I think it shows "open source intelligence" can gather too much information which is risky to the US strategic interests by harvesting the data implicit in the models.
The more inferences improve, the more likely it is somthing strategically valuable is at risk. Somebody maybe in the Fort Meade district did their homework.
It's nothing about AI per se, or the mythical hacking and code risks. It's the wider information leakage risk although possibly it showed too many backdoor opportunities in federal and state and supply chain online systems.
I think it shows "open source intelligence" can gather too much information which is risky to the US strategic interests by harvesting the data implicit in the models.
The more inferences improve, the more likely it is somthing strategically valuable is at risk. Somebody maybe in the Fort Meade district did their homework.
It's nothing about AI per se, or the mythical hacking and code risks. It's the wider information leakage risk although possibly it showed too many backdoor opportunities in federal and state and supply chain online systems.
Discussion: https://news.ycombinator.com/item?id=48511072
Ohh, time to read!