Ethan Lambeth

I guess these are all object related, what about segment detection/recognition of people and their objects?

Blur people's faces on the street, some people might opt into default anon, others not so much. But there will be open source tools on these glasses. Will people walk around with glasses that blast out infrared light to disrupt cameras recording their face? Will this even be legal! I am thinking of these demos people do with the Meta glasses where they run OS facial recognition and then go up and talk to the person. I know there are clothes/scarves you can wear to hide from this stuff but that feels too high friction. Also a lot of privacy oriented stuff (in Europe!) is getting banned.

The effects

Ok list effects of this:

Anyone is stalk-able.
Some cases of people being stalked/doxxed (I guess this also goes hand in hand with increases in model capabilities that could de-anon pretty much anyone).
Public fear/distaste for the technology, probably driven by a particularly nasty case involving someone being unjustly victimized by an offender with the tech.
Call to outlaw tech, maybe it becomes a 3D printing weapons situation. (Do there exist pro open source people who are anti this? Are there privacy pilled open source people?)
Some politicians call for regulation.
Sentiment around it bubbles, some privacy oriented startups will raise a bunch of money (assuming things don't go FOOM before this happens).
Well what if everyone is walking around with the same ASI on their face, surely it won't harm person because they are also a user? Also someone to care about? I guess that's an alignment problem and that is unlikely to be solved before Zuck ships something decent. Also these tools can run decently enough without a model needing to do things. The cat is fully out of the bag.
Do we just blanket ban all (consumer, not Palantir rip) human facial recognition software? Getting caught with it is punishable. But people will still find and run them on jailbroken glasses.

So what are the actual risks of this tech? No way it is worse than model biorisk.

Privacy is fully gone in public (arguably already the case, social media is incredibly good at finding people).
Most people act as though they are already within the panopticon.
What maybe if there is a data breach of 100k glasses, and their data stores, then that data could be combed through to see all of your IRL patterns across an aggregate of people's lens POVs. But why are IRL patterns actually that sensitive? You can argue digital patterns are richer, and are a superset of IRL patterns (ie your location data is already out there, which stores your repeated purchases are at, etc), so maybe the threat isn't the new surface for info collection, there is already a LOT out there anyway. The threat is model capabilities allowing people to comb through this. The bottleneck is incentive, these scams/hacks can be done at scale now. But why do people want them? Where does the money/damage come from?

Money either from:

Consumer via phishing, scamming, hacking etc (smaller scale, can be fast access to capital).
Institutions are bigger gains but slower.

Damage/Destruction:

Calculate the largest surface area for harm?
Spearphish more effectively, hit targeting.
I guess if you want to find a target you will no longer need their location from their phone. It is no longer about their opsec but the opsec of the people around them. If you are spotted on someone's glasses who is calculated to be close to you, they have a sleeper bug constantly scanning/sending faces until it matches the target, then immediately send location. I wonder if this can happen already with cameras and microphones (or even just location/spending data). Camera will be hard because it has to be pointed / visible but you can just turn a microphone on (but you have to be talking and get a voice match). Prob been done by a gov.