27 ms·
Google AI helped IDF drones with targeting in 2024 breaching its own policies
- bhouston 9mo agoArchive link: https://archive.is/WJ36r https://archive.is/WJ36r One key quote: "The request from the IDF email address asked for help making Google’s Gemini more reliable at identifying objects such as drones, armored vehicles and soldiers in aerial video footage, according to the internal documents included with the complaint. Staff in Google’s cloud unit responded by making suggestions and doing internal tests, the documents said." Another showing double standards: “Many of my projects at Google have gone through their internal AI ethics review process,” the former employee who filed the complaint said in a statement to The Post, provided on the condition of anonymity for fear of retaliation from the company. “That process is robust and as employees we are regularly reminded of how important the company’s AI Principles are. But when it came to Israel and Gaza, the opposite was true. … I filed with the SEC because I felt the company needed to be held accountable for this double standard.”
- PlatoIsADisease 9mo agoCant make noods, can cause painful ending of life. Optics are everything. Amoral companies don't actually care. Water is wet. Maybe I need to understand how the legal slave trade died.
- nashashmi 9mo ago> I filed with the SEC because I felt the company needed to be held accountable for this double standard How did Israel get away with so much leeway from tech companies? It can't just be money. They are too small a market. What else is there?
- dttze 9mo ago[dead]
- woleium 9mo agothey have world leading cyber offensive
- mhb 9mo ago"Google in February last year updated its AI policies to remove its pledges not to apply the technology to weapons or surveillance, saying it needed to evolve to help democratically elected governments keep up in the global battle for AI dominance."
- xnx 9mo agoTitle is now: "Google helped Israeli military contractor with AI, whistleblower alleges"
- sshine 9mo ago“Don’t be evil” — as with anyone who works with AI-powered weapons targeting technologies, there are two tricks to stay on the moral side: 1. Only sell to good countries 2. Only target bad people (Company pitch when I accidentally ended up interviewing at a weapons manufacturer for their new edge GPU research.)