Microsoft research had some good publications on generating infinite non-repeating textures by (IIRC) markov-filling an aperiodic tile set, including creating video textures. I tried to find the examples a few months ago but the old URLs I'd bookmarked were no longer working.
> no way to prevent the intentional weaponisation of chat/ai
AI psychosis is happening without any intentional weaponization. Sure you can't help intentional bad outcomes but accidental ones are much worse because they don't require a malicious party.
Consider a conman-- a harm to society for sure, but he can only con so many people and he'll only do so in ways that get him money, power, glory, etc. Bad sycophantic AI can con and unlimited number of people and isn't limited to destorying their lives in ways that benefit the AI's owners.
> Also, the privacy and the finding are in direct opposition to each-other, which isn’t always a comfortable system dynamic. ... or some sort of group key sharing
Perhaps you might consider a pinsketch in the manner proposed for cryptographic biometric security. https://arxiv.org/abs/cs/0602007
With it two peers could compare their BSSID environments and learn ~nothing about each other unless they were nearly matching.
I can see how one could use it for location based key agreement for mutual authentication-- not as obvious to me how to apply it to privacy preserving location.
The latter would probably just best be accomplished by downloading the whole database, or (less optimally) using PIR to probe for the locations of single BSSIDs.
Ideally nodes have lots of connections so that attackers can't so easy block transaction and block propagation and censor information. But lots of connections means lots of network bandwidth wasted relaying redundant information nodes already know about.
From day one bitcoin relayed transactions by offering just their hashes and only requesting what wasn't known. But even sending hashes ends up being a lot of data in total, and the bandwidth scales with peers*transactions. Using set reconciliation changes that to more like peers+transactions.
When using setrecon for authentication its important that the scheme is as close to information theoretically optimal as possible-- which makes approaches like pinsketch important.
For the transaction case faster but less communications efficient methods might be better except that latency is also a consideration for Bitcoin and minimizing latency means running reconciliation often. This stresses the inefficiencies of alternative tools as well as covers up for the quadratic decode cost of pinsketch.
I'm not aware of anyone using our minisketch library for authentication-ish uses but I'd be interested in seeing it.
Irrelevant nonsense can also poison the context. That's part of the magic formula behind AI psychosis victims... if you have some line noise mumbojumbo all the output afterward is more prone to be disordered.
I'd be wary of using any canary material that wouldn't be at home in the sort of work you're doing.
This article is about movies skimping out on the communication skills of their characters as a narrative device-- it's a lot like the fact that many pre-cellphone movies would have been over in 60 seconds if only the characters had a cellphone. We live in a world with cellphones now, so they're usually expected in movies. We still don't live in a world where people know how to communicate.
Movie logic-- or more generally "narrative bias"-- taints people's thinking in ways far beyond failures to communicate. I've encountered a lot of people who fall for scams or believe conspiracy theories for reasons that ultimately amount to "if life were a story, this is what would happen".
Yes! It's a crucial distinction. Rationalism is about being rational / logical -- moving closer to neutrality and "truth". Whereas to rationalize something is often about masking selfish motives, making excuses, or (self-)deception -- moving away from "truth".
