Scientific research has found a new national sport: slapping human qualities onto algorithms. But one team of researchers decided to see just how far that logic stretches, and the result is brilliantly absurd.
In 30 seconds
01Researchers debunk the myth that LLMs are intelligent like humans are.
02
→
💡
What this means for you
Basically: the next time you read a headline screaming "AI *understood* this" or "AI *learned* that," maybe raise an eyebrow. It *might* be true, or it might just be a lazy way to describe what the software technically did.
Thought slapping 'AI' next to a company name guaranteed its stock would soar? Well, the market had a bitter surprise this year.
·1 min·2·Beginner
If LLMs were truly intelligent by the same logic, Age of Empires II would be too.
03The real issue: science uses terms like 'intelligence' imprecisely when describing AI.
Let's be honest: lately, every time a study drops claiming LLMs (that's Large Language Models, like ChatGPT) act “as if” they're intelligent, genuinely smart, or even gasp conscious, the headlines go absolutely wild. Think: “AI has developed empathy,” “AI understands human language”—you know, the kind of sensational claims that send Twitter into a frenzy and rake in clicks.
This research kicks off with an observation so simple it's brilliant: if we're attributing human characteristics to LLMs just because they solve problems, recognize patterns, or act in “intelligent” ways, then by the same logic, we'd have to admit that Age of Empires II—that legendary strategy video game from 1999—boasts the exact same qualities. The game crunches complex numbers, makes decisions, and manages resources. So, technically, by the very same yardstick we use for LLMs, it too would possess human attributes.
📬 Enjoying this article?
Get the best AI news every week, straight to your inbox.
That irony, by the way, isn't accidental. The researchers are highlighting a very real problem in current science: we often throw around terms like “intelligence,” “understanding,” and “cognition” far too loosely when describing what AI models do. If a behavior is truly an indicator of human intelligence, then it should be exclusive to humans (and maybe a few particularly clever animals). If it's not, then we're simply using the wrong words to describe what's actually happening inside those algorithms.
Now, this isn't to say LLMs aren't impressive—they absolutely are. But there's a crucial difference between “producing output that looks intelligent” and “being actually intelligent.” Age of Empires II does the same thing, just in a simpler way: it simulates complex behaviors without having a single conscious thought in its “head.”
Ultimately, the research is a gentle but utterly ruthless dismantling of how scientific language far too often slips into anthropomorphism when talking about AI. It's not about claiming LLMs are dumb; it's about suggesting we should probably be a bit more precise about what we actually know versus what we're just guessing.
While the tech world was buzzing about OpenAI, Anthropic made its move. They just dropped Opus 5, a model they claim is almost as good as their legendary Fable 5.
NUOVO
