I don't expect much of Mitchell, but this is impressively bad. Opening by attacking the messenger, because of course it's completely unreasonable for writers to not be professionals in a space, she manages to get worse from there.
> 'Friedman notes that, for AI systems that included training text in different languages, “the AI could translate between the languages it had been trained on—without anyone ever programming it to do so.”'
She calls this "magical thinking"...and then goes on to acknowledge that it's entirely true and accurate, while trying to dismiss it on the grounds that there exists text paired to a translation available to scrape.

We have evidently different opinions of Mitchell? How so, what's her problem? Also I didn't interpret what she wrote that way you report, she's an advocate for science-based results and hype-removal when it comes to AI and here she's pushing back on bombastic claims.