Meta recently claimed that its AI systems have reached “human-like levels.” A tech bro on X eagerly reposted the announcement, praising Meta’s brilliance and declaring the AI to be “bright like a U.S. president.”
Well, he’s not wrong.
After some light testing, I can confirm that Meta’s AI performs exactly like a U.S. president – Joe Biden, to be specific. And not Joe Biden in 2012. Joe Biden in 2025. You know, the version that wanders off-stage and forgets which country he’s in.
The Hashtag Experiment
To validate Meta’s claim, I ran a simple comparison test:
-
Old Instagram (2020): Click a hashtag. See all matching posts. In chronological order. 100% accuracy. Fast, functional, and straightforward.
-
Current Instagram (2025): Click a hashtag. Get served a chaotic pile of unrelated content, mixed with a few actually relevant posts, completely out of order and often weeks old.
It’s like telling the AI, “Show me posts with #Barcelona2024,” and it replies:
“Here’s a selfie from 2021 with no caption, a salad recipe from a crypto influencer, and one meme about Ohio.”
That’s not just bad indexing—it’s cognitive dysfunction.
Redefining “Human-Like”
Meta’s claim hinges on the slipperiest of marketing tactics: redefining “human-like” to mean barely coherent, easily confused, and chronically forgetful. If that’s the benchmark, then yes—congratulations, Zuck. You’ve built the Joe Biden of artificial intelligence.
Meanwhile, OpenAI’s ChatGPT, Elon’s Grok, Google’s Gemini, and even the underdog DeepSeek are demonstrating advanced reasoning, memory retention, multi-step logic, code generation, and actual language comprehension. They’re flying starships while Meta’s bot is drooling on its shirt and trying to find the door.
What This Means for Users
Meta’s AI isn’t just bad—it’s a liability for anyone who relies on discovery. Hashtags used to be a powerful search tool. Now, they’re just another algorithmic wasteland, where actual relevance is lost in a swamp of recycled engagement bait.
It’s no surprise. Meta has long prioritized algorithmic spoon-feeding over user agency. Why let people search and filter when you can force-feed them what the system thinks they’ll engage with longer—even if it’s irrelevant, outdated, or flat-out wrong?
The Real Intelligence Gap
The tragedy here is that Meta is sitting on enormous resources, massive datasets, and billions of users—and they still can’t build a hashtag engine that works as well as it did in 2020.
This isn’t human-like intelligence. It’s bureaucratic entropy dressed up as innovation.
If Meta’s AI is the future, I’d rather talk to Clippy.
