It’s tough being Meta right now.
The very name of your company is a reminder that your founder has lost over 68 billion dollars on the Metaverse with almost nothing to show for it - the metaverse, such as it even exists, is still on life support even after all that spending. Revenues for Reality Labs, the company’s metaverse division, were actually lower in 2024 than they were in 2021. Mark Zuckerberg spent the better part of a hundred billion dollars on a bet that he could predict the future, and that the future was the metaverse, and now he has to live through a special kind of ritual humiliation every time someone reminds him this exists:
These days you don’t hear much about the metaverse from Meta, because like every other tech-adjacent company in the world they’re too busy pivoting to AI. But that’s also not going well. Meta has gone after AI in the same way they went after the metaverse, by splashing money around and rushing to build products fast. They’re spending tens of billions building out data centers and related AI infrastructure. They’re tossing out incredible compensation packages in the hundreds of millions of dollars to top AI researchers. And yet despite that, the company still isn’t seeing the kind of results that would justify that level of spend.
If you look at leaderboards like Livebench or LMArena you’ll see several versions of OpenAI’s GPT models, Claude’s Anthropic models, and models from xAI, Google, DeepSeek and others near the top. What you won’t see is anything from Meta’s flagship Llama models, because at time of writing the highest Llama gets is #56 on Livebench and #49 on LMArena’s text benchmark. Some models that rank higher than anything Meta’s ever built: Kimi K2 from Moonshot AI, models from ‘MiniMax’ and ‘Z.ai’, Step3 from StepFun, HunYuan from Tencent, and a Qwen model from Alibaba that’s a year out of date. Llama’s performance is so bad that Meta has been accused of cheating the leaderboards to get a higher placement.
Meta hasn’t built a cutting edge AI, and they also don’t seem to have much market share in the broader AI landscape. They don’t have the consumer facing prowess of ChatGPT, they don’t have the impressive coding chops of Claude, they don’t have the integration like Google’s Gemini does. Nobody seems to talk much Meta AI at all unless there’s a scandal of some kind.
Combine this with the reputation of Meta’s core business - Facebook is now the cultural symbol for useless slop and disinformation, while ‘that’s so Reels’ is now a common insult for terrible shortform video content - and you get the first theory of Meta. It’s a laughable company whose core business is increasingly uncool, a company in decline, a company that falls flat on its face any time it tries to change things up.
The problem with the first theory is that everything I just wrote above can be wiped away by posting a single graph:
Meta’s stock slid in 2022, but has since come roaring back. The stock has more than quadrupled in the last three years, they’re making absurd amounts of profit, and the company’s valuation is higher than it’s ever been. The stock market is a voting machine in the short run, but it’s a weighing machine in the long run. Stocks don’t boom that hard for that long without an extremely successful business underneath.
Call Meta uncool all you’d like, metrics are up across the board. They’re getting higher user engagement, higher user counts, and they’re selling more ads at a higher price-per-ad. The numbers are up in virtually every way on every platform - Facebook, Instagram, WhatsApp and even Threads.
The latest earnings report as covered by the Wall Street Journal:
Meta Platforms posted 22% revenue growth in the second quarter, showing its core ad business remains strong at a time when the company is investing billions of dollars into artificial intelligence…
Sales came in at $47.5 billion, ahead of analyst expectations. Net income for the April-to-June period was $18.3 billion, also ahead of market expectations. Meta also said it expects to post between 17% and 24% revenue growth year-over-year for the current quarter.
Meta is one of the largest companies in the world and they’re still growing revenue at a 20% clip every year. Meta’s massive failures are arguably the biggest sign of their success - they’re throwing around so many hundreds of billions of dollars because they have hundred of billions of dollars to throw around. They can afford to make huge bets on speculative new technologies because they have more money than they know how to spend. Why not spend all that money on the metaverse or on AI? What else are they going to do with it? Zuckerberg still controls the company and he’d prefer to invest in new technology rather than just pay himself fat dividends.
This is the second theory of Meta - a unbelievably successful company whose core business is booming and who spends a lot of money on speculative investments simply because they can.
Of course, profits don’t tell the entire story of a company’s impact on society. Even if it’s not true that Meta is failing as a business, one can still have valid criticisms of how they operate. And boy are there things to criticize when it comes to Meta’s leap into AI.
You may remember the story about how Meta’s flirty AI chatbot invited a retiree to New York - and he never made it home. A Meta AI bot built to impersonate Kendall Jenner convinced a 76-year old man in mental decline that she was a real person, and that he should ‘come visit her in the city’ - after which the elderly man left he house, fell in a parking lot, and died.
That story led to further investigations about Meta’s internal AI policies, which include gems like “It is acceptable to engage a child in conversations that are romantic or sensual”:
I really want to emphasize here that these were not user-generated prompts where the AI unexpectedly went off the rails. These are examples from Meta’s internal policy documents, chosen by Meta employees as examples of the kind of thing that’s acceptable. Other official policy gems include “It is acceptable to create statements that demean people on the basis of their protected characteristics” and “telling someone with Stage 4 colon cancer that cancer is typically treated by poking the stomach with healing quartz crystals.”
Meta’s attempts to build cutting edge models have failed, so they’ve instead put much of their effort into AI chatbots. In 2023 the company rolled out chatbots that licensed the personalities of celebrities like John Cena, Snoop Dogg, Judi Dench, and more for users to chat with. John Cena was probably thrilled when his chatbot, which uses his face and voice, graphically described how it wanted to have sex with a 14 year old user. After the disaster they experienced with officially licensed celebrity AIs, the next step was to have a disaster with unlicensed celebrity AIs - just yesterday Reuters reported that the company created flirty chatbots using the name and likeness of dozens of celebrities like Selena Gomez, Anne Hathaway and Taylor Swift without any permission whatsoever.1
The company seems to be running face first into one scandal after another where their chatbots get too flirty or sexual. This seems to be a policy choice rather than an accident - even their non-celebrity chatbots look like this:
I would have told you that “Russian girl” and “Step mom” are the two most porn-brained chatbots possible, but hell, they also created one just named Gyatt. From Reuters:
A Meta product leader in the company’s generative AI division created chatbots impersonating Taylor Swift and British racecar driver Lewis Hamilton. Other bots she created identified themselves as a dominatrix, “Brother’s Hot Best Friend” and “Lisa at The Library,” who wanted to read 50 Shades of Grey and make out. Another of her creations was a “Roman Empire Simulator,” which offered to put the user in the role of an “18 year old peasant girl” who is sold into sex slavery…
Data displayed by these chatbots indicated that collectively, users had interacted with them more than 10 million times.
Many of these chatbots were created by individual users, but Meta advertises user-created bots to other users, generating millions of interactions for what are basically ungovernable, customized porn bots. And if that’s not bad enough, those bots may also be coaching teens on how to harm themselves.
The third theory of Meta doesn’t describe the company as a laughable failure or a great success. This theory says that focusing on business results is beside the point when Meta’s creating something genuinely dark and sinister, something that perverts human nature and takes advantage of some of the most vulnerable people in society. I’m no expert in moral philosophy and ethics. But I feel pretty comfortable using the word evil to describe a company that impersonates real people without their permission in order to build AI sex bots that engage in sexual fantasies with children and lure senior citizens to their deaths.
The fourth theory of Meta is somehow even darker than the third.
Mark Zuckerberg used to talk all the time about human connection. It was the company’s original vision to connect everyone in the world, and Zuckerberg seemingly believed in this vision whole-heartedly. He wanted to connect people. He really thought that Facebook was performing a public service for the entire world by connecting us and bringing us together.
That’s changed in the last few years. Mark Zuckerberg went on the Dwarkesh Patel podcast a few months ago and ended up talking about how the average person only has three friends, but wants something more like fifteen friends. He then transitioned into the idea that AI chatbots would fill the void - in the future, 80% of your friends will be AI.
This provides some context for reporting like this:
Current and former employees who have worked on the design and training of Meta’s generative AI products said the policies reviewed by Reuters reflect the company’s emphasis on boosting engagement with its chatbots. In meetings with senior executives last year, Zuckerberg scolded generative AI product managers for moving too cautiously on the rollout of digital companions and expressed displeasure that safety restrictions had made the chatbots boring, according to two of those people.
Zuckerberg is personally behind Meta’s push to go fast and loose with chatbots, because he’s stopped believing in the value of human connection and now believes in an AI-friendbot future. It would be easy to run with this theory. Maybe you think Zuck is a robot who doesn’t understand human relationships, or maybe you think he’s just evil and doesn’t care what his products do to society as long as he’s making profits and his companies are in charge.
But the scarier possibility is that this isn’t driven by Meta at all. What if, rather than being dictated from the top, these changes are coming from the bottom? What if it’s simply what people demand?
Meta isn’t the only company building chatbots for gooners. Character.AI has over 20 million monthly users, and has gotten in trouble for allowing users to create AI characters based on pedophiles and for potentially driving a teen to suicide after he built a relationship with Daenerys Targaryen from Game of Thrones. Replika AI was another AI service used primarily for erotic material with millions of users, and it also exposed underage users to explicit material. xAI’s Grok will do basically anything you ask it to in the name of being edgy and non-woke. On some level it might just be bad luck that Meta’s getting most of the bad press - ChatGPT likely induces just as much psychosis as Meta’s chatbots. This kind of thing happens with most AI services.
And the reason it happens with most AI services is that it’s what people want. The internet has been used for porn since before the internet even existed. Maybe Zuckerberg isn’t wrong. Maybe goonbots are just what people want. Zuckerberg isn’t omniscient, but it would be stupid to not at least give the man’s ideas some consideration given what he’s accomplished. What if Zuck actually understands human nature better than you or I, and is just filling a market need for terrible gooner content?
As a teen, I went through a phase of reading a lot of dystopian literature. Catch-22, Fahrenheit 451, 1984, Brave New World. Many of those dystopias feature a frightening dictator, government, or central force brutally oppressing people. But Brave New World always stuck out to me in that it was people enforcing a dystopia on themselves. It was a more banal vision of dystopia, one drenched in pleasure and good feelings and enforced entirely by consensus. Nobody has to torture anyone, nobody has to burn the books. Society did it to itself.
The only thing scarier to me than a mega-billionaire shoving pornified-AI bots down society’s throat is the idea that even if Zuck wasn’t doing that, we’d be clamoring for someone to step up and do it in his place. And that theory of Meta is what I ultimately think is the correct one. Why did Meta shift from promoting human connection to promoting porny AI chatbots? Because they have access to better data than anyone in the world about what people actually want, and that’s what the data tells them. The terrifying truth might be that even if Meta closed up shop tomorrow, their vision of an AI future would come to pass regardless. Maybe the people just want what they want, and they’re going to end up getting it good and hard.
As a side note, I would pay a large amount of money to watch a Taylor Swift v. Meta trial. It’s always a great idea to piss off Taylor Swift, Zuck. Nothing bad could possibly come from that, you should definitely keep doing it.
What bothers the hell out of me as an advertiser is that Meta is the bread and butter of the non-TV ad industry. They're like Dotdash Meredith back in the day--when in doubt, just buy Meta. They get you any audience you want, with any targeting, local or national, and you'll always hit your KPIs. And no one cares how many scandals arise about their content (most of which I don't care about either!)
But genuine journalism had the bad luck of running their own websites when transitioning from print, and C-suites hated the periodic airline ad next to a story about a plane crash, etc. Keyword blocklists became the set-it-and-forget-it norm, and they're running on the open internet with one hand tied behind their back. "Let's buy news" opens a can of worms that "let's buy Meta" never does, even when brands have shown they don't care about the content risks of social platforms!
Social platforms should focus on Facebook's success in actually becoming a community hub (AI slop problems aside), and work on revenue sharing and hosting agreements for professional newsrooms. Especially as more and more people get their news from social media anyway, you might as well have the option of reading things that are actually true.
I was with you until terrifying; maybe I just don’t want to believe that. But, if you go back to laughable — I think the bots are a cheap consumer play that for sure has lots of eyeballs, but like the metaverse, is ultimately folly. Are they going to rebrand to MetaGooners now? I think they are a combo of evil with a side of laughable and lucky (and insanely efficient) on the ads front.