Re:To do politics or not do politics? Tech startups are divided
by singron on Monday November 23, 2020 @08:55AM
(#25186765)
From my reading of available publications by the airforce on maven, the calculus was actually that they had a fixed threshold for risk, and being able to analyze more data would allow them to act in more cases with that same risk level. To probably oversimplify, if they were ok with 25% innocent causalities, increasing efficiency 100x would just let them kill 100x more often at the same ratio. Disclosure: ex-google
Project Maven continues, now run by Palantir. Google's refusal to continue the project did nothing but slow down the project by a few months.
https://www.bloomberg.com/news/articles/2019-12-14/palantir-...
In fact, considering the goals of Project Maven (roughly making sure you're blowing up what and who you intend and avoiding collateral damage), if you consider Google a leader in AI, you could make the perverse argument that you're actually killing people by reducing the effectiveness and delaying the timeline of the program. I'm not saying I've researched this enough to say that this is a correct or even plausible argument, merely that it is possible.