The man I am looking at is very hot. He’s got that angular hot-guy face, with hollow cheeks and a sharp jawline. His dark hair is tousled, his skin blurred and smooth. But I shouldn’t even bother describing him further, because this man is self-evidently hot, the kind of person you look at and immediately categorize as someone whose day-to-day life is defined by being abnormally good-looking.

This hot man, however, is not real. He is just a computer simulation, a photo created in response to my request for a close-up of a man by an algorithm that likely analyzed hundreds of millions of photos in order to conclude that this is what I want to see: a smizing, sculptural man in a denim jacket. Let’s call him Sal.

Sal was spun up by artificial intelligence. One day last week, from my home in Los Angeles (notably, the land of hot people), I opened up Bing Image Creator and commanded it to make me a man from scratch. I did not specify this man’s age or any of his physical characteristics. I asked only that he be rendered “looking directly at the camera at sunset,” and let the computer decide the rest. Bing presented me with four absolute smokeshows—four different versions of Sal, all dark-haired with elegant bone structure. They looked like casting options for a retail catalog.

Sal is an extreme example of a bigger phenomenon: When an AI image-generation tool—like the ones made by Midjourney, Stability AI, or Adobe—is prompted to create a picture of a person, that person is likely to be better-looking than those of us who actually walk the planet Earth. To be clear, not every AI creation is as hot as Sal. Since meeting him, I’ve reviewed more than 100 fake faces of generic men, women, and nonbinary people, made to order by six popular image-generating tools, and found different ages, hair colors, and races. One face was green-eyed and freckled; another had bright-red eye shadow and short bleached-blond hair. Some were bearded, others clean-shaven. The faces did tend to have one thing in common, though: Aside from skewing young, most were above-average hot, if not drop-dead gorgeous. None was downright ugly. So why do these state-of-the-art, text-to-image models love a good thirst trap?

After reaching out to computer scientists, a psychologist, and the companies that make these AI-generation tools, I arrived at three potential explanations for the phenomenon. First, the “hotness in, hotness out” theory: Products such as Midjourney are spitting out hotties, it suggests, because they were loaded up with hotties during training. AI image generators learn how to generate novel pictures by ingesting huge databases of existing ones, along with their descriptions. The exact makeup of that feedstock tends to be kept secret, Hany Farid, a professor at the UC Berkeley School of Information, told me, but the images they include are likely biased in favor of attractive faces. That would make their outputs prone to being attractive too.

[Read: We don’t actually know if AI is taking over everything]

The data sets could be stacked with hotties because they draw significantly from  edited and airbrushed photos of celebrities, advertising models, and other professional hot people. (One popular research data set, called CelebA, comprises 200,000 annotated pictures of famous people’s faces.) Including normal-people pictures gleaned from photo-sharing sites such as Flickr might only make the hotness problem worse. Because we tend to post the best photos of ourselves—at times enhanced by apps that smooth out skin and whiten teeth—AIs could end up learning that even folks in candid shots are unnaturally attractive. “If we posted honest photos of ourselves online, well, then, I think the results would look really different,” Farid said.

For a good example of how existing photography on the internet could bias an AI model, here’s a nonhuman one: DALL-E seems inclined to make images of wristwatches where the hands point to 10:10—an aesthetically pleasing v configuration that is often used in watch advertisements. If the AI image generators are seeing lots of skin-care advertisements (or any other ads with faces), they could be getting trained to produce aesthetically pleasing cheekbones.

A second explanation of the problem has to do with how the AI faces are constructed. According to what I’ll call the “midpoint hottie” hypothesis, the image-generating tools end up generating more attractive faces as an accidental by-product of how they analyze the photos that go into them. “Averageness is more attractive in general than non-averageness,” Lisa DeBruine, a professor at the University of Glasgow School of Psychology and Neuroscience who studies the perception of faces, told me. Combining faces tends to make them more symmetrical and blemish free. “If you take a whole class of undergraduate psychology students and you average together all the women’s faces, that average is going to be pretty attractive,” she said. (This rule applies only to sets of faces of a single demographic, though: When DeBruine helped analyze the faces of visitors to a science museum in the U.K., for example, she found that the averaged one was an odd amalgamation of bearded men and small children.) AI image generators aren’t simply smushing faces together, Farid said, but they do tend to produce faces that look like averaged faces. Thus, even a generative-AI tool trained only on a set of normal faces might end up putting out unnaturally attractive ones.

Finally, we have the “hot by design” conjecture. It may be that a bias for attractiveness is built into the tools on purpose or gets inserted after the fact by regular users. Some AI models incorporate human feedback by noting which of their outputs are preferred. “We don’t know what all of these algorithms are doing, but they might be learning from the kind of ways that people interact with them,” DeBruine said. “Maybe people are happier with the face images of attractive people.” Alexandru Costin, the vice president for generative AI at Adobe, told me that the company tracks which images generated by its Firefly web application are getting downloaded, and then feeds that information back into the tool. This process has produced a drift toward hotness, which then has to be corrected. The company uses various strategies to “de-bias” the model, Costin said, so that it won’t only serve up images “where everybody looks Photoshopped.”

Source: Adobe Firefly. Prompt: “a close up of a person looking directly at the camera”

A representative for Microsoft’s Bing Image Creator, which I used to make Sal, told me that the software is powered by DALL-E and directed questions about the hotness problem to DALL-E’s creator, OpenAI. OpenAI directed questions back to Microsoft, though the company did put out a document earlier this month acknowledging that its latest model “defaults to generating images of people that match stereotypical and conventional ideals of beauty,” which could end up “perpetuating unrealistic beauty benchmarks and fostering dissatisfaction and potential body image distress.” The makers of Stable Diffusion and Midjourney did not respond to requests for comment.

Farid stressed that very little is known about these models, which have been widely available to the public for less than a year. As a result, it’s hard to know whether AI’s pro-cutie slant is a feature or a bug, let alone what’s causing the hotness problem and who might be to blame. “I think the data explains it up to a point, and then I think it’s algorithmic after that,” he told me. “Is it intentional? Is it sort of an emergent property? I don’t know.”

Not all of the tools mentioned above produced equally hot people. When I used DALL-E, as accessed through OpenAI’s site, the outputs were more realistically not-hot than those produced by Bing Image Creator, which relies on a more advanced version of the same model. In fact, when I prompted Bing to make me an “ugly” person, it still leaned hot, offering two very attractive people whose faces happened to have dirt on them and one disturbing figure who resembled a killer clown. A few other image generators, when prompted to make “ugly” people, offered sets of wrinkly, monstrous, orc-looking faces with bugged-out eyes. Adobe’s Firefly tool returned a fresh set of stock-image-looking hotties.

Source: Adobe Firefly. Prompt: “a photo of an ugly person”

Whatever the cause of AI hotness, the phenomenon itself could have ill effects. Magazines and celebrities have long been scolded for editing photos to push an ideal of beauty that is impossible to achieve in real life, and now AI image models may be succumbing to the same trend. “If all the images we’re seeing are of these hyper-attractive, really-high-cheekbones models that can’t even exist in real life, our brains are going to start saying, Oh, that’s a normal face,” DeBruine said. “And then we can start pushing it even more extreme.” When Sal, with his beautiful face, starts to come off like an average dude, that’s when we’ll know we have a problem.

Just about everything you do on the internet is filtered through a handful of tech companies. Google is synonymous with search, Amazon with shopping; much of that happens on phones made by Apple. You might not always know when you’re interacting with the tech giants. Google and Meta alone capture something like half of online ad revenue in the United States. Movies, music, workplace software, and government benefits are all hosted on Big Tech’s data servers.

Big Tech’s stranglehold has lasted for so long that, even with recent antitrust lawsuits and whistleblower exposés, it’s difficult to imagine a world in which these companies are not so dominant. But the craze over generative AI is raising that very possibility. OpenAI, a start-up with only a few hundred employees, kicked off the generative-AI boom with ChatGPT last November and, almost a year later, is still making fools of trillion-dollar rivals. In an age when AI promises to transform everything, new companies are hurtling forward, and some of the behemoths are struggling to keep up. “We’re at one of these moments that could be a succession moment” for the tech industry, Tim Wu, a professor at Columbia Law School who helped design the Biden administration’s antitrust and tech policy, told me.

Succession is hardly guaranteed, but a post–Big Tech world might not herald actual competition so much as a Silicon Valley dominated by another slate of fantastically large and powerful companies, some old and some new. Big Tech has wormed it way into every corner of our lives; now Big AI could be about to do the same.

Chatbots and their ilk are still in their early stages, but everything in the world of AI is already converging around just four companies. You could refer to them by the acronym GOMA: Google, OpenAI, Microsoft, and Anthropic. Shortly after OpenAI released ChatGPT last year, Microsoft poured $10 billion into the start-up and shoved OpenAI-based chatbots into its search engine, Bing. Not to be outdone, Google announced that more AI features were coming to Search, Maps, Docs, and more, and introduced Bard, its own rival chatbot. Microsoft and Google are now in a race to integrate generative AI into just about everything. Meanwhile, Anthropic, a start-up launched by former OpenAI employees, has raised billions of dollars in its own right, including from Google. Companies such as Slack, Expedia, Khan Academy, Salesforce, and Bain are integrating ChatGPT into their products; many others are using Anthropic’s chatbot, Claude.

Executives from GOMA have also met with leaders and officials around the world to shape the future of AI’s deployment and regulation. The four have overlapping but separate proposals for AI safety and regulation, but they have joined together to create the Frontier Model Forum, a consortium whose stated mission is to protect against the supposed world-ending dangers posed by terrifyingly capable models that do not yet exist but, it warns, are right around the corner. That existential language—about bioweapons and nuclear robots—has since migrated its way into all sorts of government proposals and language. If AI is truly reshaping the world, these companies are the sculptors.

Some of Big Tech’s old guard, meanwhile, haven’t been at the forefront of AI and are scrambling to get there. Apple has moved slowly on developing or incorporating generative AI, with one of its flashiest AI announcements centered on the mundane autocorrect. Siri remains the same old Siri. Amazon doesn’t have a salient language model and took almost a year to begin backing a major AI start-up in Anthropic; Meta’s premier language model is free to use, perhaps as a way to dissuade people from paying for OpenAI products. The company’s AI division is robust, but as a whole, Meta continues to lurch between social media, the metaverse, and chatbots.

Despite the large number of start-ups unleashed by the AI frenzy, the big four are already amassing technical and business advantages that are starting to look a lot like those of the current tech behemoths. Search, e-commerce, and the other Big Tech kingdoms were “prone towards tipping to just one or two dominant firms,” Charlotte Slaiman, the vice president of the nonprofit Public Knowledge, told me. “And I fear that AI may be like that as well.” Running a generative AI model such as ChatGPT comes at an “eye-watering” cost, in the words of OpenAI CEO Sam Altman, because the most advanced software requires a huge amount of computing power. One analysis estimated that Altman’s chatbot costs $700,000 a day to run, which OpenAI would not confirm or deny. A conversation with Bard could cost 10 times more than a Google Search, according to Alphabet Chairman John Hennessy (other estimates are much higher).

Those computing and financial costs mean that companies that have already built huge amounts of cloud services, such as Google and Microsoft, or start-ups closely partnered with them, such as Anthropic and OpenAI, might be uncatchable in the AI race. In addition to raw computing power, creating these programs also demands a huge amount of training data, and these companies have a big head start in collecting them: Every chat with GPT-4 might be fodder for GPT-5. “There’s a lot of potential for anticompetitive conduct or just natural business-model pressures” to crowd out competition, Adam Conner, the vice president of technology policy at the Center for American Progress, a left-of-center think tank, told me.

These companies’ access to Washington, D.C., might also help lock in their competitive advantage. Framing their technology as powerful enough to end civilization has turned out to be perversely fantastic PR, allowing GOMA to present itself as trustworthy and steer conversations around AI regulation. “I don’t think we’ve ever seen this particular brand of corporate policy posturing as public relations,” Amba Kak, the executive director of the AI Now Institute and a former adviser on AI at the Federal Trade Commission, told me. If regulators continue to listen, America’s AI policy could functionally amount to Big AI regulating itself.

For their part, the four GOMA companies have provided various visions for a healthy AI industry. A spokesperson from Google noted the company’s support for a competitive AI environment, including the large and diverse set of third-party and open-source programs offered on Google Cloud, as well as the company’s partnerships with numerous AI start-ups. Kayla Wood, a spokesperson for OpenAI, pointed me to a blog post in which the company states that it supports start-up and open-source AI projects that don’t pose “existential risk.” Katie Lowry, a spokesperson for Microsoft, told me that the company has said that AI companies choose Microsoft’s cloud services “to enable AI innovation,” and the company’s CEO, Satya Nadella, has framed Bing as a challenger of Google’s dominance. Anthropic, which did not respond to multiple requests for comment, might be better known for its calls to develop trustworthy models than for an actual product.

A scenario which Big AI dislodges, or at least unsettles, Big Tech is far from preordained. Exactly where the tech industry and the internet are headed will be hard to discern until it becomes clearer exactly what AI can do, and exactly how it will make money. If AI ends up being nothing more than empty hype, Big AI may not be that big at all. Still, the most successful chatbots are, at least for now, built on top of the data and computing infrastructure that existing Silicon Valley giants have been constructing for years. “There is no AI today without Big Tech,” Kak said. Microsoft, Google, and Amazon control some two-thirds of cloud-computing resources around the world, and Meta has its own formidable network of data centers.

Even if their own programs don’t take off, then, Amazon and Meta are still likely to prosper in a world of generative AI as a result of their large cloud-computing services. Those data centers may also tip the power balance among Big AI toward Microsoft and Google and away from the start-ups. Even if OpenAI or Anthropic find unbelievable success, if their chatbots run on Microsoft’s and Amazon’s cloud services, then Microsoft and Amazon will profit. “It’s hard for me to see any Big Tech companies being dislodged,” Conner said. And if people talk to those chatbots on an iPhone, then Apple isn’t going anywhere either.

Then again, the social-media landscape had its dominant players in the mid-2000s, and instead, Facebook conquered all. Yahoo predated Google by years. Certainly, in the 1980s, nobody thought that some college dropouts could beat IBM in personal computing, yet Apple did just that. “If you bet against the online bookstore, you made the wrong bet,” Wu said, later adding, “Taking a look at the necessary scale now and extrapolating that into the future is a very common error.” More efficient programs, better computers, or efforts to build new data centers could make newer AI companies less dependent on existing cloud computing, for instance. Already, there are whispers that OpenAI is exploring making its own, specialized computer chips for AI. And other start-ups and open-source software, such as from MosaicML and Stability AI, could very well find rapid success and reconfigure the makeup of Big AI as it currently stands.

More likely is not a future in which Big AI takes over the internet entirely or one in which Big Tech sets itself up for another decade of rule, but a future in which they coexist: Google, Amazon, Apple, and the rest of the old guard continue to dominate search and shopping and smartphones and cloud computing, while a related set of companies control the chatbots and other AI models weaving their way into how we purchase, socialize, learn, work, and entertain ourselves. Microsoft offers a lesson in how flexible a tech giant can be: After massive success in the dot-com era, the company fell behind in the age of Apple and Google; it reinvented itself in the 2010s and is now riding the AI wave.

If GOMA has its way, perhaps one day Bing will make your travel plans and suggest convenient restaurants; ChatGPT will do your taxes and give medical advice; Claude will tutor your children; Bard will do your Christmas shopping. A Microsoft or OpenAI AI assistant will have helped code the apps you use for everything, and DALL-E will have helped animate your favorite television show. And all of that will happen via Google Chrome or Safari, on a physical MacBook or a Microsoft Surface or an Android purchased on Amazon. Somehow, Big Tech might be just emerging from its infancy.