Erik: Joining me now is Freelancer.com, CEO, Matt Barrie. Really interesting story folks. How this one came together. We were trying to figure out who the top five macro guests were. You already know about the countdown. Lynn Alden was the big winner, okay? Believe it or not, those were just the macro guests, the number one all time download champion episode of Macro Voices was actually a Matt Barrie AI interview.
And the thing that's really striking to me is people like Luke Groman and Lynn Alden spent half of their careers building their following and promoting themselves and so forth. Matt's not even a guy who promotes himself yet he's done more downloads in a single episode than any of our guests that do, and it has to be this amazing topic of artificial intelligence. Matt, anytime something like this happens, public internet AI, it leads instantly to the obvious debate. Okay, is this a bubble that's about to burst or are we just getting started?
Or, where are we in this story and how should we think about it? Is, are we coming up on the dotcom bust or are we just getting started?
Matt: If you think back to the dotcom days, and there are a lot of parallels between then and now. Cisco got to half a trillion dollar market cap, which was about 5% of us, GDP.
And if you thought that is a bubble, what do you think about Nvidia at four and a half trillion dollar market cap and 15% of US GDP. It's become an absolute behemoth. It's about seven to 8% of the S&P 500 market valuation. It's doing about 6% or 7% revenue of the S&P500.
It is an incredibly profitable business. It does about $160 billion a year of revenue at about a hundred billion of EBIT, which is insane. It is built upon a very interesting foundation. The, they've released in the latest financials that the top two customers of Nvidia generate around 40% of the revenue and the next four customers generate another 40% of revenue, and it's believed to be that the top customer is Foxconn. And the second top customer is Quanta. And these are two Taiwanese manufacturers of boxes. They're put Nvidia chips in them. And number three is another Taiwanese company called Western. And below that, so your Dells and your Super Micros.
So you've got incredible customer concentration. The top six customers, top three being from Taiwan, but the top six customers are about 80% of the revenue. And top three customers are at 50% of the revenue. And then you look at their, product segments that NVIDIA produces, 80, 88% of the revenue is currently coming from So it's interesting.
And then out, out of the a hundred billion dollars of EBIT, the biggest supplier to Nvidia is TSMC, which has got, just like NVIDIA's got 60% concentration of the semiconductor market. TSMC has got 70% capture of the foundry market.
And so you've got this incredible revenue story, incredible revenue growth. That's all been on the back of data centers. It's all concentrated into six of which three are in Taiwan, and really, Nvidia story, which is really holding up the entire market, is built upon a fault wine, one island three customers one foundry and a prayer that the geopolitics that shakes Taiwan is less than the geophysics underneath the ground.
So it's pretty interesting thing to look at. And if you look at that, so you look at that revenue number of $160 billion a year, which is about 40 billion a quarter. The entire AI compute space. So we're talking Open AI, we're talking Midjourney, we're talking Anthropic, Etc. That entire AI compute segment in its totality is less than $40 billion of revenue a year, and every single one of those companies is losing money.
The entire ecosystem is really built around Nvidia prop, profitability, and and a whole Taiwan story underneath it. The more you dig into these numbers, which we'll do in a second, I'm sure the more the crazier the whole thing gets. And that's in the backdrop of a pretty frothy overheated market.
I mean of o f over 10 times that's already passed the dotcom peak, which was about 25%. And back in the dotcom days revenue growth was about twice as high, 20% year on year. That's where we are. And it's it's interesting with that basis of where we are today, of what the projections are for tomorrow.
Erik: Matt, this industry has already spent an incredible amount of CapEx.
It seems like the interest in it is unlimited. There's probably unlimited capital that wants to chase this. Wait a minute. You're gonna run outta energy at some point. We'll come back to that later in the interview. Is CapEx unlimited? Does it limited by energy? Is it limited by something else? When does this stop?
Matt: Obviously there's an incredible amount of money being spent not just on Nvidia, but in the CapEx to build the data centers, to put the machines that have Nvidia chips in them. Einhorn came out this week from Greenlight and said, this could be the biggest bonfire of money since the the dotcom boom.
If you look at, if you look at the amount of money that that the big sort of hyperscalers are spending, the likes of Microsoft and Google are spending at the moment around 50% of earnings on their CapEx. And the METAS and the Oracles and the Amazons are projected at the moment to spend about 70% of their earnings on CapEx with next year, potentially up to 1.3X their earnings.
If you think about the dotcom boom, and you think about, all the telco boom and bust that happened there, and the optical fiber boom bust that happened there, at its peak AT&T, when it was, dominating the market and it had about, I think it was about 60% market share back then of telecoms, it was spending at its peak about 72% of its earnings on CapEx. And, Exxon was spending about 65% of its earnings on CapEx on the shale, boom. And now come at the peak of CapEx. So we're really reaching the point of CapEx to earnings.
That hasn't been really paralleled since, the railroads were laid across America. And those numbers are getting to a pretty insane predictions. I think Altman came out just recently and then I think Jensen's kind of backing them up on this, on spending by 2030 7 trillion dollars on CapEx.
Now, if this sounds like a big number, it is, that's about one third of the M2 money supply in the us and how are they gonna finance that? A lot of, in fact, possibly most. The CapEx is being financed by ads. So what are you gonna do, you're gonna finance $7 trillion of CapEx on banner ads. The advertising market's kind of interesting to look at.
In the last a hundred years, advertising as a percentage of us, GDP has never really changed from about a 2% level. The composition of advertising in that spend has changed. Digital advertising now is the line share that's about 70, 75%. Of advertising spend, in the US it's 2% of GDP.
In Australia, for example, it's 2% of GDP. Worldwide it's about 1% of GDP, but in the rest of the world you don't really have a strong consumer market like you do in the US and so forth in the Western economies. So there's pretty limited ability for Meta and Google and so forth to increase I think their advertising. I don't think that increasing more ads in the feed or placing more banner ads on the page. Is going to be able to sustain a ever increasing exponentially growing CapEx spend. And if it's not ads that's financing this CapEx spend, it's cloud.
And so it's, you've got three major players in cloud. You've got AWS, you've got Microsoft Azure. And I was very surprised actually to see that in the last quarter. Microsoft's actually have taken Amazon. In terms of the revenue, it's about $47 billion in the quarter versus a AWS is 29 billion.
And then third place you've got Google Cloud and. For a long time, there were pretty decent margins. But now with, the ai, CapEx boom, those margins are eroding quite dramatically as the CapEx spends ramping through the roof.
And you've got now fourth and fifth players coming into the market. So you've got Oracle now trying to get into the market and they're trying to do it on price. So they're saying they're gonna undercut. The cost by about 40%. And then behind that you've got coreweave and the newer clouds coming in.
And if you reflect back on the dotcom days, AT&T had, one analyst I think said, better, better profit margins than drug dealing. When it was just three players in the market and they had 60% market share, and again that's the market share that Microsoft has in software, and that's the market share that NVIDIA has in ai.
And it's the market share that TSMC has in foundries. And now, if you look back at the telecoms boom and the fourth player and the fifth players came in, all the economics just fell apart because of the CapEx. You've got that dynamic happening now in the market , for AI data center build out.
And the funny thing behind all of this it's the demand isn't there. They're all building this out a ahead of time. And if the money's not coming from banner ads and it's not coming from cloud, it's coming from VCs.
So you've got, huge amounts of money sloshing to the space. The numbers are absolutely astronomical. Nvidia I said they're doing a one 60 billion a year in in, in revenue. And the rest of the AI compute space is making less than 40 billion in totality and and losing money.
Each of these big hyperscalers, they're spending, they anticipated to spend next year, a hundred billion dollars plus or minus 20 billion on CapEx alone. And they're not even doing that in revenue. AWS at the moment is doing about 110 billion a year in revenue. That's gonna be spent in its entirety next year on CapEx, just by each other's hyperscalers.
So these numbers are astronomical. In the back, in the dotcom days, at least when you had that boom the software market generated around 300 billion of revenue in its totality and employed about a million people. This AI boom is generating what, 20, 30 billion a year in revenue.
It's hard to tell 'cause all the companies are private. Losing money and not employing anywhere near as at that level. And in fact, if anything, they're saying, we're gonna take everyone's jobs away and create unemployment. The numbers are very fantastical and the more you dig into that, the more you dig into just how big those numbers are.
I just saw literally about an hour ago, Mark Andreessen tweeted a an interesting graph. It was, uS real GDP growth contribution from Tech CapEx and at the moment in the second quarter of 2025, the US real GDP growth is around 2% per annum, and half of that is CapEx from big tech. And they wanna lift that CapEx from 400 million to a $7 trillion spend by 2030 in totality.
Erik: Matt, as energy investing is my personal passion, and I gotta tell you this, AI one is a completely different ball game than anything else I've looked at.
I'm not really thinking about, okay, what's the demand of this and the supply of that? I'm thinking this fundamentally changes society in a way where potentially AI consumes so much energy and it needs it to grow that it. Exacerbates an energy price escalation that I think was coming due to inflation anyway, and all of a sudden you've got everybody feeling like AI is both stealing our jobs and consuming our energy.
It has to be outlawed, it has to be shut down, and I think it potentially adds to a social divide that's already outta control. So to me, there's almost nothing you could say about the scope of how much energy it's going to use or what the consequence of that. Could be, that would surprise me. I have to admit though, I'm a narrative oriented guy.
I haven't done enough homework to have any data to back the things I'm saying. Am I crazy to think it's that big? And how big is it?
Matt: In all these things, in all these bubbles, you have blue sky and then you have reality. And energy's where the reality practically starts to hit. Data centers are drawing huge amounts of energy from the grid and the US already, it's about 4.5% of the whole US energy demand.
And projection go to 9% by 2030. And this is at a scale now that it's causing real problems for a lot of different countries around the world. In fact, Bloomberg just came out with about a week or so ago with a, an analysis of energy prices in the US and they found within within vicinity of data centers, I think they looked at, within semi kilometers of a data center.
They did a survey of wholesaler energy prices in the United States, and they showed that within the last five years that those prices have gone up 267%. So you can imagine the average person in the street's probably not very happy with their, if their bills got, by triple to power someone else's chatbots.
At the same time, it takes a long time to put energy generation on the grid. It's highly political as you well know, you were just out here in Australia. Even though we're an energy superpower in five different capabilities, uranium, coal, gas, we've got some oil, Etc we're not allowed to use most of it.
And so it's very, we're shutting down coal plants, Etc, so it's very hard to get new supply on the grid. If the average data center now is what, 300, 400 megawatts gonna a gigawatt. If you try and put that on a lot of city grids, you'll brand it out. So in addition to grid cap capacity problems and energy, price problems, you've just got the practical reality of building these things, right?
Schneider Electric has an order book backed out till 2030. So I don't know how they're gonna get this CapEx, done. At the moment they're drawing down the capacity that's come from the office market rolling off. So as the work from home and the COVID thing hit, the office market in in the US and other places around the world, there is some construction capacity that's been drawn down to to build these data centers.
But, if you can't get the, electrical transformers and the switching equipment and you can't get the builder to, to be available at any reasonable price to build the data center. And then on top of that, AI demand is moving. Electricity markets like OPEC used to move oil.
You're gonna have some real problems in terms of your ability to actually deploy this. And then, at the end of the day. The economics aren't there because no one's making any money yet other than an Nvidia. So you've got this whole infrastructure that is built upon, very a great story a great promise.
And I do think in the long run, it's very similar as a dotcom boom. In that, I think I said this in the last, one of the last MacroVoices episodes, right? Cisco, networks. Networks. That's their motto. That's their tagline. As the internet got deployed in, in, in around the western economies, people thought, everything's gonna be on the internet.
Your clothing will be connected in a personal area network to your watch, to your handheld device, to this, that, the other, to your computer or to the world. And Cisco makes all the routers, the switches, the hubs, and ultimately the chips that will your network, the fabric of mankind and industry.
And so they're gonna be the richest com company in the world. You look at AI and it's a very similar issue where you go think AI's gonna power everything. It's great for highly personalizing experiences and really taking automation to the next level.
Which is the essence of computer science and, but we may be getting ahead of ourselves in the stories here. You've obviously got some personalities that are great. Showman, you're Sam Altman's, the, one of the PT Barnums of our time in terms of weaving a story and telling a story.
But the numbers now are so fantastical and so large and just so incredulous that you have real constraints in the physical world. Building up with numbers that get, you keep adding zeros to them and then you, at the end of the day, that may be way ahead of where the current usage is in time. Just we had the telecoms blow up in the dotcom boom.
Ultimately people, the law of compute computing basically is that, ultimately you'll use up on the memory, all the bandwidth, all the disk space, all the CPU capability, available to you. But it may take time, some time to get there, and you may have a massive o capacity for a period of time as people get over exuberant.
Erik: At the beginning of the public internet and the then the ensuing dotcom bubble and so forth, there was a lot of contention over completely unorthodox business models, Google giving all kinds of things away just to establish market share, which seemed crazy, but it was actually very effective. It seems to me like we're going through the same thing again, and I've gotta tell you, I went from Chat GPT3, I felt like it's a cool novel, but I'm not gonna spend 20 bucks a month on this thing.
Now at Chat GPT5, I've got the $200 subscription. I wouldn't go down to the 20 if you paid me to because I just, even for that small little edge I really value it and I would frankly love to, to be offered the $500 a month option because I've come to use it so much. Okay. If they're mostly only able to sell the $20 option, is this.
Business staying? How do they make money doing this?
Matt: It's clear that there's no sustainable business model with a foundational model. At this point in time, there's zero switching costs. If open AI, deep research.
O3 is better than, SONET, whatever from Claude I immediately switched my activity there and I'm sure you do the same as you just go to whatever the latest, greatest is. There's nothing holding you back. And what's very obvious though, as you become more of a power user, and I do spend a lot of my day admittedly in in GPT making all sorts of queries, Etc, when you kind of run outta, usage in your $20 plan or your 200 plan, or whatever the plan is. It the platforms don't say, please put in your credit card and buy more credit. They say you're on a timeout in the Nordic corner for six hours, seven hours, six. Six days, and in some circumstances I was given a several week time out on one of the platforms.
So you can tell that they're actually, the unit economics of inference is not there. They're not making money on you actually using the product. It's more like a gym membership. They're making money and you not turning up and paying the subscription. And this is I think a real tangible problem that you're getting like this incredible build out and data center usage for companies that are completely unprofitable and, you may be willing to pay $500 a month for greater usage of your model, I would think that would still probably be unprofitable in your usage at that level. Are you prepared to pay 2000 a month, 20,000 a month, or whatever the true number is for profitability? I don't I dunno. One, one of the, one of the big usage ca use cases that are coming out now, which is which is burning a lot of tokens are these sort of, AI enabled, in integrated development environments. So you see like your cursors and so forth where programmers can get, ordered complete on steroids effectively. Writing software cursor makes no money on on that token burn. They've got an incredible revenue ramp rate.
In fact, they're the one of the only three companies in the entire AI compute space that has. $500 million a year or more of annual recurring revenue and, no annual recurring revenue, not revenue. There, there's a lot of funny calculations being used to project the hype.
But 100% of cursors revenue is sent to Anthropic who provides the underlying foundational model to basically power their products. And, they're not making money's, not making any money. They're losing money, losing billions a year, Etc. And at this point in time, even companies like Perplexity is sending a room to send like 165% of their revenue to the underlying foundational model providers to run their search service. And then those foundational models, providers are sending it off to the cloud computing guys who are sending it off effectively to Nvidia, who sending it to TSMC.
So it's so what are the, so what do these companies have to do? So the open AI is the world. They've come out now saying, okay, we're not gonna make any money on 20 a month. We're not making any money on 200 a month. We have to come out with some sort of a story to justify these sky high valuations.
Now we'll note that Open AI has raised about $64 billion worth of money, I think today. And and they've done it at a $300 billion valuation. They're about to sell some employee stock. I think they've, they're upsizing the employee sell down round at half a trillion dollar valuation. So it's interesting, the open AI guys are cashing out, by the time that's done, they'll cash out about 11 point something billion dollars worth of stock, and they're only doing about, four mil, 4 billion of revenue in a quarter, sorry, a half or billion revenue a half, and losing eight in the first half of this year.
So to promise to, to pry their justification for these sky high valuations, they've gotta come up with these fantastical stories in the current one. I saw was Sam saying, I think about two weeks ago that 40% of jobs, so in order to justify these sky high valuations, and open ais employees sell down at a half a trillion dollar valuation of $10 billion worth of stock.
They've gotta come up with these fantastical stories. And the latest story I saw from Alman about two weeks ago was that 40% of people will potentially have their jobs taken. Thanks to AI and by implication of that, what he's saying is that open AI will generate the income that 40% of people would in the past be paid, and that's gonna be the revenue stream in the future as they take over all these different categories of work.
Now. If that were remotely true they would launch Jihad on on Sam. They'll burn down the Open AI headquarters and they'll put an ax to the data centers and governments would regulate and ban it. The Great Depression unemployment was around 32%. And so these numbers are extreme.
And to date, I still don't know. And I know there's dislocation that's gonna come in certain segments, but I still dunno, someone amongst any of my friends or even online who could put their hand up and truthfully say. I've lost my job thanks to AI despite all these stories around, agents and so forth and so on.
Now, I do think it's gonna come in some parts. I think, in customer support in some areas of sales, in some areas of administrative work where people have highly paralyzable task based workflows, which lend themselves very highly to automation and very good automation, I will say.
Because remember, the whole art of software is around automation. This is really just taking things really to the, a 10 x step up. And I think you can do some of those task-based functions very well. It's still not there yet. Nobody is making a profit doing this yet, but it is coming.
But it's not gonna be on the scale of 40% of the world people, the people in the world being unemployed. And you've got this, if you just step back and just look at the ecosystem, right? So you've got Nvidia 160 billion in in, in revenue, a hundred billion in earnings. They're making bank. And then you've got this, a whole ecosystem around them basically pumping Nvidia to the moon in terms of their revenue.
And by, by a result of that, their valuation. None of that around them is making any money. You've got open AI pitching all sorts of fantastical things. You can remember that they still have to revenue share with Microsoft. There's still a nonprofit this, that, the other, but they're, talking about, trillion dollars a year of CapEx 7 trillion by 2030.
There's the Oracle deal, which we should talk about in a second, Etc. And they don't have the money for any of this stuff. And it's almost. It feels much bigger than Enron in terms of the of the look and feel of the whole story around open AI and all the funny things that are happening now around the financings and the space and the justifications for how the CapEx is gonna be gonna occur.
And by example of that, there was what the calling now on social media, I think Elon called it the Infinite Money Glitch. You've got this announcement by Oracle that have just come out. The other day, Oracle share price jumped by about 39%, because they had like this it was like 380 or so billion dollars worth of cloud compute bookings that they've got out till 20 30 or so. Now, Oracle hasn't built their cloud yet. They're saying they're gonna build one, they say they're gonna do it, the price competitive approach of, roughly a 40% discount, Etc.
But then they showed this order book, and the order book was quite. Interesting actually when it came out, 'cause it was like a hockey stick, I think it was like 2026? It was like 18 billion in in, in bookings. And then the next year was 32 billion. The next year was 73 billion.
The next year was a hundred and 14 billion and 144 billion. And so it showed this huge hockey stick ramp and their share price went up 39%. But then a day or so later I think it was the Wall Street Journal published that 300 billion of that is OpenAI. Now open AI is only doing was it $4 billion a quarter of revenue.
They don't have 300 billion to spend. But what they've done is they've done the, they've done the, this deal with Oracle where they're gonna they've committed to this order book and off the back of that, Oracle share prices has pumped. And then in parallel, all of that.
So Jenssen's come out and he is now offering, this is the infinite money glitch. So he is offering a hundred billion dollars worth of GPUs to Nvidia under vendor financing. So he's effectively saying, take the GPUs now, pay me later for them. And then what OpenAI has done with those GPUs is gone to Corewave and Corewave is a neo cloud.
And we'll talk about them in a second. That's like the new generation cloud. AWSs and Azures and so forth, and it's gone to them and said, okay, I've got hundred billion dollars of GPUs. I'll give them to you to to, to use. And I'll book with you $22 billion worth of compute over the next couple of years.
And we take or pay. So I'll lock that in and I'll guarantee that I'll spend $22 billion with you. Corewave then go off and going, look at my bookings. I've got $22 billion of the bookings. And then open arrow on the flip side saying, look, I've got 22 billion of committed spend I'm gonna make on compute, so I must have a big customer pipeline.
So pump my valuation and give me a sky high, half a trillion dollar thing. So I can do $11 billion stock sell down. Then Corewave is saying I've got 22 billion of bookings from OpenAI. I'm gonna go to the private credit markets and JP Morgan and Goldman Sachs and so forth.
And I'm gonna raise money in the debt markets. And what I'll, from private credit, the lender of last resort. And what I'm gonna do with that is I'm gonna take that money, I'm gonna go back to Nvidia. I'm gonna buy the data center, gear that to build out my my AI, GPU powered, data centers.
And the money just goes around a circle. And at no point in time yet has a single customer spend a dollar. And so you, you look at all of that and you go if you send a check, check the numbers. And there's actually a really interesting blog called Where's Your Ed at which I encourage everyone to read where someone's gonna look, looked into this in far more detail.
And just going through the numbers, Etc and so forth. And if you believe the Oracle story, and if you believe about this 300 billion dollars in spend, then what that means is that in five years, Oracle. Is going single handedly grow the AI compute market 500%. They're gonna take it from sub 40 billion to 200 billion.
They are going to overtake AWS by I think 2029 with a hundred and AWS does about 115 billion a year in revenue. And by 2029, Oracle is predicting they're gonna do that from their cloud business and then overtake it to 144. So they're gonna do that and the whole thing.
Is gonna come from one customer being open AI who doesn't have the money. And I asked you before coming in here, I just checked Oracle's balance sheet. Oracle's a big business. They've only got a billion, 11 billion in cash. So these numbers of a hundred billion a year of CapEx for each of these companies.
400 billion up until now, 7 trillion being a third of the M3 money supply that some of is gonna be financed with ads or cloud or VCs. The numbers are just. Insane. And it is this, all this infrastructure and all this whole environment is now is built around Nvidia who's making all the money.
It's the only one making any money other than TSMC. And then you've got this new paradigm coming in with the the neo clouds, right? So the neo clouds that's Corewave, it's Lambda and it's Nebius.
Erik: Matt, hang on a second. Neo clouds, I have to admit, I don't know that one. Fill us in.
Matt: It's one of those things where, you know, from time to time I browse the internet and I see these new businesses that pop up that look like they're very large businesses, and I guess think to myself, I must be really stupid or just not understand.
And so you read through their website and so forth. And you go, wow, this is really cool. It's like really advanced. It's really huge. This is the whole industry I didn't know about. These are next generation cloud computing providers. So these guys are aiming to provide the GPU powered version of AWS or Azure.
And they're coming into the market as the fifth entrance, the sixth entrance and seventh entrance to compete against those incumbents. And, the interesting thing about these companies are obviously I've talked, just talked about the infinite money glitch, where it seems to be like a bit of a circle joke going between, in Nvidia Open AI CoreWeave, and they're all using the same money that's just sloshing around in a circle.
These neo clouds, they're like the WeWorks. Of GPUs, right? It's basically take taking some pre bookings. Use those pre-booking to raise some money. Rinse, repeat, and go in a circle. And these guys are building out infrastructure and they've got huge spend.
I think they're spending about $50 billion a quarter. Of those three, they're financed by the by Nvidia and by the the customers are Nvidia, so for example, super micro funded Lambda. And they're basically just building out data centers that are GPU powered and, doing so in a very much a buildup, and they'll come business model because to the best of anyone's knowledge at this point in time, not really making any money outside of your traditional open AI's measures and of course Nvidia itself.
And it'd be interesting to see where NVIDIA's going with this. 'cause because ultimately it feels that Jensen's heading towards. A, his own foundational model being, being launched on Nvidia hardware, but we'll see if that happens or not.
But the these Neo Cloud providers are basically trying to enter the market and through price pricing and through arguably a better products because it's using Nvidia GPUs rather than nutritional cpu, Etc. Basically trying to enter the market and, captures, potential huge revenue streams in the future that may or may not eventuate.
I think core has spent this year about $20 billion building up their infrastructure and it, and yeah it just feels, it feels like a very strange business model if you get down the bottom of it.
Erik: Matt, I wanna move on to the social implications of this. I am, as I get my head around this, I'm increasingly convinced this is much bigger almost than anybody is thinking about or talking about. And I'll use this idea of the competency crisis that people have talked about. Let's use Uber as an example. Back in the day getting a job as a limousine driver.
Required a fair amount of skill. You had to make sure you knew how to find addresses and so forth. They dumbed that job down to the point where it's just so idiot proof with a moving map in front of your face that shows you exactly where to go. That just about anybody can be an Uber driver. And my prediction is AI is gonna make it so that just about any numb skull whose gut.
Some kind of of device that they can use AI for is suddenly qualified to do all kinds of jobs that they might not even understand, but they can still get through it because they got AI to coach them. It's like they've got their own personal little coach for it. If that's where we're headed, I think it dumbs down society.
Why the heck would you want to graduate high school or be bother finishing, nevermind going to university. If a lot of the jobs that are out there, you just need to learn to operate AI and that's it. So is that where it's headed or what do you think the social implications can be of this just huge explosion of AI into society?
Matt: It's, yeah and in some regards, are we gonna see a huge spike in the Dunning Kruger effect as a result of everyone becoming an expert in 15 seconds from querying something on chat GPT, without doing the hard work to. Learn something over a period of years. Look, what I see is.
Oh generally what we are seeing, so I've got 83 million people in my marketplace who are using AI on for every type of job you can imagine now, we're seeing a big lift in skills, right? I think I've said this before, if you're an average copywriter, you're now an exceptional copywriter with GP or Claude.
If you're an average Illustrator, now you can become exceptional quite quickly with, your Midjourneys and the like. The same thing's gonna happen in all the white collar trades, whether you're a programmer or whether you're a, industrial designer or an architect or what have you.
So it does dramatically with the skills of people. I think it's like the world bef when you went to work and I was just at the cusp of this when I did the workforce, you you entered the workforce and there used to be no computer on your desk during the day.
And then you came to work and there was a computer on everyone's desk and all of a sudden everyone was a lot more productive and so forth. And yeah, you had some dislocation and a lot of jobs went away that the typing pools of secretaries and large companies disappeared. But, they got redeployed, Etc, ultimately to different jobs, Etc.
I do think that, skills lift. We're seeing a lot of productivity. We're seeing a lot of liquidity. So for example, our contest functionality where people can put up a prize and people compete for the prize. And we do that all the way from, simple things like logos.
We're up to, $6 million gene editing. Innovation challenge for NIH. We are seeing a lot of entries coming in a lot faster, a lot quicker, a lot higher quality. In fact, we had to recently rate limit that because we're getting over 700 entries per contest. So we're getting a liquidity effect and, and a speed effect as well as a quality effect and a bang for buck effect.
So all in all, you can probably encapsulate in the word productivity goes through the roof. But, yes. On one hand, if you're at university and you just to cheat by looking up Google the answers you know that cheating's now on steroids by getting GPT to do stuff for you, and that's certainly gonna continue.
If you get your, if you get your, big brother do your homework, or you get Google to do your homework, or you get wolf from Alpha to do your homework, or you get GPT to do your homework Yeah, and you're not learning the materials you're gonna really suffer. The flip side of that is that AI is going to be probably the most powerful forces in history for education in that.
It will be able to personalize a lesson plan for you and just what your pace is that you wanna learn at, and I know your, you, Eric, are very good at asking, I have the right questions to really maximize your learning rate on a new topic or a new area and so forth. And the same for me.
It does have the power particularly is, from the, the high school to tertiary education and above really empower the ability for people to learn new skills, new trades, new knowledge. Super quick. Now the question is just gonna be, and that, and this is one of the questions, the trillion dollar question that you kinda look at when, you know a lot of these AI guys hype these models to the moon is, will the AI be able to create new knowledge and do you independent scientific research and come up with new things and new inventions and will we enter this sort of hyped up, AGI super, super singularity sort of event.
Event horizon. Is it just a really good autocomplete that may be better than any autocomplete you've ever seen in your life. But basically. While it can find correlations and this, that the other it does struggle to come and find new advances, to push scientific breakthroughs.
That that's gonna be the big question. And I use a lit test for that. Will we see sort of creativity emerge from these models And, for example, will there be a song trending in the top 100 soon? That's gonna be completely done by AI designed, composed, produced, and executed and deployed by AI.
Are we gonna start seeing that with literature? Are we gonna see it, start seeing that with music? Are we gonna start seeing it with entertainment? If that starts to happen, we might also be there with scientific progress and technical advances, or is it gonna start capping out?
And there are real fundamental limits on access to data, which is fundamental. All the cheap oil of data has been mine, drilled out and used to train these models Now. Increasingly there's tariffs, rules, regulations, and restrictions on the ability to use it.
And is that gonna be, which is required exponentially more, with every generation of the models that are coming out. Is that gonna be the fundamental limit? And in fact, yeah, we get a computer in our desk, that computer is faster, better, 10 x better than you've ever seen before. And everyone's gonna be super productive.
But there is gonna be a limit from where it's gonna go from here, particularly in, in the practical reality that the underlying economics of this AI at scale is not actually working.
Erik: You mentioned something at the beginning of that answer I wanna come back to, which is the fire hose the excess amount of productivity.
I've felt this myself, Matt, on your platform. I've been using freelancer.com for several years now. If I need logos, graphic things I'm really bad at, just, hire somebody. It's amazing to me using the contest function a few years ago when I started using that. Boy, I could put up a hundred dollars contest and a whole bunch of well-qualified designers would submit over just a few hours.
I'd have a bunch of really interesting designs. I'd get to pick one, the guy gets a hundred bucks. It was just a great deal. Recently I thought, Hey I'm gonna up it to 300 bucks and I really wanna do a good one this time. What I had no idea was coming, Matt, is if you put a $300 contest on freelancers.com, you better have a staff of people ready to help you sort through the thousands and thousands of submissions you're gonna get.
It's that much, and it's literally to the point where I feel like I need AI. In order to filter all of the AI proposals. 'cause all of the sudden that freelancer community has so much capacity that I literally can't give each designer a fair shake or a fair chance because there's just so much coming.
And that's a good thing but it also. Presents a problem.
Matt: Yeah. That when you mentioned $300 in graphic design, I, you're gonna be, I could be overwhelmed.
Erik: I sabotaged myself by paying to and think about it, just a few years ago, 300 bucks, if you wanted a logo done by a professional graphic designer where you're gonna get one person's best attempt, and they were gonna put a few hours into it, that used to cost 500 bucks.
I only put 300 bucks on the contest and the result was. I almost couldn't use any result because in order to, not that I didn't get plenty of value back for my 300 bucks, but I didn't have time to sort through it all and pick one in time.
Matt: Yeah. We recently had to deploy something to actually rate limit the ability for people to, to enter in contest.
And and we were gonna use AI to actually sort through and so forth and rank and so forth, but also as the reputation system as well. We order all the entries now by basically the quality of past work of the particular entrance. But you are right.
We're definitely seeing a lift in not just the quality of work that people produce, but the speed at which they can actually do it. And ultimately the bang for buck bag for money. What we have is an incredibly deflationary business. I think NASA published a white paper while.
Between 80 and 99% of what they'll traditionally pay are going to traditional labor arrangements and not business, but crazy. We. Steps to make it manageable because there is an absolute explosion in productivity.
Erik: Matt, final question. Tell me about Sam Altman your sense of him, his sincerity, his intentions whether he's a good guy or a bad guy.
I, I have to admit, I can't read it. I thought his interview with Tucker Carlson was incredibly revealing. I'm not sure exactly of which. And there was a tweet also. Which really got my attention because I've seen this before. I know over, you guys in Australia went through this with social media and, oh we need to protect the children.
We we need to protect the children from profanity and bad things on social media. So therefore. Everybody is not allowed to use it until they identify to the government. With a government Id exactly who you are before you're allowed to log into your social media site. Sounds to me like Sam Altman is pushing for the same thing for AI, and I don't trust his intentions.
Am I being paranoid and more broadly, what do you think of this guy?
Matt: Yeah, I think you always have to take a look at the, what's been said and just think about it carefully and think through what, what's really going on. There's this, there's, I'll take this the second question first and then we can come back to the Tucker Carlson interview.
So he said that I think in the context of the was he was saying that they're gonna allow adult content on open ai and as a result of that, they need to protect the children and therefore they're gonna do age verification. There's a few things going on there.
Obviously Elon has been using, sexualization of content, I guess for not of better word, on Grok to try and encourage a bit of a cult following of the Grok AI.
Erik: But hang on, timeout there, there's just only one sane way to do that, which is to say if you want the adult content, then you can opt in at your sole personal discretion to, to show your ID.
That's not what he's pushing for. He is saying, everybody has to show their ID in order to be able to use the thing because they're not gonna, it doesn't make sense.
Matt: Yeah. So on one hand, Elon's kind of saying we can get a Grok that's uncensored. And I think I, I think if you look at the sustainable competitive advantages between the foundational a AI models, there's basically none.
They all do the same thing. They all can write a story for you. They all can. Give you advice, they can all generate an image, Etc. The only different competitive advantage is how censored they are and how left wing biased they are in the training, right? And Elon's kind of said you can come to, and, you can talk about anything and we're not gonna censor you.
And that, that's my advantage. Now maybe Sam is saying to himself the internet. I think last time someone published something, 80% of the traffic on the internet is PornHub. And so maybe on one hand, Sam's saying, okay, how do I get usage up of my AI?
And what are people potentially willing to pay for? I don't know. Maybe we'll go into kind of like the, pornographic style content and maybe that will get more people subscribing and more usage up I don't know. What I think is probably really going on is, in Australia, for example, we've got a effectively a censorship division of government which is run by someone who's nicknamed E Karen.
And in December of this year she is going the law comes in place which has already passed through government. Which is going to require, under the guise of protecting the kids. Every single social media platform in the world that operates in Australia will need to do age verification to prove that you are not under 16 now that has absolutely do nothing to do with protecting the kids, and everything to do with requiring digital ID to be rolled out in Australia so that everyone on the internet can be identified.
As the source of a comment on Twitter or the, on a YouTube video or whatever it may be. In fact, our Prime Minister, Anthony Albanese some years ago, there's a clip online where he was asked if he was ever gonna be dictated for a day, what's the one thing he'll do?
And he said, ban social media because he's not so thick skinned. And doesn't like. People about things on the on, on the internet about him. So this law is coming in now. The tantrums are going to be insane. Obviously. They just tried banning social media and was it Nepal and everyone rioted and set fire to parliament and so forth, but the ramifications of trying to stop people saying nasty singles on the internet is more than just, X and Facebook and Instagram and so forth. The eSafety Commissioner is now looking at whether to ban Roblox, Lego Land and GitHub because they're all places that people can leave comments, Etc.
So that shows you how insane this law is. And so by virtue of that, they've also said they're gonna require it for search engines. Now AI is basically a modern version of a search engine. And while we have not seen, I think yet substantial deviation away from Google and traditional search engines to the chat lock interfaces for search, that's basically what a lot of people are predicting.
And I believe the reason he's rolling out the age verification is because he knows that the ety regulator is gonna require, them to do age verification and, starer over in the UK is now trying to push digital ID as a excuse to of what he needs in place in order to solve the mass immigration problem that he's been running at full speed over there as well and it's causing all sorts of social unrest.
And he's saying these people coming across from Africa into the uk, we can't stop from working unless we have ID, although pulled up. Said they've gotta show a national insurance card, and if they're not doing that, they're getting work anyway. From, doing illegally working and not showing an insurance card, how's digital AI gonna help?
They're still gonna work the same way. But yeah, I think I dunno what's going on there. There may be a few things, a bit of confluence of events, but it feels that it's just getting ahead of that, 'cause there's quite substantial fines for the companies that allow kids to operate on them.
And it'll be interestingly where it plays out or whether they're actually gonna ban GitHub. Yeah, you I come home and watch YouTube at night sometimes, and I watch prospecting videos and geology videos and how to repair old electronic equipment or what have, it's a fantastic resource for kids and, but yeah, they're gonna ban it.
I get the feeling that they're just gonna apply that rule to OpenAI as well. In regards to your first question. I was pretty amazed by the Tucker Carlson interview. Isn't Tucker an incredible interviewer?
Erik: That interview pissed me off because we had this scheduled and I was thinking to myself, boy, we are gonna rock it, Matt.
We're gonna do the best AI interview ever done. And I, Tucker nailed, Tucker, blew me away. 'cause that guy is not technical at all. And he absolutely nailed that interview.
Matt: It was like a story a. Rivaling Breaking bad, watching that interviewer. Just how you set it up and it was, for those of you who haven't watched it, I highly encourage you.
You, you watch it. It's incredible. Work of art from Tucker. Yeah, it was interesting. Obviously all these guys who run these, multi-billion dollar companies talk a big game. They need to talk a big game because, whole venture space is kinda like a base of power law of investing.
You gotta kind swing for the bleaches, and you gotta get one investment to return a phenomenal amount of money to make any mo to make a return in your portfolio. So they all talk huge games, whether it's Elon, whether it's Sam, whether whoever it is. But yeah, it, and so you he started off asking him just things about, have you seen.
Higher powers, of some, is there something spiritual happening inside the AI that are they alive? Are they, and so forth And use, setting, setting Sam up and then and then got to the belief system. You know how, what, how's the belief system set up in, in these foundational models?
In, in chat GPT right? If it reads everything on the internet. You can see if you look at the YouTube comments that the sum total of humanity is actually quite bitter and and, argumentative and negative. And how do you get how does GPT train on the all the world and then end up, being so politically correct and so forth.
And it got down to the point where it's like, who inside OpenAI is deciding, what the belief system is of GPT? Is it you? I want names, give it their names and titles. And then of course yeah moved on to the. Copyright issue that's happening. I didn't mention before, obviously all the data in the world needs to be consumed to get these models to the next level.
And Anthropic just lost a 1.5 billion class action over copyright infringement because they were scanning in half, they scanned in half a million books without, paying the authors. And at $3,000 of work it's just been ruled to pay 1.5 billion. So there's a real compute cost there that's added on top with the training data.
But yeah, there was, I think there was a whistleblower inside open AI that was talking about potential copyright infringement or what have you from OpenAI. I don't the full details of that. But then that whistleblower ended up dead. And I don't know if I probably a preempt people watching Tucker's interview, but it was absolutely incredible the questioning that, that kind of came out around that and asking Sam what he thought of it.
And I don't know I thought the average person in the street, if you asked, did you kill someone? They would go, no. And then they'll go, and then don't. Stupid, you're an idiot. Sort of response, but.
In which that was an, I think Sam was completely thrown by the question. But, yeah, it was interesting how the responders came out.
Erik: I thought that was amazing. And as an interviewer, boy, I was just envious of Carlson because he nailed that. I guess we should explain what was going on is he was, Carlson was asking Altman about a very high profile murder of one of open AI's employees. And Sam...
Matt: or suicide? Hopefully suicide.
Erik: Oh, yes. I'm sorry. Murder or suicide. Could it be either? Sam tried to take the high road and show Tucker up and say, Tucker, come on. We're talking about the dead here. Don't you think we should have a little more respect for the subject matter? And he was trying to be the holier than thou, I'm gonna put you in your place Tucker, and make you look the bad guy.
And Carlson just kept his composure so perfectly and said, oh. Gosh, Sam, you would definitely be right. It would be, frankly, very inappropriate for me to ask that question at anything less than the behest of his mother who believes that you ordered his murder. So let's just skip to the chase. Sam, did you kill him or didn't you?
Or? It was something almost that direct and I thought, wow. I don't even fantasize about being that good at interviewing someday. On that note, let's let's end it there, Matt, because we are running out of time. freelancer.com is an amazing platform. That is your day job. It is just a phenomenal to me you pretty much upstage our entire episode 500 countdown.
With something that you do as a hobby. This is not even your day job. What do you do for a living in your day job and how can people learn more about how they can outsource a whole lot of cool stuff freelancer.com?
Matt: What, whatever you need done we can do it for you at Freelancer, right? So we make it real.
We turn your dreams into reality. What? You start wanna start a business. You want to grow your business. You wanna find people to build a product for you. We have 83 million people around the world with every skill you can possibly imagine. You'll get it done for you. A fraction of the costs, right? So you can just down a credit card and start the next.
I don't know, maybe start the next Open AI, start the next, Uber for cats, pizza review site that delivers whatever it is you can get done. The freelancer and freelancer.com. Just post your projects free, give it a go and you'll see the magic. And all the freelancers now AI powered and can get your job done, your product builds, your service delivered and a fraction of the time, fraction of the cost.
Erik: And I can definitely give a personal endorsement because I do use it regularly, and I would say it's not really a complaint, but the only really urgently needed repair at Freelancer is you gotta filter out how much you can get at a low price because it's overwhelming to the point that it, you just can't handle it.
Don't pay 300 bucks for a logo design. You will get three weeks of sorting through. Contest. Entrance be before you're done. Matt, I can't thank you enough. We're gonna wrap it there. And Patrick Ceresna and I will be back as Macrovoices continues right here at macrovoices.com.
