Cerebras IPO, the Dead Internet, and the AI Backlash

Transcript generated automatically from the episode audio and lightly cleaned up (filler words removed, misheard names corrected where clear). Speaker labels were added afterward and may contain errors.

Rob May: I'm Rob May. I'm the co-founder and CEO of Neurometric.

Ryan Eppley: And I'm Ryan Eppley, co-founder, CEO of RootAccess. This is the AI in NYC show.

Rob May: All right, everyone. So, Anna's sick and, our guest got hung up in with some travel plan changes. So, you just have Ryan and I today. But the show's getting enough traction that we want to keep putting it out there. So let's get let's let's actually let's get started talking about that really quickly. So, there's a couple things you can do if you like the show. Obviously, you can go rate us, send it to your friends. We have a newsletter you can subscribe to if you go to the website. And, lots of things we're offering.

We have a job board now. So, if you're a New York based company and you want to post some jobs, contact us about that. We had a president be we had a sponsor be present. Which you should go check out. They are done. And we have two new sponsors coming up. So, yeah, growing and everything else. And so, those of you guys that have been watching us for a while, thanks for thanks for growing with us. Thanks for being part of this.

Ryan Eppley: Yeah, seriously. And honestly, we have a ton of great news. So, we didn't want to skip this week because I think so much so much is happening in the world right now, Rob. I know you teed up maybe like four or five articles. So, let's let's hop in. What's what's first on the list?

Rob May: So, the one that's most interesting to me today is a Cerebras IPO and this is interesting to me for for two reasons. Number one, I'm a former hardware engineer. You run a company that deals with hardware. So, so you and I have a interesting perspective on this that I think a lot of other people don't. The fact that you know there's other ways to process AI other than GPUs and other than Nvidia and AMD chips and so that's part of it and the second thing is a family office I know actually asked me you know Cerebras originally filed for their IPO back in 2024 late 24 decided not to IPO at that point in time looks like it was smart decision given what happened today and so I went pretty deep on it back then and it's a it's an interesting chip do Do you do you know much about the chip or should I

Ryan Eppley: Yeah, enough that you know I felt like people were asking me if they should invest and I you know gave them the technical breakdown. They're like all right I don't know what all this means.

Rob May: Do do you want to explain to the group what that it's a wafer scale chip and what that means?

Ryan Eppley: Yeah. I feel like you know first and foremost people need to understand before you even dive into some of the wafer and things like that. It's like everyone's been talking about transformers. Everyone's been talking about GPUs. There's even a bigger trend in the industry that people are going back to CPUs now as well, right? So I think the opportunity is not just hey GPUs will dominate all of computation. I think we're finding that there's so many different ways that people are looking to and I think Cerebras is cashing in on that opportunity that they saw years ago.

Rob May: Yeah. So one of the things that you have to remember is inference workloads vary quite a bit, right? The difference between so the ChatGPT if you think of that as the killer app of AI ChatGPT was conversational search conversational search requires a very large model because it has to know lots of things and be able to talk about a lot of things but when you look at other opportunities of AI like you know image classification that needs to run on an edge device like a robot or a camera or a car that does not require a large GPU does not require a large model necessarily image generation requires a larger model.

So these types of inference workloads which is like what are the thing you're trying to do with the AI are quite different and as Ryan was saying like a lot of the as we're moving towards agentic AI which is probably going to be significantly bigger than conversational search at scale. So, so give it give it 18 months. But, these agents, they kind of need LLMs to handle unusual use cases that they don't know how to do or to coordinate and break down tasks for them, but they don't need LLMs to do the tasks. They can use smaller models. They can use they they can use CPUs for a lot of that work. And so it's definitely changing the structure of compute. Not that GPUs are going anywhere. We're still going to need many many more of them.

Ryan Eppley: I mean, Nvidia just hit a all-time high again. Intel, though, the dark horse of the, you know, computer kind of landscape just crushing it lately. I think they're at like a hundred now in their stock. And they were down at 20, I think, less than 12 months ago or something crazy, right? So, yeah. One, both of us just being like hardware, you know, lovers for for all these decades. Seeing what has happened, I feel like this IPO has really been a new entrance into this into this like landscape that I'm just happy they IPOed at all because when they did an IPO in 2024, I thought maybe this is going to be one of those like, oh, they're just going to raise a series E then F, then G, and they're going to play this this whole alphabet of we're actually never going to go IPO. Are just going to raise ungodly sums of private capital. But I'm just glad that they did. I think people want to invest in these types of companies.

Rob May: Yeah. So there's two interesting things about the Cerebras chip. One is so the vast majority of computers that we use if you're not familiar with this use what's called a von Neumann architecture and you have a compute cluster of circuits that perform computations and you have memory cluster of circuits, right, that store data. And as models have become bigger, the reason this has become the bottleneck is you have to move the so think about this as ChatGPT as a program that lives in memory and you have to load some of the neurons into the compute process, process them, store those new nodes back out in memory with the new values, put some new and so for a big model, you have to shift back and forth a lot. That gets even more complicated, which is as models get too big to run on one GPU, you have to shard those to multiple GPUs. So now you got to run on four GPUs or eight GPUs or I don't I don't know what a trillion parameter model I don't know what 10 trillion like Mythos I don't even know how many GPUs it takes probably clusters worth just for inference.

So there's there's a couple ways to fix that. There are many ways to do compute that are not vineo architectures. We're seeing a rise in analog compute particularly for edge related things. But one of the things that cerebras does is Cerebras tries to put the memory particularly SRAM memory which is super fast spread closer throughout the chip to the compute clusters. So instead of having like think about it as like a hard drive right having a hard drive over here that loads memory into something you have memory dispersed through the chip. So it's much faster shorter distances and all that kind of stuff. So that's one innovation. The second innovation is

Ryan Eppley: I don't even think people appreciate just like one how more optimal that is because so much energy is just spent sending ones and zeros around to either store and local or even just the ability to not have to transition large amounts of workload in itself is saving not just energy and time. It's also allowing people to start pushing their hardware a lot farther. And I think that's like one of those things that people in the hardware space for so long have been asking for, but at scale they never had it. There was never a company that said, "We're going to build this at a large enough scale for you to really get behind." so this is why it's so gamechanging in so many ways.

Rob May: Yeah. And then the second thing is when you make you know circuit boards like or chips GPUs whatever you're making they come out on these big wafers and these wafers are sort of like this big and these wafers are sliced up. They'll be you know eight or 10 or 60 like I don't know how many but you know depends on the circuit but there'll be a bunch of circuits on the wafer and they slice them up. Now what Cerebras did is they said we're just going to use the whole wafer. So whereas you know you have chips in this phone that came out you know many many many of them on a wafer a Cerebras chip is the whole wafer

Ryan Eppley: just a giant pancake one.

Rob May: Yeah exactly. So it's this giant wafer with all this SRAM and so it means a couple of things. It is it can hold a really big model the kind of model that you have to break up across multiple GPUs can run on a single Cerebras chip and with all the SRAM split you know through throughout the like it's lightning fast. So if you're worried about the speed of your AI like and you can you know you're not an edge case this is a data center case I mean Cerebras is the chip for you. So things like like AI search I think perplexity uses it meta uses it like it is the chip to run models fast and large models fast. Now, it, you know, it burns power and like crazy and apparently gets super hot and so, you know, you have to be careful with some of that stuff. So, there's there's downsides. There's trade-offs in all this.

But anyway, you know, good for them. They so they IPOed today, the day that we're we're shooting the show. So, you guys will probably see this in a couple days. I've been tracking the stock in the last like couple hours, but I saw it went up. I don't know how much. Basically, I mean, at one point in the day, it was up more than 100%. So, I don't know where it ended. Pro probably a little bit less than that, but I think it ended up 80 85%.

And so to put some perspective on that, they have about $500 million in revenue currently. They have heavy customer concentration. So I think two or three customers are like 75% of that. You know, and so these are these are rough numbers. I don't know the exact numbers. And so I think the end of the day over $50 billion over 100 times revenue it's trading at now the I market is hot at the moment. Cerebras definitely has a big market opportunity in front of them but to grow into that I mean they've got to be at you know to you know when you think about growth slowing down and like where they got to be I mean they've got to be in tens of billions of dollars in revenue. They need to be there in a couple of years. So that that is aggressive. Put it this way. I didn't buy the stock, but I'm excited about it.

Ryan Eppley: Yeah, I feel I feel like everyone is trying to figure out a moat in the AI space and undoubtedly when you have hardware and the asset itself, yeah, has something more than like software behind it, it feels it feels worthwhile to put some money in. But you're right. I thought I thought the pricing was like pretty high overall, but still a huge supporter and I still told people that they should look into it more.

Rob May: Yeah, I defin you know, no opinions on the stock at the moment. I kind of want to see where it settles in and want a little more about the company, but big fan of the chip. I definitely think you're going to as you see inferenceware workloads fragment, you're going to need lot lots and lots of time.

Ryan Eppley: Let's be honest. I don't know how many viewers have been following this, but like people have tried building all different sorts of new chips, AI, you know, native workload chips and things like that and they haven't panned out. You know, there's at least on one hand, I can start counting in the last couple months I've seen founders unfortunately have to shut down companies that have tried building.

Rob May: I was an investor in RAIN neur which is neuromorphic chip and it was a great team too.

Ryan Eppley: Super smart team. Yeah. And so I feel I feel like people are like, "Oh yeah, as long as you build like this great hardware, like all of it will work out for you." Like that's not quite the case. Like there is a bit of timing, luck. Obviously the engineering under under the hood has to be really sound, but I feel like this is a great success story because they they definitely were one of the pioneers in so many ways.

Rob May: Yeah. So the criticism of Cerebras is really a market competition criticism which is one of the chips we mentioned that you're probably thinking of was Groq. Groq with a Q, not Elon Musk's Grok with a K. Groq with a Q. Similar to when Cerebras's father IPO Groq with a Q, it was thought they were going to go out of business. They weren't going to make it. And you know they they got through and they actually got acquired by Nvidia. And there's a couple of reasons people think that acquisition happened. One is the CEO of Groq with a Q was the lead one of the lead developers for Google's original TPU tensor processing unit. So very early on the AI compute side. And then the second thing is I think Nvidia recognized that the world is shifting from training to inference and Groq was an inferenceon chip and so merging those together could present a serious threat to Cerebras.

Ryan Eppley: Yep.

Rob May: So we'll we'll see what happens but okay. So, so, so jumping on from the Cerebras issue, which, you know, we've talked about, you know, the heat and the power in data centers. We got this article from the Washington Post. 71% of Americans oppose data centers in their communities. So, I guess two two questions for you. Number one, is this rational if you live in the community? Like, should you should we be doing that? And number two, where do the data centers get built? Do they have to go to Alaska or the wilderness?

Ryan Eppley: The nimbies are are hot on this topic. I also, you know, maybe this is just I haven't looked enough at the data and where it's being sourced from, but like did how big was the sample size? Is this like a 100 people? Is this a thousand people?

Rob May: That's a good question. I will I will

Ryan Eppley: Yeah, because anytime I hear like numbers like this on on a hot topic, I'm always like, you know, is this a hundred guys on a you know, forum just like putting in a yes or no?

Rob May: Yeah, people that still have land was this people that still have landlines and they called and [laughter] pulled them like of course they're going to oppose data centers. You have a landline.

Ryan Eppley: Yeah. And what you know town are they from? What zip code? Things like that. I think there is truth no matter 71% does sound high but I think it is you know a topic where if you don't have an opinion people lean no because it feels like oh I don't want to it's almost like the I don't want to live near a recycling facility or trash facility. I don't want to live near a highway. I don't want to. People don't like the idea of living near things unless they are parks and neighborhoods, right? So I think in it's a default answer, right? If you're like, oh, do you want to live next to something? It's like, well, if it doesn't sound nice, then probably not.

Rob May: Yeah. So thousand Americans surveyed so not a huge survey but 70% worried about the data cent's environmental impacts which is you know excessive use of water power my my actual concern you know would be more like noise or power outages I did you read this there was an article maybe like two years ago I guess Elon Musk had built a data center somewhere for like Tesla or SpaceX or something and they were they were arguing like there was this like buzz was going through the community that they couldn't get rid of and people couldn't, you know, people tuning to the data center but they couldn't prove it came there. So that you know weird stuff like that could happen. That's what I would be more worried about.

Ryan Eppley: Yeah. Just personally speaking like I don't think I'd like to live next to a data center. Would I be fine with it in my like general zip code though? Probably. Would it be an isore for sure? You know, so I feel like where do they go is the question you asked. Let's build them underground.

Rob May: Let's [laughter] build them underground, right? I mean, moving dirt is super expensive, though, right? Like, so even just moving the earth to make that happen.

Ryan Eppley: I put it this way, they they have to go somewhere.

Rob May: Yeah. Right. It is infrastructure. Well, they want to put them in space now. And most of the people that I know that know anything about thermodynamics are like, "How do you radiate all that heat out in to space? It's hard."

Ryan Eppley: Well, that was the big assumption about six months ago and a lot has changed, right? You have StarCloud at a $2 billion valuation. You have they just rebranded. So, it it's now Cowboy Space, right? They just raised I think it's like close to $300 million and they're going to build their own launch vehicle that turns into the data center, right? In like the stage two or something like that. So I think there is this notion that the science is either there or it's going to be pretty close to being more not just economic, but now we don't have to worry about the NIMBY thing, right?

Rob May: Yeah. Oh, all the data centers are just going to go into space. No one has to live near them anymore.

Ryan Eppley: Yeah. Obviously, you still got to transfer all the data. So, there's there's still so much infrastructure to be built.

Rob May: Well, it's interesting to think about what you would have to do to make micro data centers work, right? Because there's so many advantages to scale when you're building a data center particularly the power coming in but also like the land and then the networking have them all there you know next to each other and so what would have to change about the world or about technology if you were just sort of like put these micro data centers with a handful of clusters of GPUs and yeah put them in the basement of buildings all over Manhattan or wherever

Ryan Eppley: I'm trying to find it live right now but someone someone on X posted it was a hilarious like meme or photo you know how there's those like free library boxes in neighborhoods and you go pick a book? Someone posted like a small like set of like hardware in there. It was like community data centers. Yeah. [laughter] I thought hilarious. But to your point like these these small clusters inside neighborhoods instead of them being like data centers, it's like these small micro ones. Well, I have seen some companies that are rising up that are trying to say like, you know, if you have a device with a GPU in it, you know, like your laptop be able to, you know, let somebody use that when you're not.

Rob May: I So, first of all, the last wave of those didn't work out, right? When I was when I was early in the backup industry in the early 2010s, sort of cloud storage was a thing. People tried to build like several companies tried to build distributed S3 where you know the files you were storing were on partially on your computer my computer and Anna's computer and whatever and like it just it never worked for a whole bunch of reason one is just really hard technology and you have to keep track of everything and you have to you have to you have to store things in multiple places actually because like what if your computer's not turned on and I need it right security and privacy it's like get gets real weird.

Ryan Eppley: Yeah, you know, there's like multiple handshakes and how like they come back together, right? You have to like package these all back together. And I wouldn't be so worried about my if I was my hard drive maybe, but like I don't know. Will you wear your GPU out faster if you're constantly running it on these intense? Yeah, of course. And honestly, isn't this like the backbone of the fictional story of the Silicon Valley HBO show? Like weren't they wasn't the Pied Piper team like trying to build distributed compute on everyone's like phone and things like that?

Rob May: Yeah, there was something about Yeah, they had a compression algorithm and it it was doing some stuff like that. I remember.

Ryan Eppley: Yeah. [laughter] So aside Silicon or Valley it needs to come back, right? Like where is HBO when the people need them?

Rob May: Oh, I know. Can you imagine? They should do one new season. Was it Mike Judge?

Ryan Eppley: I think so. They did it the same butad guy and whatever.

Rob May: Like we should we should send them a note and be like, "Hey, you need to make you should do one more show, one more season, and it's about AI and it's about the AI craziness cuz they even started foreshadowing it, right?" You know, so I don't know. But yes, data centers, people aren't happy with them. I think it's easier to always just default to no, I don't want them.

Ryan Eppley: Put it this way, they got to they got to be a part of our long-term infrastructure. This is like a non-negotiable need at this point. So, I feel like either a we're going to have to find ways to hide them, build around them big bushes around this, you know, kind of community development plan, or we're going to have to spend the money to transfer large amounts of Ethernet cable into the outskirts of cities and things like that. We'll we'll find ways.

Rob May: Yeah, Gilfoyle named his AI at one point.

Ryan Eppley: Yeah. What was the name? I don't remember, but I haven't seen that show in a long time, but that was a fantastic show and had like an A in it. Was it like Arthur or something?

Rob May: He ordered a bunch of hamburgers. Anton.

Ryan Eppley: Yeah, it had an A. Had an A. I was close. That's right. I remember that now. All right, what's next? What else we got on the news?

Rob May: Okay, so dead internet theory. M so this is the idea that there was a prediction that you know the world is going to turn into AI slop and the internet's going to become useless and I think these Stanford I think it was Stanford researchers who were looking at it are now saying a third of all pages are AI generated now probably a large part of the I mean our website's mostly AI generated with claude but like I reviewed it so part of the question is how how much of it is reviewed and what does that But, you know, do you do you worry about that? Because I you know, you're you're I think sort of like me and that you kind of came up in a time when you were like the internet was awesome and young and different and it was all the tech geeks on it and we loved it and you know, even before AI, it was kind of become filled with crap.

Ryan Eppley: But, you know, I feel less inclined to care about if a web page had like AI generated content on it when the content is, you know, it's commercial content, right? Like it's just like sales content, you know, imagery, things like that. But what I don't like is blogs and written articles that are supposed to be technical or somewhat detailed. They they do feel they do feel not good. [laughter] I don't know how else to say that in a more sophisticated manner, but I don't like the idea of like, hey, I'm just going to vibe code. It's not a vibe code. I'm going to vibe write this article for my SEO, right? I feel like that's a part of dead internet theory in my opinion. I don't want to read an article that's just like regurgitated. I want to read someone's thoughts, right?

Rob May: I feel similar in it's like when you find those pages where you Google them and they're going to tell you something, you know, and it's like you have to keep clicking and expanding because they want the clicks and they want to show you more ads and the page is designed to serve ads. Not really. And like you're like, dude, you could tell me what I wanted in the first paragraph and instead I got to scroll down, you know, y 30 seconds. But there the whole articles theory here is probably true that like a good chunk of the internet is now getting to sloppification, right? We're starting to starting to feel that.

Ryan Eppley: I' I've seen that when I click on a company's website sometimes I'm fine with like it it's clearly made by clawed like you can see it right there's some type of Yeah. General thesis that you can immediately just be like yeah this Yeah. But what I don't like is when they have an article or blog and it's technical specifically and you're like, "This was definitely not written by a human." Yeah. Right. That doesn't make me feel good. That almost like makes me not want to use the service now. I'm like, "Ah, man." Like, you can't you can't have one of your engineers like deep dive on this for a thousand words.

Rob May: Yeah. This is part of why I love SemiAnalysis. We should have Doug back on or or somebody from there soon. I, you know, maybe they use AI to write their articles, but I'm guessing they don't just from reading them because they're just so jam-packed. They're just written in a way that I don't see AI, right? Maybe they have a great cloudmd file and maybe that's what's doing it, but you know, it's they're really well done and really, you know, they're they're they're spec heavy, which AI doesn't normally write that way. And so, yeah, it's cool.

Ryan Eppley: Honestly, I started following more like writers and editors and newsletters in the last year than probably my entire professional life so far mostly because I start following them on, you know, X on LinkedIn. I start understanding their perspectives and like now I almost I go to them for insights rather than relying on ChatGPT because I already I already know what ChatGPT is giving me is just amalgamations with no real clear where is it coming from you know it just feels it feels like I'm almost having to second guess whatever Claude and ChatGPT are saying versus I can still guess why this author or why this journalist is saying something But because I've been following them, I almost like I almost appreciate it more even if I disagree. It's just the human element.

Rob May: Yeah. I mean I and you know we'll see where this goes with the world, but like I like to write and so I still you know I write a newsletter every Sunday and

Ryan Eppley: Yeah. You've been doing this for years.

Rob May: Yeah. Since since 20 2016 and it's you know before that I wrote a blog for 15 years and so it's so I Yeah. So I just like it helps me think through things. You know sometimes I would say like maybe two out of every five or six articles I kind of like write half of it and then give it to Claude to finish. But I have not had much success just giving it pure ideas and having it crank something out. Like it it really needs a lot of like excuse me like a like an outline or or something or like what you know what I'm thinking to get anything good out of it. And then I don't like it if it doesn't sound like me and says things I wouldn't say. So I still am very limited in my usage there.

Ryan Eppley: I have honestly maybe the best use case of for this topic of why ChatGPT and Claude and I've tried a couple others really failed me. Yeah. So we just publicly launched our first product last week and we had a welcome email. It was just like one welcome email, super short. And I was like, honestly, we should have like, you know, a couple of emails over the coming weeks, right? And I asked ChatGPT to make like a series of like welcome emails, a drip a drip campaign for new users, features, things that they didn't probably know about the product. And I read them. I was like, this is awful. I was like, I wouldn't want to read this. This is my own product, you know? And I just manually wrote them one at a time. And then I honestly gave it back to Claude and I said, "Hey, could you write a fifth one based off of these four?" And it still did a terrible job. Just the cadence, it just it felt awful. I was like, "Oh man." Like when you want to be personal, it still is so important to just do it yourself.

Rob May: Yeah. I think people will really start to You're always going to have that problem, right? Are going to have this problem that models that are probabilistic are going to play the probabilities. And if they've seen that 90% of articles like that read a certain way, that's the way that they're going to lean. Which means if you have the capability of standing out in what you write as a human, it's going to become even more valuable than it has been in the past because some of the people that could do what you do aren't going to do it for various reasons. And there's going to be so much more content out there that I think real human content is going to be really valuable.

Ryan Eppley: So, though I did, we talked on it maybe the other episode. What was the company that came out that basically, you know, you can upload something that was written by an LLM and it will output it in a way that feels very human. It's like an authentication, you know. I don't know if you talk about

Rob May: I so my kids who are both in high school say no tools like that where they're like make they know how to prompt it to make it sound like it was written by LM. [laughter] I don't know if I should be proud or concerned.

Ryan Eppley: That's good. They're they're hacking their way through life, you know. So I think we got one one more good article, right?

Rob May: Yeah. Well, related to this, Madi didn't add with this one in there, but related to this topic, the one we shared on chat was speaker at University of Central Florida who mentioned kind of like you're coming out of school at a time. It's so exciting, blah blah blah. AI got booed. So, I What do you think? Like I hadn't really thought about this until that moment, but like do you like I know younger people are a little bit worried about it, but do do you think they're skeptical? They're anti- AI. Aren't Aren't kids always going to be like anti- something, right? I don't know.

Ryan Eppley: I would imagine that there is a sense of job insecurity. People are very very concerned. We, for instance, had 4,000 applicants for our internship role over the summer and it already felt bad letting most of them know that we couldn't take we couldn't take let alone all of you. Couldn't take even like a small percentage of you. Yeah. So I do think there's there is angst around like AI is making companies so efficient they don't need me anymore. I also think it's just like cool to hate on on things sometimes if your friends hate on it. I think there's like some of that posturing that happens sometimes. As soon as AI has become cool for, you know, CEOs of super bland companies to like care about, maybe it's not the coolest cutting edge thing anymore.

Rob May: Yeah, I think that's also true. Yes. You know, related to that, I think I don't know if I mentioned this on the show, but my kids think I you know, we think Tesla's so cool. My kids think Tesla is an old person's brand, right? Yeah. Because they're like everybody's got one. Everybody that we have one is like 50 and so it's not cool. They they like Rivian and Lucid a lot more. So that's kind of cool.

Ryan Eppley: Yeah, human nature is weird like that.

Rob May: Well, this is why we go through cycles, right? Every generation sort of rebels against the generation before it in terms of the music taste, the style of dress, the standards. People have old old school cameras all of a sudden again.

Ryan Eppley: Yeah. I haven't seen a pager, though. Where's the flipones?

Rob May: Right. So, clearly people aren't ready to go all the way back.

Ryan Eppley: Yeah. Do you ever have a pager? You had a pager?

Rob May: I never had a pager. No. No. No. I never I never was important enough to be paged in that [laughter] time between, you know, getting a cell phone. I mean, I but I didn't I didn't get my first cell phone until I was 23 or 24. So, I did not grow up with one. My kids find this fascinating, right? They're like, "But dad, how did you what happened when you had to meet a bunch of people and plans changed?" You're like, "Well, if we were if we were meeting at this bar and we said 7:00 and then we were going to go somewhere, it's like if you weren't there by 7:05, we told the bartender like, "Hey, if this person that looks like this comes in and asks for Rob, here's where we went." You know, it was just I don't know.

Ryan Eppley: You made you made it work, right? I mean, people have made life work at every phase. So would you be booing this professor?

Rob May: You know, I would not. I so my point of view on these things is I it's funny you know doing the show and working in an AI I'm probably a little more of an AI skeptic about how beneficial it's going to be to society compared to most people which which I was about the internet as well but look it it boils down to like it's coming and you have to deal with it for better for worse right so like hating it and trying to avoid it is not a solution right so you can you can make make a conscious decision to live a different way. You know, and that's going to be hard. I mean, I don't know if I ever told you this, but like two years ago, my go because I tried to do like a big life change every year. It's like, what am I going to do this year?

And in 2024, I was like, I'm going to cut Amazon out of my life. I couldn't do it. It's just like so easy to send stuff and it's like I don't I don't have a Netflix account. I never have. So, my videos are on Amazon Prime. My music's on Amazon. And I was like, man, I got to start over. Yeah. And the whole thing and now part of it there was no easy solution to move to. If there was another easy solution, I would and so I think you're going to get that way with AI, right? It's going to become so entangled in your life. So I think you have to be careful and thoughtful about how you implement it in your company. And Mhm.

You know, you're going to lose I don't know what junior people are going to do in terms of learning how to code because it's great that suddenly everybody can be a programmer except that you know you're going to go from it's sort of like the phone trees, right? And having all this automated customer support. This automated customer support is awesome if you just have a simple question and you can get a quick answer. But if you actually have something complicated, it's way worse now than when you just just used to call and get a, you know, get a person who then could triage your thing. So now triaging takes many many many many minutes to get to a human.

Ryan Eppley: Yeah. I mean going back to the technical point maybe it's just our bias and it's because we have a lot of hardware you know in our focus but we started finding that for software engineering roles we love hiring electrical engineers because they even in junior roles they have a really good understanding of the concepts and they have to really understand how the computer works. So, as soon as they're like, "Oh, I don't know how this piece of software works." They're quick to figure it out. They'll look at the diagram. They'll go look at some of the, you know, sources of that information. So I feel like if you understand hardware, even if you've not done a ton of software architecting, you can you can go really really fast with AI now.

Rob May: Yeah. Yeah. It's super val I think a lot of what AI is going to create, not just in technical jobs, but in business jobs, is more systems thinkers are going to thrive, right? So people that understand how things work at multiple layers and then how those layers inter interact. And you know one of the things that people forget is you when you're designing a system you may suboptimize some pieces of the system to optimize different pieces of the system somewhere else. And that's just not a normal way of that most people think.

Ryan Eppley: Yeah. All right. So you got one more here around trademarks for likeness.

Rob May: Huh. Yeah. Celebrities are filing trademarks for themselves. Should we should we file let's let's ask Madi to look into this. Should we fire file something for the AI in NYC show?

Ryan Eppley: I mean, we're being scraped one way or the other. So whether it comes out, you know, blatantly or it's just a part of the training data.

Rob May: We I mean look, if if people will watch robotic knockoffs of this show, we'll take it, right? Like at this point trying to grow the traffic. [laughter]

Ryan Eppley: So, you know, as long as they keep you in your, you know, hat, right? They they got the Rolling Stones hat properly generated. It'll be fine.

Rob May: Yeah, I've had a couple shows without hats, but yeah. We would, you know, but we you know what we should do? We should get a company that makes humanoid robots to have one come sit in here with us one day.

Ryan Eppley: The only time I've ever seen it work is when it's like teaoperated though.

Rob May: Yeah. Like could we get one 100%, you know, non-taoperated actually, you know, autonomous?

Ryan Eppley: Probably can't say much. It'd be tough for it, you know. And it is strange. I saw last year I saw Palmer Luckey speak at Reindustrialize via a humanoid on stage. He was sitting in California, I believe, and the humanoid robot was in Detroit on stage with Ashley Vance. And it was a strange experience. It was definitely weird to watch a robot gesture and speak in an interview. Definitely uncomfortable. Maybe in a good way. Uncomfortable in a good way, I'd say.

Rob May: Yeah. Did it did Did it Did it cross the what what what do they call the uncanny valley?

Ryan Eppley: It definitely was like entering the uncanny valley, but because you know who's on the other side and it's the same voice. It's not like the robot had its own voice. It was the audio was Palmer Luckey. But just like a visual of a human sitting next to a humanoid as like it's a late night talk show. Mhm. And the humanoid is there acting as if they're the guest. Definitely was like starting to enter that uncanny valley like territory.

Rob May: Yeah. When you think about the late night shows, I think it's why I want to have a comedian on sometime because I do think humor is going to be one of the I it's one of the hardest things for humans. I think it's going to be one of the last things to go over to AI to make them funny, you know, because there's a it's there's a timing component, there's a surprise component, there's a like it's not it's not probabilities.

Ryan Eppley: Yeah. Yeah. There is there's a whole art to it as well. Though we have seen art was one of the first things that AI got really good at.

Rob May: That is true. You know.

Ryan Eppley: So, my girlfriend and I were just watching the new Kevin Hart show around like Funny AF, and it's a bunch of comedians. It's almost like American Idol, but for comedy, standup comed comedians kind of like vive for their place to get a Netflix special with Kevin Hart or something like that. Trying to imagine like AI like being a humanoid trying to like compete with all these humans. You're right. It would be awkward, but I honestly think it could play into it really well.

Rob May: Yeah. Cool. So, should we wrap up on a New York question?

Ryan Eppley: Yeah. What do you got?

Rob May: Actually, since Anna's not here, let's wrap up on an Anna question.

Ryan Eppley: On an Anna question.

Rob May: Yeah. Okay. Since Anna's not here, fa fa favorite trait of Anna Kirk?

Ryan Eppley: This might be a while for us to figure something. No, I'm just kidding. It's definitely when you know that she's very excited to speak on something, you can like see her like itching in a little closer to the mic from the corner of your eye at least cuz I usually sit over here. Yeah, it's very it's very good. She has great energy. So always love that.

Rob May: I would say something similar. I would say Anna Anna's curiosity is unbounded. She you can ask Anna you could be like I want to go if you're like we're going to go to this junkyard and crack open phones and look at whatever and see the thing. She'd be like yeah like let's go. Like she's just she's she's of the century. Yeah. So that's cool. We miss you, Anna. Be glad to have you back for the next show. Thank you guys for watching. Please rate us, send this to your friends. We have a newsletter now. Subscribe to the newsletter. We're going to start covering New York events. Send us your events. Related to that, if you know somebody that should sponsor the show, if you know somebody that should be on the show, if you yeah, have any ideas, let us know. And thank you for continuing to be part of this and for watching. Heat. Heat.