Social media companies are working on adding AI to their products, and some of them are seeing more mixed success than others. Why?
AI is the hot product of the day. Even the worst-case scenarios are still good for their owner companies. Look at Grok. A branded LLM (LLMs are frequently called AI) brought a ton of media buzz to Twitter, now X, because Grok lacks rails or guidelines to interact politely with users or source accurate answers for questions. Briefly it was incredibly racist. However, the effect of this up-and-down behavior on Grok’s part is that X itself got a bit of a break from the headlines because Grok was soaking up all of the negative press. Additionally, while Grok topping the web results tied to X wasn’t good, it also wasn’t horrifically bad like it should have been, because X’s new branding as the edgy free-speech website was only bolstered by Grok’s bizarre posting.
A company sees this and thinks to itself, “we could do that. Probably not with the racism, but adding an in-house LLM would boost us in search and benefit us regardless of users using them or not.” They aren’t necessarily wrong. Either users get these AI features, dislike them, and then choose to complain loudly elsewhere about the features (alerting other people that Facebook and Instagram, two of the most saturated social medias on the market, now have AI) or they like them and maybe start using them, and when these companies inevitably unveil a subscription service later, they’ll be more likely to toss a couple of dollars to Facebook every month to keep using them.
But it’s not as simple as that. Some of them genuinely want true adoption, because Grok’s a success story in that regard too. Many of these companies have spent millions on research and investment into LLM companies, expecting customers to adopt them with the same voracity that Twitter/X users adopted Grok, and many of them are seeing numbers that disappoint them. This is why the UIs are changing. Because companies are worried their userbases are not using the AIs they spent tons of money on.
One conspiracy theory is that the companies – seeing less usage than they want – are resorting to trickery to get people to open the AI features they’ve been avoiding. I know when I upload a photo, I definitely already did all of the tweaking to color and background in a better editing app than Instagram. What, why does Instagram need picture-tweaking AI? Instagram’s trying to like, edit my photos? By itself? Better than regular Photopea, a fully-fledged and functional picture editing app? Better than the Photos app that comes on the phone? Better than Adobe?? What is this? Obviously I never open the AI editing on purpose. However, when they move buttons around, my muscle memory will occasionally lead to me hitting a button I don’t intend to because it’s where something else used to be. Discord removed the “send” button on it’s desktop app so it would have more room for another button leading to a feature you need to spend money to use – the way social medias are setting up their LLMs is going the same way.
From the consumer’s point of view, the AI (generative content, not “true” AI, remember) is either a cool feature that saves them some time, or something they don’t want and don’t need. X is an exception because it’s full of people who stayed on X for years after Elon Musk took over, and the crowd there is uniquely receptive to LLMs. Regular, socially-neutral apps are chasing a result they can never obtain. Apps providing picture editing – cool. Okay, so they can remove stuff from the background and add stuff that will eventually look good once the tech catches up. Some people probably want that built into their social media app. They want an LLM, they don’t want Photopea, they see room to improve and they think Instagram can do that for them. The final problem is that on the same timeline, a bunch of actual photo editing apps launched the same product. Including the ones these people were likely using before. This is a move that gets people who’d been uploading raw photos to start using the AI feature, not the people who’ve been editing already, whether because a different app is providing AI or because they themselves already edited the photo and don’t want the AI retouching it and messing stuff up.
The AI a social media app comes up with to edit may also be worse than the editing app’s version. In the same way a combo dryer-washer is worse than a regular dryer and a regular washer, keeping the apps separate gives the person who took the photo much, much more control over the tweaks they make. The market is already satisfied one way or another. In retrospect, there’s no way adding AI photo editing to anything was ever going to make people use it productively.
Using it wrong is also an option that these companies are introducing, and if they’re not X, they’re realizing they let a genie out of a bottle. Grok’s effectively being used to harass people as often as it’s being used for anything else: you ask Grok to put someone in a maid outfit or make them bald and Chinese, it just does it without question. Then there’s a photo of “you” effectively in digital yellowface right under your own post. It’s absurd! And probably a sign these programs don’t need to be fully integrated with the platform such that they can reply to any dumb request anyone makes!
