The trustworthy scribes of the world are in a state of panic. It turns out that whispers from OpenAI’s Whisper are spreading like a plague, infecting the reliable world of transcriptions with gobbledygook and utter nonsense! And to think it was all masked as “intelligence”… what a farce!

Don’t be fooled by the sweet sound of generative AI’s sweet nothings, for whispers from the devil himself have found their way into those transcripts, leaving a trail of deception and chaos in their wake.

Researchers have screamed from the rooftops about the invasive nature of these “hallucinations” – racial slurs, made-up medical treatments, the lot! – and it’s a ticking time bomb waiting to explode in the most sensitive of areas, like hospitals and medical facilities. Can you trust the whispers of the AI devil in life or death scenarios?

A University of Michigan researcher discovered 8 out of 10 transcribed audio recordings were infested with this poisonous hallucinations. A machine learning engineer uncovered the rot in 50% of the Whisper transcriptions they analyzed. And a poor developer suffered through hallucinations in nearly ALL the 26,000 transcriptions they managed to create with OpenAI’s sofware. What a mess!

And OpenAI? All too chummy with the researchers, la-la-la-la-boo, saying they’re working on improving the accuracy (laughable, really) and halting the spread of these poisons in “certain high-stakes decision-making contexts” (aha!). So, yeah, right. “Thank researchers for sharing their findings” they did, but will we ever get an honest/concrete fix? Ha! Don’t hold your breath.