science

Spanish neuroscientists confirm ai fails the spark test

Stephen King calls it ‘soulless’. Spielberg swears it can’t dream. Now a Barcelona lab has strapped electrodes to the claim and watched the brain light up in a way no transformer can fake.

The verdict: when humans invent, they don’t shuffle data—they rewrite memory, emotion, yesterday’s heartbreak and tomorrow’s hunger into something statistically improbable. Algorithms, meanwhile, stay trapped inside the bell curve.

Why the study lands now

ChatGPT drafts half the posts flooding your feed. Midjourney illustrates the ads chasing them. Creative industries are quietly auditing headcounts, so the University of Barcelona decided to run a creativity stress-test before boardrooms decide who stays.

Researchers pitted 54 volunteers against three leading language models on the classic Alternative Uses Task—turn a paperclip into anything but a paperclip. Humans produced ‘lock-pick for a dollhouse’, ‘microphone for ants’, ‘tooth for a zipper monster’. The machines recycled the same five synonyms for ‘hook’.

Neuroimaging showed the moment of genuine insight: hippocampus, amygdala and prefrontal cortex fired in a chaotic salsa that no GPU replicates. The AI’s activation maps looked like tidy spreadsheets.

Data is not experience

Data is not experience

Feed a model 50 million scripts and it will average them into mush. Feed a toddler one sunset and you might get Close Encounters. The difference, the researchers argue, is autobiographical anchoring: creativity is the revenge of the lived.

That asymmetry scales. Studios already use AI to storyboard; publishers use it for slush-pile triage. The danger isn’t replacement—it’s dilution. When the pipeline fills with competent-but-generic output, the statistical middle becomes the new ceiling and audiences forget what edge feels like.

King and Spielberg sensed this without needing an fMRI. The neuron scans just handed them a peer-reviewed mic drop.

The profit motive won’t wait

The profit motive won’t wait

Start-ups pitching ‘AI-generated writers’ rooms’ have raised $700 million this year. Expect marketing decks to ignore the paper and trumpet ‘human-level fluency’. Investors rarely read footnotes about hippocampal dynamics.

Meanwhile, Amazon’s Kindle store added 5,000 AI titles last week. Reviews complain they read like ghostwritten Wikipedia. The long tail is turning beige in real time.

Barcelona’s team closes with a warning: every time we outsource a first draft, we train the next generation of creators on synthetic mediocrity. The loop tightens. Metrics rise. Culture stalls.

Hollywood’s strike wasn’t about salary scales—it was a last-ditch attempt to keep the bell curve human. The guilds lost the wording; the neurons just handed them the receipts.

Keep the receipts.