The media still can’t get AI right, and it’s getting worse

Look, I’ve been watching this circus for years, and Oscar Schwartz’s 2018 Guardian piece nails a problem that’s only gotten worse. The media coverage of AI is a dumpster fire, and social media has given self-proclaimed ‘AI influencers’ a megaphone to parrot Elon Musk’s worst takes for clicks. The result? A public that’s either terrified of Skynet or convinced AGI is five years away. Neither is true.

The classic example Schwartz uses is the 2017 Facebook AI ‘panic.’ Five researchers at Facebook AI Research published a paper showing bots could simulate negotiation dialogues. Mostly they talked fine, but occasionally they’d spit out nonsense like ‘Balls have zero to me to me to me to me to me to me to me to.’ The researchers noted this was because they forgot to constrain the bots to proper English grammar—not exactly earth-shattering.

Then Fast Company ran with ‘AI Is Inventing Language Humans Can’t Understand. Should We Stop It?’ They framed it like the bots had gone rogue, and the researchers ‘pulled the plug’ to contain the monster. The story went viral. The Sun compared it to The Terminator. Actual experts like Zachary Lipton at CMU watched this transformation from ‘interesting-ish research’ to ‘sensationalized crap’ with predictable frustration.

This isn’t new. Schwartz traces the pattern back to 1946, when the ENIAC was unveiled. Journalists called it an ‘electronic brain,’ a ‘mathematical Frankenstein,’ a ‘weather wizard.’ Physicist DR Hartree tried to set the record straight in Nature, explaining what it actually did. The London Times promptly ran ‘An Electronic Brain: Solving Abstruse Problems; Valves with a Memory.’ Hartree’s letter to the editor calling the term ‘misleading’ was ignored. The ‘brain machine’ label stuck.

Same story in 1958 with Frank Rosenblatt’s perceptron. It was a rudimentary pattern recognizer—could only handle limited inputs. The New York Times declared it an ‘electronic brain’ that could ‘teach itself’ and would soon ‘walk, talk, see, write, reproduce itself and be conscious of its own existence.’ That level of hype helped get funding, sure, but by the late 60s, pioneers like Marvin Minsky realized they’d badly underestimated the problem. Minsky had predicted machines would surpass human intelligence in his lifetime. In 1969, he co-authored a book proving Rosenblatt’s approach had fundamental limits. The first AI winter followed.

We’re repeating the cycle. Every time a language model generates a slightly weird sentence, the headlines scream about machines becoming self-aware. Every time a researcher mentions potential risks, it gets twisted into ‘AI will kill us all.’ The discourse is unhinged, as Schwartz puts it, and it’s not just annoying—it’s dangerous. Unrealistic expectations lead to boom-and-bust funding cycles, policy based on fear rather than facts, and a public that can’t distinguish between a chatbot hallucination and actual artificial general intelligence.

I’m not saying AI doesn’t have risks. It does—bias, job displacement, misuse. But we need sober, accurate reporting, not clickbait dressed as journalism. The media needs to stop treating every incremental advance as the dawn of the machines and start treating AI like what it is: a powerful tool with real limitations. Otherwise, we’ll keep getting burned by the hype cycle, and the real conversations about responsible development will get drowned out by the noise.