College students used Meta's smart glasses to dox people in real time

> Google Glass originally failed due in part to public backlash at being recorded without consent in public spaces. However, it’s also true that in the decade since, people have become more accustomed to being filmed due to the rise of smartphones, vloggers, and TikTok.

This is a massive "citation needed." People around me are still almost always not filming, and I absolutely have not become accustomed to being filmed. I expect that's generally true for most people?

Really glad there are researchers out there coming up with bad actor scenarios, and I hope corporations, privacy advocates, and the government take these into consideration.

These are solvable problems. Not always 100% but the harms can be mitigated. But the truth is that some people don't want these problems solved.

Another example in the wild was Apple Air Tags.

Are there new license plate readers that also do object and face detection? Are the AI corporations starting to train their algorithms on all of reality instead of just what is online? Tesla at least has long been collecting video data that could be used for this purpose, besides being used for training self driving car algorithms. Vehicles can be tracked using subsonic vibrations, where every vehicle in the world moves could be tracked with an array of geophones and nobody would ever know it is happening. Perhaps SQUID magnetometers could do the same with many fewer sensors. SQUID magnetometers are extremely sensitive, so much so that the output looks like noise without shielding, but maybe AI could make sense of all that signal that looks like noise, but which is actually derived from every electromagnetic event on Earth? I talked to a scientist about this and he thought the confluence of so many signals would be impossible to distinguish at the time, but maybe this is a problem that AI would excel at by being able to see minute correlations that people or heuristics can not? Not sure I'd want to live in such a world though. Instead of an LLM it might require a more time-domain oriented algorithm.

[dupe] https://news.ycombinator.com/item?id=41726385

https://news.ycombinator.com/item?id=41724957

among other submissions.

Just focus on the students' source instead of repeating these news posts over and over and over

Source Doc: https://news.ycombinator.com/item?id=41724310

I'm curious about how they do the reverse image search. Do they just use Google?

Stuff like this makes me glad that my family and myself practice digital hygiene such that there is not a single photo of myself online that could lead back to me.

For everyone else's sake I hope employing automated (facial) recognition on private citizens becomes so highly illegal, no sane person would ever dare to do crap like this.

All I've ever wanted from AR glasses is the ability to look up people's names so I don't have to remember them.

While I like what they stitched together, it has nothing to do with meta ... at the end of the day, any camera can do the same