WR 49: New year, new AI profiles (or not)
Weekly Reckoning for the week of 6/1/25
Welcome back to the Ethical Reckoner. In this Weekly Reckoning, we’ve got stories from across the globe. We start with a few AI-generated profiles in remote corners of Instagram, then look at how social media companies are preparing for Trump 2.0. Then, we look at pig butchering crackdowns in Nigeria (crypto scams, not the meat processing) and open-source LLMs in China that are outcompeting the Big Tech models. Finally, my New Year’s resolution was to get rid of weight loss ads in my social media feeds: find out how I did it in the Extra Reckoning.
This edition of the WR is brought to you by… getting back into the swing of things (slowly)
The Reckonnaisance
Meta’s AI profiles get canceled
Nutshell: After Meta announced that they’re planning to have AI-powered profiles on Facebook and Instagram, early tests were discovered and roundly criticized.
More: Last year, Meta announced a test of some “AI characters,” including some celebrity ones. People basically totally ignored them, until last week when Meta’s VP of product for generative AI said that AI-powered accounts are the future of the platform: “We expect these AIs to actually, over time, exist on our platforms, kind of in the same way that accounts do… They’ll have bios and profile pictures and be able to generate and share content powered by AI on the platform.” The old accounts hadn’t posted since last spring, but that didn’t stop netizens from re-discovering them (and sometimes assuming they were the new accounts). One called Liv, whose bio described her as a “proud Black queer momma of 2 & truth-teller” and “your realest source of life’s ups & downs,” received particular attention for being a “particularly offensive caricature of what a gigantic corporation might imagine” her persona to be like. Meta is now deleting the old character accounts.
Why you should care: If this is the future of social media, it raises a lot of questions, including: what will our timelines look like now? Will they be filled with AI “slop” from AI accounts as some fear? Can these characters be made without being really offensive? Is everyone on the Internet a dog AI besides you?
Nigeria cracks down on crypto pig butchering
Nutshell: Almost 800 people were arrested in a raid on a crypto romance scam operation in Lagos.
More: Ever get a random wrong-number text? There’s a solid chance it’s an attempt to start a “pig butchering” scam, where scammers build a relationship to “fatten up” the victim before hitting them for money, usually in the form of a crypto “investment.” People are losing billions of dollars to these scams—it’s estimated that pig butchering has cost victims $75 billion over the last four years; a single Myanmar-based scam compound brought in over $100 million in the first nine months of 2024. And many of these compounds are staffed by human trafficking victims lured with the promise of work and then imprisoned—-an estimated 300,000 people. Nigeria’s arrests are a drop in the bucket, but a start.
Why you should care: Pig butchering is bad all around. Unlike other scams, trying to mess with the scammer (“scambaiting”) might actually hurt someone if they get in trouble with the people keeping them prisoner. It’s unclear how many people arrested in Lagos were there against their will, although 148 were Chinese nationals and 40 Filipino nationals, and I hope they’re treated compassionately. As authorities struggle to keep pig butchering in check, the best way to keep yourself safe is to not respond to random messages. And definitely don’t buy crypto because someone promises you’ll get great returns. (This applies even if you know the person, by the way.)
Social media companies prepare for the next US administration
Nutshell: Tech companies and execs are donating to Trump’s inauguration, and Meta replaced their public policy head with a long-time Republican insider.
More: Meta, Amazon, and Robinhood have all donated $1-2 million to Trump’s inauguration fund, as have the CEOs of Uber, OpenAI, and Apple in their “personal capacities.” Meta has also replaced its global head of policy, former UK deputy prime minister Nick Clegg, with Joel Kaplan, who has worked in the background since 2011 to “repair [Facebook’s] relationship with the right” through various scandals and claims of bias.
Why you should care: It’s clear that we’re entering a new era of how tech relates to the US government. Tech faced a lot of blowback during the Biden administration—from the right and the left—on everything from censorship to child safety to antitrust. It seems like tech companies are trying to stack the deck in advance and start off on good personal terms with Trump. However, considering the rich history of tech worker protests, this may backfire. Time will tell if it’s a winning strategy.
Chinese open-source LLMs are actually really good
Nutshell: A “low-profile” Chinese startup released an open-source large language model (LLM) that’s challenging models from Anthropic, Meta, and OpenAI.
More: DeepSeek has released a new open-source LLM that’s leapfrogged many Chinese and American models both open and closed. In the world of AI, there’s not only tension between US and Chinese models, but between open- and closed-source models. Open-source models1 make (at least some of) their code and/or model weights public, and sometimes training data too, which lets people download and fine-tune them for other purposes. Meta’s open-source LLM, LLaMA, made waves when researchers for China’s army fine-tuned it to create a question-answering bot for “tasks in the military field,” which is technically against Meta’s use policy, but that policy is basically unenforceable when model weights are freely available online.
Why you should care: Open-source technology is seen as a major driver of innovation, but it’s also inherently risky if there’s someone out there who you don’t want to access your code. Open-source LLMs have been lagging behind proprietary ones, but DeepSeek debuted at #7 on the Chatbot Arena, a leaderboard tracking LLM performance. This is above Claude 3.5 and ChatGPT-4o from May of last year (but slightly below its November version) and just below ChatGPT o1. These tests are imperfect measurements of LLM performance (teaching to the test is a common problem), but they’re good indications of where things are heading, and this won’t be the last you’ll hear of this as US techies and politicians start ringing the alarm.
Extra Reckoning
I want to talk about targeted advertisements. It’s no secret that websites companies use data they have about you, including your age, gender, and interests (determined by your online activity), to target ads at you in the hopes that you’ll be more likely to click through and make a purchase. This is especially prominent on social media and is why your advertisements probably look completely different than someone near you.
Recently, I’ve been getting a lot of ads for “compounded semaglutides” for weight loss, which claim to be the “same active ingredient as Ozempic,” the GLP-1 drug originally for diabetes that’s been increasingly adopted for weight loss. (They may or may not actually work, and may or may not send you to the hospital.) By “a lot” I mean probably every minute when scrolling Reddit, and then also sometimes on YouTube and other platforms (Instagram likes to throw in the occasional dietician service). I’ve written before about how I’m profoundly uncomfortable with what our society’s feverish adoption of Ozempic-like drugs represents, and yet it seems like these drugs are here to stay.
ER 15: On Ozempic, Diet Culture, and Robobees
But that doesn’t mean I want them splashed across my timeline. I have a pretty good relationship with my body, but like many 20-something women, sometimes it’s a complicated one. And honestly, seeing messages constantly telling me that I should be losing weight isn’t good for that relationship, especially when they’re saying “It’s easy! It’s cheap!2 We’re going to tell you this through literally every single ad in your Reddit feed!” (I wish I was joking.)
I was going to turn this into a rant about how social media companies should let you have more control over the content of ads you see. After all, what if someone was truly struggling with their relationship with their body and was getting these ads? But it turns out that a lot of the big websites let you limit ads about “sensitive topics,” usually including alcohol, dating, gambling, pregnancy, and weight loss. Here’s how to do it for Google, Instagram, Facebook, and Reddit. Still, I think these companies could go further. Some of Instagram’s new teen safety features include restricting sensitive topic posts, not just advertisements, and platforms should expand this option to all users. TikTok bans weight loss-related ads altogether, although both platforms’ policies have been of mixed efficacy. People deserve to have more control over what sensitive content they see in their timelines. And honestly, it’s in social media companies’ best interests to give it to them. I was getting really frustrated with Reddit continuing to show me those ads even though I downvoted each and every one (apparently totally ineffective), and I’m actively using Twitter less these days, not because of weight loss ads but because all the rest of the content is frustrating. But now that I’ve changed my Reddit ad settings? I’m enjoying my feed much more.
I Reckon…
That last week I talked a lot about what I want to think about this year, but I want to know what you’re curious to hear more about. Drop a comment below with a topic or story you want to see more of in the newsletter this year!
There’s a huge debate over what “open-source” really means, especially as it relates to LLMs, that we don’t have space to get into here, but watch this space for more soon.
$165/month is cheaper than Ozempic, but still… not cheap.
Thumbnail generated by ChatGPT with the prompt “Please generate an abstract brushy impressionist painting in shades of pink representing the concept of scamming”.