Episode Categories:
Resources:
Join the discussion: Got something to say? Let us know on LinkedIn and network with other AI leaders
Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineup
Connect with Jordan Wilson: LinkedIn Profile
Uber’s $1 AI Training Tasks: The Real Blueprint for Future Workforce and Data Strategy
The recent move by Uber to pay drivers as little as $1 per task to train large language models is more than just a headline—it’s an early warning and blueprint for how businesses must adapt their workforce and data strategies in the coming years. As detailed in Everyday AI’s latest episode, this shift is a direct response to a fundamental issue in artificial intelligence: the scarcity of high-quality, human-generated training data in an Internet increasingly saturated by AI-created content.
Inside Uber’s Digital Task Program: Structure and Implications
Uber’s digital task program invites drivers to complete short, non-driving tasks such as recording voice clips, uploading photos, or submitting documents through the Uber driver app. Each task pays at least a dollar, with more complex tasks earning slightly more, and funds are deposited within 24 hours. Participation is optional, targeted at Uber’s existing pool of millions of US gig workers, and now expanding from a pilot in India.
These microtasks build valuable datasets for Uber’s internal AI solutions, and the company also sells this labeled data to enterprise clients and AI labs. Notably, Uber claims this initiative isn’t designed to support self-driving vehicle programs, though the nature of the data—speech, location, images—suggests clear utility in autonomous technology. The tasks themselves are notable for requiring no prior experience; a phone is the only prerequisite.
The Critical Shift: From Generalist to Specialist Data Labeling
Traditionally, data labeling for AI model training has relied on crowdsourcing platforms, featuring low-paid, global generalist workers. Companies such as Scale AI, Appen, and Sema have built billion-dollar businesses in this space. However, this model has led to both economic and ethical challenges, particularly with U.S. firms hiring overseas labor.
The trend is now shifting toward domestic and specialist labelers. This evolution is not merely a function of labor economics; it’s a strategic necessity. AI models require smarter, more context-rich datasets to avoid performance plateau and “model collapse”—the phenomenon by which models trained mainly on synthetic or AI-generated data suffer quality degradation over time.
“Dead Internet” Theory and the Value of Unique Human Data
Business leaders face a stark landscape. The majority of online content is now AI-generated, and the quality gap between models trained on novel human data versus synthetic data is growing. Recent studies cited in the episode indicate:
Over half of new online content is generated by AI.
Bot traffic now exceeds human traffic on the Internet.
By next year, most open training datasets suitable for AI will be exhausted.
Gartner estimates 60% of AI training data this year is synthetic—a figure whose upward trend threatens to undermine AI’s effectiveness.
This means any organization seeking to develop proprietary AI applications must shift from relying on public Internet scraping to actively building company-specific, human-created datasets.
Economic and Workforce Impact of Microtask AI Training
For gig workers, the short-term benefits are clear: additional income streams that don’t require professional experience and can supplement earnings between traditional tasks. Over the longer term, however, microtask-based work introduces significant uncertainty. By training models, workers may be contributing to the automation—and eventual replacement—of their own jobs. The line between supplementing labor and rendering it redundant continues to blur.
It is projected that commonplace full-time employment, especially for college-educated workers, will decline as microtask-oriented roles proliferate. Many future jobs may involve nontechnical work aimed at customizing, fine-tuning, and feeding proprietary knowledge into AI systems.
Strategic Imperative: Building Internal Data Pipelines
Enterprise firms must recognize that classic nine-to-five roles will taper off as data labeling and microtasking become essential functions. First-party company data—unique, exclusive information generated directly from business operations—will serve as the competitive bedrock for AI projects. Expect the emergence of internal “fine-tuning” teams whose job is to collect, curate, and label operational datasets for bespoke AI training.
This is not just a technical concern; it represents a larger business initiative and competitive advantage. Companies unable to source fresh, high-quality human data will be limited to using recycled, synthetic, or AI-generated information, eventually losing ground as model performance plateaus.
Secondary Ripple: Higher Education and Data Partnerships
Financial and enrollment pressures are forcing universities to rethink their approaches. As academic skills and training lag behind industry needs, particularly in AI, expect more partnerships or “aqua hirings” between higher education and technology firms. Universities may become key data pipelines—not only as training grounds for future labor but as sources for rich, human-generated datasets from research and classroom activity.
Final Takeaway: Action over Sentiment
The Uber digital task program highlights the urgent need to rethink workforce roles, internal data strategy, and the type of human capital business leaders should prioritize. The move toward microtask-based work is likely only the beginning. Organizations must adopt proactive approaches—creating in-house protocols for human data generation, reevaluating job functions, and ensuring they’re not just participants but architects in the next phase of AI adoption. Companies that ignore these shifts risk being left behind in a landscape where model quality, not quantity, will drive competitive advantage
Topics Covered in This Episode:
- Uber Digital Task Program Overview
- Uber Paying Drivers $1 for AI Training
- AI Data Labeling and Microtask Economics
- Impact on Gig Worker Job Security
- Human Data Training for AI Model Quality
- AI Model Collapse and Internet Data Exhaustion
- Future of Nine-to-Five Work and AI
- Enterprise Strategies for First-Party Data Collection
- Universities and AI Company Partnerships Prediction
Keywords:
Uber digital task program, Uber paying drivers to train AI, AI model training, $1 micro tasks, Large Language Models, AI data labeling, crowdsourcing human data, gig workers, AI gig economy, job automation, job replacement by AI, AI impact on workforce, autonomous vehicles, computer vision, Waymo, driverless cars, AI-powered ride sharing, AI data collection, first party data, AI data scarcity, synthetic data, AI regurgitation, model collapse, reinforcement learning with human feedback, incremental AI model improvements, knowledge cutoff, AI-generated content, dead Internet theory, bot traffic, Imperva bad bot report, Reddit data deals, Quora, AI partnerships, Fortune 1,000 companies, enterprise AI adoption, fine-tuning teams, domain specific models, nontechnical AI training, business leaders and AI, universities and AI, aqua hiring by AI companies, labor economics, global South cloud workers, AI training documentation, restaurant menus, voice clip recording for AI, photo uploads for AI, US economy shift, short term economic relief, long term economic uncertainty, AI slop, unique human created data, 2025 AI predictions, model quality plateau, agentic nature of AI, data curation, content scraping lawsuits, OpenAI, Google, Anthropic, Microsoft, Scale AI, Appen, Sema, benchmarking AI models, multimodal datasets, lecture hall data, college enrollment decline, university partnerships with AI companies.
Podcast Transcript
Jordan Wilson [00:00:45]:
Maybe you missed this recent headline. Uber is paying drivers $1 per task to train large language models. And I think that this has grabbed headlines in traditional media and has really set off a lot of discourse on social media. People saying, is this a good thing or is this a bad thing? Well, I think regardless, this is actually a sign of what's next. And regardless on if your take of if this is a good thing or a bad thing, I think this is going to become very commonplace in The US, especially as we head into a time where the internet is essentially dead and AI labs and companies need human data. All right. So on today's show, we're gonna be talking a little bit about this story, and I'm gonna give you my hot take on, yeah, get used to this. We're probably all gonna be training AI models to do our jobs.
Jordan Wilson [00:01:49]:
All right. Let's dive into it. What's going on y'all? Welcome to Everyday AI. My name is Stuart Wilson. I'm your host, and this thing is your daily livestream podcast and free daily newsletter helping everyday business leaders like you and me not just keep up with all these AI headlines like this, but how we can make sense of them and grab the most important information to help us make decisions that grow our companies and our careers. That's what you're trying to do. Great. Starts here with the unedited, unscripted livestream podcast.
Jordan Wilson [00:02:16]:
But if you wanna take it to the next level, that's where our website comes in. Your best friend, youreverydayai.com. Go sign up for our free daily newsletter where we recap the highlights from each and every episode as well as, give you what you need to know on the AI news. So if you want the AI news, make sure to go check out, today's newsletter. Should be an exciting, announcement coming from Google on vibe coding inside of AI studio, FYI. Alright. But let's talk about what we're going over on today's show. Yeah.
Jordan Wilson [00:02:46]:
Uber paying a dollar per task for humans to train AI models. Well, I'm gonna detail what Uber's new digital task program is and tell you how it work. I'm gonna tell you why I think it's both a good thing and a bad thing. I'm gonna preview what it means for the future of The US economy and more, and I'm gonna give you my hot take on actually what this means. Alright. Live stream audience, good to see you. If you have any questions, let me know. Go ahead.
Jordan Wilson [00:03:14]:
Get them in now. But, yeah, if you listen on the podcast, FYI, we do this thing live, unscripted, unedited. It's something I I I like to do. I think so many, you know, podcasts out there or, you know, sources of information are overly polished and, you know, we just like to give you just the real stuff. So good morning. Dino joining us from Italy. Jay joining us on LinkedIn from, Minnesota. Brian and Marie, good to see you all.
Jordan Wilson [00:03:44]:
Tim from the, YouTube machine. If you have questions on this or if you want your opinion to be heard, go ahead and drop it. But here is the new digital task program overview. Yes. Uber drivers are getting paid sometimes a dollar per task, to help Uber train AI models. They swear it's not for autonomous vehicles and to replace, the actual drivers who are completing these tasks, but I'm sharing a little screenshot here for our livestream audience showing about how it works. So the way that Uber is launching this, it's inside of their work hub, and drivers, can do these tasks for sometimes as little as a dollar, sometimes more. So Uber says they're quick and easy.
Jordan Wilson [00:04:28]:
Each task just takes a few minutes to complete. They say you can earn money on the side. So to do these tasks offline while you're not driving my gosh. I hope people are doing this offline while they're not driving. And they say no experience is needed. A phone is all you need. Alright? And then, drivers can log in and do these things kind of while they're not driving. So, this was just announced this past week, and the digital task program pays drivers small amounts to complete these micro tasks that ultimately help Uber train its own models, and they sell that data to others.
Jordan Wilson [00:05:03]:
It's optional, obviously, for drivers who opt in or don't opt in, but it's essentially for their, Uber's current, pool of millions of these gig workers who, drive or maybe are doing other gig work. This has expanded now just to The US after a successful pilot in India, and Uber uses the data to improve its own services, but they also sell this data to other AI labs. So here's how it works. Drivers opt in to receive invitations for these digital, digital tasks through the opportunity center in the Uber driver app. The tasks are reportedly quick and can be completed on a phone, often while away from the vehicle or just while they're idling, waiting for their next ride. Payments for each test start at just a dollar, sometimes more for more complexity. And then the earnings from drivers, you you know, kind of get added to their account within twenty four hours. This is normal, and I think it's actually important to have a conversation.
Jordan Wilson [00:06:05]:
We're not gonna get into too much into the weeds on data training, but it's important to know this isn't some one off from Uber. Right? There's dozens of companies valued at billions of dollars, multiple billions of dollars a piece that this is essentially what they do. They pay sometimes low wage workers from other countries, to make sense of data and to complete tasks. So, companies like Uber, Scale AI, Appen, and Sema, Essentially, they're crowdsourcing platforms that hire workers globally to perform these micro tasks that train AI systems. And many of these jobs, we've seen some, ugly headlines over the years. Well, oftentimes, they're low paid, you know, and sometimes they face payment or labor law issues, especially when they're, US based companies hiring, workers in developing countries. Firms are though moving away from these large, kind of pools of low paid generalists. Right? So, we saw some stories back in 2023 and 2024.
Jordan Wilson [00:07:11]:
A lot of these bigger companies were paying, ultimately workers a dollar or $2 an hour, but not necessarily specialists. They were just, sometimes anyone these companies could find, paying them a couple of dollars an hour. So now there's been this shift toward number one, at least for The US companies, bringing, some of this talent abroad or sorry. Bringing it, domestically, bringing it in house to The US, but also moving away from generalist to specialist, which is why I think it's important for all of us to hear this. Alright? And workers doing these tasks may be helping develop AI systems that could eventually automate or replace their own jobs. And I think this is gonna become, unfortunately, very common in the late twenty twenties, where you're you might have a role at your company, where it looks like you're helping your team better adapt to AI. You're, helping collect, data internally, first party data. And ultimately you might just be training a model that replaces your job.
Jordan Wilson [00:08:16]:
And this is a much bigger conversation, right? Than tackling Uber's, new digital task program. But it's actually both a opportunity and a threat for traditional nine to five work. So let's look a little bit more at what this program is actually offering up for these tasks. So the, examples that Uber drivers are doing, so they may be recording voice clips in their native language. They may be uploading photos of specific items or locations into the app and then submitting documents in various language such such as restaurant menus. So again, some of these tasks, you might look at them and say, okay, this is for autonomous vehicles. Uber wants to compete with, Waymo. Right? They want photos and, certain location based data that will help Uber better understand what it is they do.
Jordan Wilson [00:09:12]:
So I think part of it is true. And even though Uber says, hey. This isn't to replace jobs. I think, ultimately, if this is a successful program for Uber, I think it ultimately will lead to, replacement of jobs from these people that are doing it because that's where this industry is heading. Right? We don't talk a lot about autonomous vehicles. Right, Waymo and, what, Tesla is attempting to do, with kind of the robo taxis. We don't talk a lot about that. The whole computer vision autonomous vehicle, industry, it's huge, and it obviously has a big crossover with AI.
Jordan Wilson [00:09:46]:
But the future is autonomous fields. You know? Sure. We've been promised them for decades, I feel, and they're not really just becoming a reality, I feel, until 2025 where it's now commonplace at least in certain cities. Right? Whether you're in California, I think Texas is another big place, right, where it's kinda common to jump in a Waymo, and have an autonomous ride from point a to point b. So what is the purpose of this program? Right? I think the $1 task, kind of entry point is what rightfully so grabbed a lot of headlines. Right? Yes. These are, kind of micro tasks that may only take a couple of minutes to complete. So, yeah, it could be a good source of income.
Jordan Wilson [00:10:31]:
So I think, you you know, looking at the good and the bad of it, isn't a great short term money making opportunity for gig workers who are already Uber drivers. Absolutely. Right? So you have to look at a time where economically things are challenging. Right? I can't tell you how many recent grads, especially here in The US, just can't find jobs. Right? So the reality is a lot of them in the interim are maybe taking on positions that they may not normally take on that are outside of the area that they got their degree in. Maybe because colleges weren't teaching AI and not preparing them the necessary skills, but, I've already tackled that plenty here on the show. But what this has led to is a lot more gig workers. Right? Especially in the younger generation that are finding it harder and harder to find jobs in their area of study.
Jordan Wilson [00:11:26]:
So there's good and bad to it. Good is it does provide some, short term economic relief from people that are struggling to find full time employment or those who are already, you know, doing some gig work on the side, doing some Uber driving on the side. Now there's some more opportunities that don't necessarily involve them driving. Right? Things that they can do at home or, you know, between rides as an example. So short term, I think there's some good to this. Long term, in terms of normal economic security, there's not a lot of good. Right? A lot of people think and assume, oh, well, because this Jordan guy talks about AI every single day, you know, he wants AI to take all jobs and wants all companies to be AI native. Right? That's not necessarily the truth, but this is just the reality.
Jordan Wilson [00:12:17]:
It's not my personal opinion, but the reality, I think, is the nine to five is not going to be like it is in, you know, five or ten years. I do think, and I've gone in-depth, you know, when we kind of do our yearly prediction and road map series. I've said this for multiple years. Traditional nine to five work is going to eventually not be the norm anymore. Right? I think especially for college educated people, the nine to five, kind of career path has been the absolute norm for decades, and I don't think that's going to hold true. So, you know, kind of little programs like this, Uber's, you know, digital task program that are paying a dollar, I think things like this are going to become increasingly more common. Alright. So let's talk a little bit more about the program and then I'll get a little bit more to some of my takes.
Jordan Wilson [00:13:08]:
But the data collected is used by Uber AI solutions. So Uber does have a dedicated data labeling unit and they do use this internally, and they sell this to other AI labs in enterprise companies. But it doesn't come without its fair share of criticism. Right? So kind of how I opened the show, you've seen this on traditional media and on social media. It is kind of a polarizing topic depending on what your views are. Again, I think it's short term economic relief, long term economic uncertainty. Right? This is, I think, will be one of the earlier, kind of news stories where this concept becomes a household conversation, right? Where the average person will be training AI models in a very non technical way that eventually will lead to longer term job displacement or replacement. Yes.
Jordan Wilson [00:14:13]:
This isn't gonna turn into a long rant on, you know, AI's impact on jobs, but the reality is, yes, AI will create more, you you know, opportunities and full time jobs that don't exist today. Millions. Yes. But I've been on record for now almost three years saying that I think AI will have a net negative impact on job creation in the long run, and I think in a big way. But like I've said, I think the major it's it's gonna be very common for people to have multiple part time jobs or multiple businesses that they own in the future because of AI and because it's gonna make it, easier and even programs like this. Right? Uber's data labeling program that they're gonna sell this to, you know, other enterprise companies. It's gonna become easier for all of us, all everyone here listening to this show to launch your own business. Even if you aren't necessarily thinking of yourself as an entrepreneur, I think it's gonna be very common place, for this, to to happen.
Jordan Wilson [00:15:17]:
So this has launched this program, not the best timing, for Uber, but it's because it's launched amid concerns and criticisms about AI automation and job security, for gig workers as autonomous vehicle tech evolves, specifically in this Uber case. But Uber does claim that this data is not going to help them create self driving cars, but critics do note that AI data labeling has historically just been low paid work for cloud workers in the global South raising questions about labor economics and fair play. But my hot take is this. We both need this sorely, and it's absolutely terrible. Alright. I delivered you the facts and the stats. Let's get into my hot take here in a second after a quick word from our sponsors. This podcast is supported by Google.
Jordan Wilson [00:16:43]:
Here's the harsh reality of where we're at with data and large language models. Without getting too specific and probably too boring for much of our audience. Here's here's the way that it works, right? The big AI companies, your Anthropic, OpenAI, Google, Microsoft, right? They've essentially been scraping the Internet for anywhere from, you know, four to eight years, specifically with the goal in mind of training large language models. So that's the open closed Internet third party datasets, even copyrighted works. Right? We're seeing a lot of these lawsuits now finally pay off. Right? We saw the $1,500,000,000 fine levied against Anthropic, for reportedly training on copyrighted books. What this has led to, essentially, all of these big AI labs have kind of hit this, this point where there's no more really unique datasets to train their models on. And everyone's playing off the same data, at least that has historically been connected.
Jordan Wilson [00:17:59]:
But the data Internet theory is very real. Alright? And I'm gonna get into that here in a minute. But AI labs and the thousands of businesses that now rely on outputs from these AI labs sorely need unique human data that's not available anymore. It's already been scraped. It's already been ingested, regurgitated, spit out, and reused. Right? There's no unique data on the Internet anymore. You know, maybe maybe, you know, data will start to become a little more unique or exclusive. Right? As, certain big providers like, I'm gonna talk about Reddit here in a minute.
Jordan Wilson [00:18:39]:
But as they sign exclusive deals, with certain AI labs and once they can successfully block, all the other AI labs from accessing it. Right? We've already seen, many, big name lawsuits where essentially some of these big websites and media companies have entered into exclusive agreements with, you know, AI frontier company a have restricted AI company b. AI company b, doesn't pay attention. They still scrape that website, and now there's lawsuits. Right? But the reality is there's, you know, back in 2023, 2024, you know, there was a little competitive advantage for these AI companies that could, number one, not just successfully scrape everything. Right? All the, you know, legal and illegal ways that AI companies scrape data, but that if they had the right pretraining, if they had the right reinforcement, you you know, reinforcement learning with human feedback, if they could properly take the, essentially, entirety of of the Internet, entirety of human knowledge, and properly train a model on that. But I think that gap has shortened to essentially zero. Right? So I think earlier on in the case of, especially, 2023, you you know, there was an advantage to be had when you had the most talented people that could look at every single piece of human information that is scrapable, and could do the best job of training a model to provide good, examples.
Jordan Wilson [00:20:18]:
Right? Good outputs. That's not that's not a factor anymore. And right now, it's not just companies like Uber or companies, that are data, collection and curating companies. Individual businesses need to start thinking if you are in the c suite and an enterprise company. Right? Let's just say a fortune 1,000 company. And I know there's a lot of you out there listening to this program that fit that mold. If you're a decision maker at a fortune 1,000 company here in The U S if you're not already doing something like this, you have to start. Right.
Jordan Wilson [00:20:58]:
And I know what this means. This does mean you're gonna have employees internally that are helping with this process, that are essentially training models that are going to replace their day to day jobs. Yet companies have to do this. You are not going to be able to compete, in the next two to three years if you're not already collecting first party data in this way, in this example that Uber is having people do. The Internet's dead though. Right? There's a lot of analysis that I don't necessarily agree with how they came to their conclusions, yet the conclusions are overwhelming. Right. There was a a, kind of two more recent studies.
Jordan Wilson [00:21:46]:
So one, that bot traffic hit 51%, last year. So, essentially, there's more AI bots, perusing the Internet than humans, which is crazy when you think it's billions. Alright? And that's per, Imperva's bad bot, report. Also a more recent study. Again, I don't, necessarily agree with how they classified AI content versus human content, but regardless, there's been a lot of recent studies. One was a, a graphite analysis, I believe, that said over half of new online content is now AI generated as of this year. There's been other studies that have projected that more than 90% of content by the 2026 will at least be partially AI generated. You might be wondering, like, okay, Jordan.
Jordan Wilson [00:22:37]:
Like, what's the big deal? Why does this matter? Well okay, if more than 90% next year, if more than 90% of new content that is published on the web is somehow AI generated or AI augmented with a human creator, This creates this regurgitated cycle of sometimes AI slot. Right? And if you want your business to succeed, you have to be able to tell the difference between what is high quality data and what isn't. Right? Which is one of the reasons why Uber is even doing this in the first place. Right? Why they're having, who they hope are educated, humans making educated decisions. Right? So, the Reddit co find, founder, recently said this. So much of the Internet is dead due to bots and AI slop. Sam Oldman has recently said this, said so much as well. So why is this important? Like I said, companies have run out of training data.
Jordan Wilson [00:23:39]:
So according to an epoch, AI research, public training data could be completely exhausted by next year. Alright. Elon Musk said earlier this year that quote unquote, we've exhausted basically the cumulative sun of human knowledge. In other words, AI models have already consumed every single piece of recorded human knowledge there is. Right? Everything that's been published on the Internet, videos, works of art. Right? Because models are multimodal can ingest content in a multimodal fashion, offline datasets, etcetera. And Gartner says that even in 2024, 60% of AI training data was synthetic or and this is in 2024, so we'll see what Gartner releases next year. I would assume that number is probably in the mid seventies to 80%, but already the overwhelming majority of new information hitting the internet is synthetic.
Jordan Wilson [00:24:44]:
It is from AI. It is made up. It is AI generated. So I hope you can see this is a problem. It's this, I, I call it AI regurgitation. All right. The more, sophisticated name for this is model collapse. So there was a and we shared this in the newsletter when it first came out.
Jordan Wilson [00:25:07]:
A 2024 nature study proved that AI models essentially collapse when they are trained on their own outputs because that's what that's what's happening. It is AI slop regurgitated. Alright? Because when humans are getting lazy, and let's be honest, humans are lazy. And a lot of humans look at large language models as an easy button. Right? They wanna maximize time savings and put sometimes little less effort in creating something. And if all we're doing ultimately is using AI to create more data that will be consumed by AI and trained on for the next AI model. This is what leads to model collapse. In other words, models are not going to be as smart if there isn't a shift in strategy, and that's what we are seeing here.
Jordan Wilson [00:26:02]:
And that's what I think this Uber example is one of the first big shifts in companies going a different direction. Right? So essentially model collapse. Think of it like if you just keep photocopying a photocopy and you keep going and going and going after ten, twenty, 30 iterations of photocopying a photocopy, it's going to become unreadable, unusable. And that's what we're I think that's where we're at today. And that's why a solution like I said, there's good and bad out of this Uber, but this is technically what we need because AI companies and I think enterprise companies desperately need fresh human data to survive because model quality has essentially plateaued. Right? You can look at all the benchmarks. Every new model that probably cost billions of dollars to train is only now seeing incremental gains over previous models. And you might be thinking why? Well, essentially, if you look at how large language models are trained, there's something called the knowledge cutoff.
Jordan Wilson [00:27:05]:
Right? So essentially, these big AI labs, it takes a while. Right? So when we see, you know, a GPT five model that was released in August, I believe the model training cutoff was about a year prior. Right? So it takes all of these smart people at the big AI labs sometimes a very long time to look at their, hey. Here's what we've collected. Right? Here's all the data that we've scraped or we've entered into partnerships to, collect all this data, and there goes through this curation and cleaning and model training process. But by the time models, you know, quote unquote new models hit the shelves, It's already very old data. Right? Sometimes the model training cutoff is about a year or more for new models. And that's why when you see all these new models released.
Jordan Wilson [00:27:50]:
Right? SONNET four five, you know, GPT five. Right? I I I I do feel we'll we'll see, you know, a Gemini three point o here, you know, pretty soon in the coming, days, weeks, or months. But that's why now there's just such these small gains in a lot of these benchmarks. Right? Where eighteen months ago, you would see huge jumps. Now it's not. Competition is now, I believe, more about features and UX. It's not about intelligence anymore. I think that gap has closed because of the data and the model training.
Jordan Wilson [00:28:24]:
Right now, I think it's more about the scaffolding and the tool calling and the agentic nature than it is about the actual data that these models are trained on. Reddit is a great example. Right? Reddit is one of the most cited, at least when large language models cite their sources, right when they go out and, go out to the web and grab new and fresh information to answer queries, which is now the defacto way that most large language models work. You know, Reddit is one of the most cited or source pieces, or, authorities out there. And they make right now reportedly more than a $130,000,000 a year from AI deals. Right? So I think it was about $60,000,000 from Google, $70,000,000 from OpenAI per ad week, and that's a big chunk of Reddit's money. But and Reddit has also blocked, you know, such as the Internet, Internet archive, and other companies, from getting their data. But why am I bringing up Reddit as an example? Reddit is, well, humans.
Jordan Wilson [00:29:27]:
Right? Sometimes it's a great source for AI training or for outputs depending on what you're using a large language model for. Sometimes it's not. Right? You might not for certain business inquiries, you know, if you see Reddit as a cited source, you know, some random, you know, person in Ohio talking about, I don't know, a solution they found to a problem. Sometimes it's good, sometimes it's not. But human unique human data in the case of Reddit, Quora, etcetera, is extremely valuable. So let me start to wrap up today's show and why I think Uber's $1 digital tasks are actually a sign of what's next. This is not a, you know, one off small story that everyone's going to forget about. I think this is going to become the norm in the coming years, and here's why.
Jordan Wilson [00:30:21]:
And I've been saying this for a long time. In 2023, I said full time nine to five work is gonna become a thing of the past pretty soon. People thought I was crazy until LinkedIn CEO Reid Hoffman said the same thing a year later. And I do think that many jobs in the latter part of this decade will be kind of similar to what we're seeing now from this Uber $1 digital task. Your full time job may, in two years, be training a model in a nontechnical way. I think there's actually a huge, a huge business opportunity, to bring, kind of the UI or UX of what Uber is doing to companies. Right? Nontechnical ways to get your skilled humans to train, to train models for your own company's use. Right? And I also said this in 2024 and, again, more deeply this summer.
Jordan Wilson [00:31:20]:
Companies, like I said, many companies are already doing this, the big Fortune one hundreds, but I think this is for enterprise companies in The US. This is gonna become the norm. You are gonna have a whole department pretty soon doing similar tasks, right, that are obviously domain specific, category specific, for your company. But you are going to essentially have fine tuning teams, right, that are gonna have to take work with these large datasets and essentially train models. And I do think small, you know, small categorical or small niche models are the future, for enterprise. I've been saying that for a long time as well. But I think this is gonna become commonplace to have teams of people working with first company data. It's gonna be huge.
Jordan Wilson [00:32:06]:
And, also, I think colleges, let's let's be real. Enrollment's already starting to decline. It's going to hit a cliff, probably here in two years. Companies or or sorry, colleges and universities are struggling financially because enrollment is going to start to dip. And because for the last three years, colleges and universities has have essentially stuck their head in the sand, when it comes to preparing students for the real world because companies want AI skills. Universities aren't teaching this. It's gonna be a huge backlash. Alright.
Jordan Wilson [00:32:40]:
I already did a full episode on this, but, essentially, I think you're gonna have these big AI companies aqua hiring universities. Right? Universities are going in the same way. Right? I'm a big North Carolina basketball fan, and North Carolina is kind of sponsored, by Jordan Brand. Right? So they're, you know, a jump man athletic department. I think the same thing is gonna be true for universities as a whole, especially those universities that aren't thriving right now and are facing enrollment problems. They're essentially going to have to take on in, investments from AI companies or to essentially get aqua hired, by big AI companies or, data companies in order to survive. But I think that there's a huge source of human content. Right? You need right? Like, what happens in lecture halls, you you know, amongst doctoral students debating, certain certain issues of today, you know, trying to explore, new scientific breakthroughs.
Jordan Wilson [00:33:39]:
There's so much unique data happening both inside companies and inside universities, which is why I think this $1 digital task from Uber is just a sign of what's next. And it is going to start to unfold, maybe not tomorrow, maybe not next month, maybe not next quarter, but I can guarantee you in 2026, we are gonna see stories like this from companies, from universities. It is going to be the norm. So if you are a business leader thinking of what is the best way to use and implement AI within your organization. I think this Uber example is both absolutely terrible and absolutely crucial to understanding the future of how to work with unique human created data when AI models and the training data is essentially just going to become and has become regurgitated AI slop. So this is kind of a blueprint for succeeding and getting ahead in the era of the dead Internet. Alright. I hope this was helpful.
Jordan Wilson [00:34:52]:
Y'all, if it was tell someone about it, tell someone about it. All right. We put in a lot of work to bring you usually unbiased, right? This is little hot take Tuesday episode, talking about a recent news piece and giving you my, kind of hot take. But we put a lot of work into everyday AI so you can have a place for nontechnical business leaders to come and just get the no BS version of what's happening in the world of AI. So if this is helpful, please, if you're listening on Apple Podcasts, on Spotify, please subscribe and like the show. And if you haven't already, tell someone about it. Share about this. Repost this if you're listening live on LinkedIn.
Jordan Wilson [00:35:31]:
And then when you're done, the most important step, please go to your everydayai.com. Sign Sign up for the free daily newsletter if you miss anything. If you need to hear more on this, we're gonna be recapping the highlights in today's free daily newsletter as well as keeping you up to date and making you the smartest person in AI at your department or your company. Thanks for tuning in. Hope to see you tomorrow and everyday for more everyday AI. Thanks, y'all.