Ludicrously bad fails. Surely AI language instruction is better than this with AI - were these developers trying to save money on compute and had poor models?
Here are some youtube videos, or articles that caught my eye - from the New York Times, Consumer Reports, Popular Science etc.
Monday, November 2, 2026
Tunnel warfare in Ukraine
https://youtu.be/iGyuC3xHPiA?is=MfHqQ0IKd1N9V3Vh
There are many aspects of this war that are dishearteningly tough on soldiers, but this is a particularly difficult one.
Solar panels over irrigation canals
https://youtu.be/89uDjx32BQY?is=M-rMxokXz-u-ptZ5
Putting solar panels over irrigation canals is a win-win: covering the canals prevents huge evaporative losses, and the underlying water keeps solar panels cooler so they're more efficient.
But, a simple sensible idea turns out to be difficult to carry out in real life as multiple governmental agencies have to sign off on it.
Using meta glasses to recognize strangers
The creators emphasized they did not release the code or software to the public. Their primary goal was to raise awareness about accessibility of public data using consumer hardware and off-the-shelf AI tools, acting as a "warning shot" regarding privacy risks.
The user wore the Meta glasses and streamed live video to a private Instagram account.
Facial Recognition: A computer program monitored the live stream, detected faces, and sent cropped photos of target individuals to reverse image search tools (such as PimEyes or FaceCheck.id). Once a matching image or name was found on public web pages, news articles, or social media, the software queried public data broker databases (like FastPeopleSearch). An LLM parsed the returned search results to extract key details—including the person's full name, home address, phone number, work history, and names of relatives—and pushed this "dossier" directly to the wearer's phone via an app within seconds.
Subscribe to:
Posts (Atom)