Live data from Hacker News

China using big data to detain people before crime is committed

theglobeandmail.com

31–40 of 185 posts

Re: China using big data to detain people before crime is committed

#31
For those of you who find this interesting and also enjoy anime, Psycho-Pass explores this premise.

https://en.wikipedia.org/wiki/Psycho-Pass

TLDR:

> The story takes place in an authoritarian future dystopia, where omnipresent public sensors continuously scan the mental states of every passing citizen. Collected data on both present mentality and aggregated personality data is used to gauge the probability of that citizen committing a crime, the rating referred to as that citizen's Psycho-Pass.

Re: China using big data to detain people before crime is committed

#33
post #26

Earlier quoted context omitted.

For the uninitiated, it's an amazing movie starring Tom Cruise which talks about "precrime" - although no big data involved. Highly recommend that one watch it https://en.wikipedia.org/wiki/Minority_Report_(film)

Based on a short story by Philip K. Dick https://en.wikipedia.org/wiki/The_Minority_Report

For those into Black Mirror or PKD, new show called "Electric Dreams" is based on PKD short stories - I recommend it, a few excellent episodes.

Re: China using big data to detain people before crime is committed

#34
post #22

We’re on our way there too in the US: https://www.theverge.com/2018/2/27/17054740/palantir-predict...

Companies which name themselves after dystopic objects or ideals with a 'nod and wink' always creep the living hell out of me. There's never a good answer when they inevitably ask, "well what did you think was going to happen?"

http://lotr.wikia.com/wiki/Palant%C3%ADri

Re: China using big data to detain people before crime is committed

#36

Many countries are using these pre-crime estimators. What we need to see are the false positive and false negative rates. Anybody got measures of those?

Well it's pretty easy to generate the ROC curves for this one. We're talking discrete events over a defined timespan, right? Just create a control group and monitor the difference.

Re: China using big data to detain people before crime is committed

#37

The NSA is already doing the same thing https://arstechnica.com/information-technology/2016/02/the-n... Machine learning is still a black box, and politicians don't seem to understand this.

There is a massive difference: the NSA/USG isn't using that against it's own citizens.

Re: China using big data to detain people before crime is committed

#38

This will be rolled out further and further, first the rest of China, then other parts of Asia, eventually Europe and the Americas. Strongly resist the rollout of this technology, along with the social credit rating systems. This is like the last chapter of humanity before we stagnate for hundreds of years.

I'm confident in Europe's ability to resist this trend, their privacy legislation is looking positive.

Re: China using big data to detain people before crime is committed

#39

Earlier quoted context omitted.

They are not "criminals" if they never committed a crime, they are people being detained based on probabilities.

Has it had an effect on crime rates? If so, this is interesting. It means crime is predictable based on the inputs the system is measuring. It also means the Chinese AI effort is more fruitful than I expected it to be. If this is reducing crime, then we have a discussion of trade-offs. That is interesting, and potentially nuanced. If it's the latter, however, this is not interesting and there is no reasonable discuss…

To rephrase: You could define whether it is being used corruptly by measuring whether it reduces crime.

This is a loaded research question, of course, because it's studying something that we find abhorrent: Prosecuting people for crimes they haven't actually committed.

Scientifically it's super interesting. Socially and ethically it's a minefield.

Re: China using big data to detain people before crime is committed

#40

Earlier quoted context omitted.

They are not "criminals" if they never committed a crime, they are people being detained based on probabilities.

Has it had an effect on crime rates? If so, this is interesting. It means crime is predictable based on the inputs the system is measuring. It also means the Chinese AI effort is more fruitful than I expected it to be. If this is reducing crime, then we have a discussion of trade-offs. That is interesting, and potentially nuanced. If it's the latter, however, this is not interesting and there is no reasonable discuss…

>if this is reducing crimes

Every apprehension under this program is a crime by any definition other than the arbitrary declaration of an illegitimate regime.

It would be absolutely astonishing if the people apprehended were likely to illegally kidnap and hold at least 1 citizen in their basement for at least the length of time that these people are detained.

Thus it can pretty much only increase crime.

Post reply on HN