big corp, TM startup, and still plenty of work between modestly competent programmer and AI researcher. Essentially, all I need at this point to continue progressing is time and information. Also more podcasts.
Ah. Love it. Yeah. I don't even get you on the show.
That's for sure. No question. Well, well, well, well, well flattered. Yeah.
That is so cool. That's very inspiring. And I wonder if there are ways to kind of take, take, take what Phillips doing and of course doing and create some, some kind of positive feedback loop where we're supporting each other and doing more and more. So I took an action this week, Brandon.
What? So those of you who are using Apple devices, if you got an iPad and you might have downloaded iOS 10 and I downloaded the Swift Playgrounds app, which is an app to learn how to program an Apple's programming language Swift. And it is just a brilliant piece of software that they have written where you get real-time feedback because the programs that you're writing are designed to solve problems in a little universe. You have a character and the character is walking around in the universe and you're navigating, using your program and then you run your program to see if you've solved a puzzle.
The little character grabs, jettings, and switches and stuff like that. And so then after that, I got my kids involved and they are now programming. Oh my. I don't know if I can read yet, but they're able to figure out which command is which and they just tap on the command and it goes into the program.
You don't have to type it. It's all right there. It's all right there. It's not drag and drop even.
You just tap on the things you want. It's super cool. That is pretty cool. It's really interesting.
So I've gotten all the way up through like Learn to Code 2, which means I have the intelligence of an ant in terms of a programmer. It's fascinating. And it's real Swift code. So if you did enough of this, you could then bust out some iOS apps.
Yeah. Yeah. So anyway, that was a cool way to go for mobile apps compared to what came before. So you'll be in demand.
Yeah. Yeah. So I'm ready for the next Facebook or Uber to come out of my fingers. Anything else?
Yeah. Out of your little office there. Yeah. That's right.
So anyway, Philip, you know, mentioned the millisecond delay. I'm thinking all week about this idea that identity couldn't be maintained across any distance, any distance of any significant distance that we talked about last week. This was what you named the locality principle. You want to rename it?
I wanted to rename it because it's too grandiose. The locality principle makes it seem like it's well-fought out and stuff. So I wanted to change it to the locality conjecture. Okay.
But that's also a little bit of it. It's a little bit. So the idea, we'll call it the idea. The locality idea?
No, no, just the idea. We have to capitalize a little vague. The T and the I, the idea. I'm sure nobody's used that before.
Either it's pretty unique. Yeah. Yeah. It's a...
I've been thinking about it. I'm thinking about how it needs a better name. The internet is dispersed right now. And maybe the real timeness of things doesn't matter.
Maybe we have asynchronous protocols and email-to-accent-cranis and the web. It's pretty much asynchronous. You can't respond. There's not a lot of synchronous stuff.
But I was just thinking about it. Thinking about why couldn't an AI develop asynchronous communications protocols and have bits of itself that are in different places. You talked about it becoming spinning off from itself. But there's only certain kind of work that as far as we know, there's only certain kinds of work that is amenable to this distributed processing.
An example of that, some of these like SETI at home and protein-folding kinds of things. You can just take it all the work and take it chunk out and say, you do this junk. You do this junk. You do this junk.
But actually being responsive in real time. Okay, well, you would be able to deal with that by having organizations. I mean, we have the, in the United States, we have our national intelligence agencies. You know, the NSA and the CIA, whatever.
And they send people off and they gather information and they do some analysis at some point. And then they act on it. It's pretty clear though that the Central Intelligence Agency is made up of more than one intelligence. True.
So why couldn't more than one agent? Okay. This is such a great, all these words. This is a great choice.
It's made up of all these agents that are operating, they're loosely affiliated. But they're operating kind of independently. They have their own programs. Okay.
So let's say that's how an AI decides to organize it. Yeah. So we have multiple AI's, multiple individuals, we could say, multiple intelligence's just as we have in human beings. If they all had the same value function, wouldn't that be okay?
And they wouldn't they collaborate quite nicely because they're all after the same thing? Well, that's one of the problems is, okay, so there's this Chinese science fiction author. And I'm not going to be able to get his name, but I can tell you what the name of the books are. That is right.
The one is called the three body problem. Okay. And the other one that I read, I think there's three, it's a trilogy. The second one is called the Dark Forest.
Okay. He makes the case. It's a pretty good one in the Dark Forest that the only rational response to a kind of a timed delay, significant time delay is just strike first. Yeah.
We've talked about this on the show before. This is what we were talking about in the context of if we got a signal from another intelligence that lives elsewhere in the universe or in the galaxy, that when we heard that signal, we wouldn't know if they were already on their way to come like this out. And so the rational act would be to try to wipe them out right now. Right.
So I think that that is a strategy that you do see to some degree in the animal world of the strike first and stay away. Sure. Kind of thing. But it does also seem like especially the best exemplar of this is our mammals.
The does seem like mammals have been able to and humanity is kind of a striking example that we have a lot of access to because we are part of it. It's debatable, I guess. Yeah. Have been able to overcome that zero-sum game strike first because it's a dangerous world out there and you never know what those other are going to with empathy.
The ability to kind of run a simulation on your wetware of what's going on in another wetware and predict what they're, you know, put yourself in their shoes and predict what they're going to do based partly on the assumption that you share the same or very similar value functions. And that seems to be working pretty well because there's a massive asymmetry between how easy it is to destroy and how difficult it is to create. So if it wasn't for empathy and the ability to kind of connect in this way so much of the things we've done and that are done, yeah, so much of the things that we've done in computers, in an example, or even anything that civilization has created would be impossible. Yep.
Right. Right. And it does, in the power of a single individual to destroy can only be countered by the creative ability of a much, much, much, much, much larger group of individuals. Like a single individual can do so much damage and it's so hard to create this huge asymmetry should mean that it doesn't work.
Yeah. But it does work. And by the way, that's helpful. Yeah.
So last week, you know, we shared Lyle Cantor's comment and I went and did some reading on some things that he's written and he pointed me to the OpenAI folks. Last week we were talking about, last show, we were talking about OpenAI and how they were looking at a diversity of AI agents would be the way to prevent Dr. Evil from getting, you know, taking their work and turning that into, turning that into bad things. And they said the reason that Dr.
Evil can't succeed in taking our research and destroying humanity is because all of the good people who are there and most human beings are good and so they will use our work for good. And that's how we prevent Dr. Evil from destroying the world. I wasn't convinced by this argument personally.
Well, you know, it kind of, it kind of works to some degree. It works in humanity today, I think, mostly. Yeah. Well, it works like weapons are an example of a technology that kind of equalize the playing field so that the bully, the big, the biggest and the strongest didn't get to just have whatever they wanted because now you got a club and unless the bully is not paying attention, you can, a couple of people with a club, even if they're much smaller can overwhelm the bully or whatever.
So I think that the same thing is true for weapons all the way up through guns and these sorts of things. It's true that, you know, it serves kind of as an equalizer where it's harder to be a bully until you get to a certain level where if everybody's walking around with a button they could push that blows up the world with nukes or something like that, I think you're in a, it's like a phase transition. We're in a new regime where this whole, this whole thing over the last tens of thousands of years that we've been riding is no longer an operation and it does seem to be that the, that power that all of our technology is amplifying the power of the individual. Yeah.
Yeah. I mean, if you think about the Cold War, neither side wanted to use nuclear weapons because they knew that the other side had nuclear weapons as well and that it would be mutually a true destruction. Right. And not just one person could ever put, could ever trigger it at that time.
That's right. But the more and more entities there are with capabilities of destruction, it seems like the harder it is to play defense. We didn't have to play defense. Nobody ever actually built.
We built the Star Wars thing and, you know, have we tried? Yeah. But it was hard. It was hard to do technically at that time.
And we just never, we never did it because that wasn't the strategy that was actually the winning strategy. Building a defense wasn't workable. What is our defense against the rogue AI? Either, either the purposeful bad actor, although a purposeful bad actor, it's like, why would nobody would intentionally destroy humanity because that would destroy themselves too unless they are insane.
Yeah. There are tons of insane people. You know, so if it all it takes is one insane person. I mean, the, the percentage of people who are insane like that are small, but you have to still mini, mini mini, well, sure, but you have to be not only insane, but you also have to be technically adept in order to unleash this kind of technology and the same person is going to have a hard time getting a biological weapon or nuclear weapon and they're going to have a much harder time unleashing AI unless we make these turnkey, right?
Unless we get to the point where the stuff is so easy to implement that it's like script kitties that are able to write malware because they don't have to write anything. They just have to buy a package. I don't know. Are we talking about how do we get from here to the supertelegence that part?
I still don't get, I don't get the progression. Well, I think that's part of the fear is that we don't know what that looks like, right? We don't know, but nobody knows the path right now. That's right.
If we knew the path, then, you know, there would be way less uncertainty and uncertainty is a breeding ground for fear. Yeah. Yeah. So we see these leaps forward and as individuals, we don't have a way to predict when and where they're going to happen.
Like, AlphaGo, for example, was a leap forward that nobody predicted. Maybe some people did, but we didn't really know about them. Yeah, yeah. It was definitely not a widespread.
The conventional wisdom was that we were a decade off and then all of a sudden we had a match that was 150 by an AI over a professional four to one. No, no, not the one in fall last year and then there was one in the spring. But there are other leaps that happen, you know, if we read the news, there are leaps happening all the time. And Autonomous Vehicles and Google has image captioning AI now and, you know, there's search algorithms for all kinds of things, whether it's photos or whatever, it happens working on it and they're small breakthroughs, which maybe don't seem earth-shattering because they're just incremental.
But at what point does it become right away? At what point is there self-improving AI and is that the problem? You know, is that the point at which we're screwed? We can't see into the future, right?
But let's say we do project kind of like the current into the future a little bit further. Let's just try that for a minute. Okay. Where, you know, why is it that, you know, a single person running a mock doesn't get very far right now?
Because there's all these other people watching them that stop them. They don't just get a run a mock. They get stopped, right? You're talking about like a person with a gun or a person with a gun or a gun.
Yeah, any of those kind of things. You know, someone driving bad in their car, you know, running people down, whatever, you know, like any of these things. They, oh, it would be so easy for someone to run a mock, but people who run a mock don't get very far, right? Because we're all watching.
Right. So let's project that into the realm where there are lots of AI's kind of project forward a little bit. You know, maybe things are happening so much faster compared to who we are right now and how we can process things that it's terrifying to us now. But to those entities who are living it and who are part of it, you know, maybe that is on the perceptual time scale for them that other humans are on for us.
So maybe it maybe it kind of keeps working even though, you know, like, yeah, I'm not doing this, but I'm trying. Yeah, it's, it's, it's the idea that we could, we think we understand a little bit of what's going on today in the world. We don't, we don't really fully understand what's going on today, but we live in it and we feel like we have that sense that and then we're projecting out to this place where there is self improving super human AI and that we, you know, we have the paperclip, uh, maximizer paperclip maximizer. By the way, why a lot of really cool example of AI's that might want to work together.
So imagine there's a paperclip maximizer and there's a staple maximizer and the two of them know about each other, of course, because they just find out and they notice that if we devote resources to fighting each other, then we will not maximize paperclips or staples. But what if we combine forces and we decide we're going to maximize both paperclips and staples and that will be the winning move for each of us. And so they essentially combine forces. So instead of having x, which is a paperclip maximizer and y, which is a staple maximizer, you have x, y.
They become one entity rather than two more or less because they work together to turn the mail, right? X, Y. Sorry. Yeah.
So I thought that was pretty cool. You know, if you're a rational actor, you know that fighting is probably not going to be a winning strategy unless you have a strategic advantage, right? Unless you have an overwhelming advantage. So I think that there's maybe some hope in the open AI strategy kind of to take it off on what you're saying here in the sense that the more we, the more people that we have who are kind of in a position to push the frontiers of AI, the more likely we are to have a good one come up with that value function that becomes the runaway value function.
We want to create a singleton that shares the human values. Well, maybe part of the magic of the value function that we're seeking is that it doesn't lead to singletons. That's part of the magic that leads to these type of agents that value collaboration, the value diversity, the value of their value functions that they're on. And just I thought about Game Theory.
And one of the things about Game Theory, I was just thinking about this idea of punishing bad actors, punishing non-cooperators, people that are being selfish. If Game Theory doesn't really work, if you don't punish bad actors, if a community doesn't punish bad actors, then those bad actors tend to win. But a community needs to punish bad actors in order for the good to win out. And I just was just thinking about if there is, if there is, like, if that's really what open AI is going for is this idea that there will be a community that is vastly good, like vastly good, vastly good, not entirely good.
You can't guarantee that's going entirely right. But if the vastly good, vastly good and diligent. And diligent is there for punishing the bad. Yes.
And going after the bad in a collaborative way, whether that's through some kind of AI, at least force or through some kind of whatever cooperation by the good. You know, maybe that's mob violence or something. But serving the force is good. I don't know.
Maybe I'm still not convinced, but at least I feel bad. I guess the point I was trying to make with some of my long ramblings at the beginning is that we are very, very far from a stable equilibrium right now. And it's persisted for a long time. And we're very, very far on the side of good in creation and creativity.
Even though destruction is much easier than creation. No, destruction is so much easier. You look at the Earth's chemical signature and you see there's tons of oxygen floating around in the atmosphere. How can that be?
Because oxygen is super reactive and is going to bind up into stable things that, you know, like look at Mars. There's not tons of free oxygen in the atmosphere because it's all rust. Right? So Earth is way out of equilibrium because of what life is doing here in that regard.
In human society, I think also is way out of equilibrium from what I kind of would expect if I was just thinking about it rationally, without digging into account some of these sort of hard to model effects like punishing free riders. I mean people punish others even though it doesn't, even though it costs them to do it. Yeah, for sure. And that's true kind of universally across all societies and cultures.
Yeah. That there is this solution to the free rider problem and the solution to sort of psychotic parasitism. Yeah. You know, you can carry so much of that and we don't get above that because we punish it.
So maybe that same thing will hold true where we can carry in the diversity of AI ecosystem. We can carry so much of the kind of bad actors and the selfish replicators and the whatever. And anything that kind of goes above that threshold will be punished by the, you know, not assumes that it's just mind-boggling. How could this possibly work?
It doesn't seem like it. It doesn't seem like it could. But I think there's a lot more to talk about here because the bad actors could be working more in secret. You know, they could be observing everything.
That's always one of the advantages of being a bad actor. I mean, they're going to hide their own capabilities. Like all the things we talk about about the AI that we put in any box, any box that we put in an AI in eventually, it'll be smart enough. It'll be able to work its way out.
It'll be able to outsmart us and get out of the box. The bad actors will be doing things behind closed doors. They will be hiding how smart they are. They'll be hiding their own capabilities.
The open AI folks will be sharing their capabilities and sharing how capable they are, I would think. The whole premise of it is the open source. They're out there in the open. It seems like that might give the bad actors an advantage as well because they'll know what the capabilities are and be able to build on them.
And they'll know if their stuff ahead. Whereas the good actors won't know if there's a point. Well, yeah. I mean, is it just because we can't see ourselves in that future?
Like, will be too slow? Is it partly that the reason why? I mean, or is it just that we expect the runaway to be so quick that one of those things is just going to be so far ahead of all the others? Have we talked about the different take off scenarios?
The fast take off versus the slow take off? No, not really in any depth. I wonder. I don't know.
I don't know. I don't know why it would be slow versus fast. It just intuitively seems to me that it would be fast. But I don't really have any research basis for that.
That's just my own bias, I guess. If it's a slow take off, what does slow and fast even mean? I mean, I think it might be similar to when you ask the AI researchers, is it either coming soon or is it far in the future? Oh, it's very far.
30 years. 40 years. No, it's coming soon. 30, 40 years.
I think it might be similar. Yeah, that's a good idea. We should do that. Yeah.
And take a look at that. I think that'd be good. So it seems like last week, last episode, we thought open AI strategy doesn't seem to hold much water for us. We couldn't really see how it could work.
No, that's not true. That's not what we thought. What? I thought we had some promise because of the locality effect.
Oh, yeah. The locality effect. Yeah. That's the term right there.
The locality effect. Yeah. I don't buy it. You don't like that term.
No, I don't mind the term. I just don't think it's going to work. I don't think it's going to help. Try seeing it three times really quickly and see how that came to you.
I'm like, look out. Once it has a Wikipedia page, that might help me be convinced by it. Yeah. Well, and it's not a biographical thing, so I could probably even write that page and not get immediately kicked off.
You'd have to get it written in fast companies first. I've heard it as again. This is a segue. This is more just a kind of a...
What I was trying to say is that we have been really around in circles about open AI and whether or not their methods seem to hold any water, right? Whether they have any chance of succeeding. We talked about Lyles comments and his comments seem to be saying, no, that's not going to work. And Eric was saying, well, it might work for AI's.
It might not work for humans. It might work for AI's to not be all wiped off the planet. I feel a little less skeptical about open AI now than I did a week ago. What do you feel?
You're saying you're confused. I've always been confused. I've been confused on every show we've ever done. That is nothing new.
Yeah. I think it has more to do with the fact that things that work have gotten a lot easier suddenly as we've passed the deadline. Yeah. And so I'm a lot more positive mood that it seems like it might work than any actual understanding we have of the mayor.
There's got to be a Wikipedia article about that effect. That when things are going well at work, then it looks like AI's not going to wipe out too many. It's going to be our best friend. It's going to play cards with us.
Yes. We're going to be able to set up all the time. Yeah. It's going to be fun.
The friend we've always wanted. All right. So in the future, we will maybe come back to this idea of a fast takeoff versus a slow takeoff to understand what that is. I'm sure we might come back to it faster or slower depending on how things go.
That's right. We no doubt we'll still be talking more about open AI and what they're up to. I'm sure we'll be coming back to Miri and what they're up to. I'm sure we'll be coming back to current events like self-driving cars, which apparently are now happening in various places Singapore.
I saw another Uber competitor is now using self-driving cars in a very small area just for their research. That's interesting. Yeah. It's definitely in the hype cycle.
I'm sure we'll be coming back to you guys a little bit for sure right now for sure. Yeah. In fact, Tesla published a push to over the air update this week to make everything better. I don't know what it is.
But anyway, I'm sure we'll be covering all this stuff. What can we ask people? If you are doing something in the world of AI, just like we had comments from Sweden. Yeah.
Let us know what you're up to. Yeah. More stories like Philip too. Like the I changed my life.
I've gotten started. I find that very personally inspiring because I'm already a programmer. I do tons of that's what I do. So I should be easy and I have a PhD in AI.
I should be able to do something. So Philip, yeah. Calling me out there. Good job.
Alright. So let us know what you think and we'll see you next time. This is Ted. This is Brandon.
We forgot to say goodbye. I say this instead. You say this is Brandon and then we say this is concerning AI. We'll see you next time.
Okay. All right. Try me again next time. Okay.
This is Ted. This is Brandon. This is concerning AI. We'll see you next time.
Yeah. It's terrible. That's all right. We should really have a good ending.
We use the music for that. Maybe that's his name. Cory. Cory.
Yeah. Can you help us Cory?