Hacker News Reader

Comment by LeeHunter | Hacker News Reader

Some of this is disturbing and some is not. I don't know why the article makes a big deal out of predictive policing and it's roots in counter-insurgency. Putting police officers where crime is likely to happen is just good management. More power to them. I could care less whether the software was developed for Fallujah. Why would you want to put police resources where they're not needed?

View on HN

Haha, this is funny. I'm one of the co-founders of PredPol, the startup that makes the predictive policing software LA uses. I think it's a stretch to call anything I wrote in my dorm military grade, but I suppose I'll take that as a compliment. For the record, the DoD called us once asking if some of the models we use to predict crime could be used to predict IED locations, but nothing major ever came of it as far as I know. Also, I live in Santa Clara not Fallujah.

P.S. PredPol predicts crime using only the time, place, and type of past crimes using some simple statistical models. All that is public information for municipalities in the US.

Can you back this up with some links please?

And anyway, even if the LAPD may not be using anything more sophisticated than PredPol with growing their CCTV array, they will in the future. Think of it like this: right now the data is stored. In 5-10 years it will be cross correlated and profiles will be built on everybody. Thanks to the efficiency, the definition of crime will be expanded. Look at Singapore's law against gum wrappers. Look at local laws against homeless blankets. More to the point - the sites you surf can be cross correlated with the places you visit to indicate a reasonable belief that you may harm your kids and they will be taken away after you've been aubpoenaed in court.

20-30 years from now: court cases initiated by computer and suggested to the DA, based on cross correlated databases. Theories built and presented against you by the eloquent lawyer, with the jury thinking, "yeah this cant be all a coincidence." And a vastly expanded private prison complex.

All in all maybe not a bad thing, but people arent used to living like robots.

Not only is omarforgotpwd right, but the article is completely wrong with regard to predictive policing. It was invented by the LAPD w/ the help UCLA researchers in 2008. It is not the result of military research, and was not used by the military until recently. LAPD has repeatedly been credited with it since 2008.. and I was not able to find a single link that backed up the article's claim that it came from the military.

And the program is far from futuristic. Basically, if a certain type of crime has happened in an area recently, the LAPD patrols the area more often. Big deal.

Here's a summary from the FBI:

http://www.fbi.gov/stats-services/publications/law-enforceme...

And here's a longer article about it (including the history) from RAND:

https://www.ncjrs.gov/pdffiles1/nij/grants/243830.pdf

I don't really have much to say about the rest of the article, but just thought it was funny they mentioned my startup and quoted one my cofounders.

I will say though, based on my experience working with the LAPD, that there is significant red tape around everything. For example I know cops have wanted traffic camera footage for a while but are not allowed access.

There are all kinds of horrific things you can do with technology. We just have to make sure nobody builds them. I'll do my part.

Using Baysian type on data is very damaging on society, for those with strong maths:

We'll move from causality( we noticed you are stocking up on baking powder, lets chat) to inference (according to our data there is 85% chance you are not compliant - w/o a cause).

One outcome: The populace will be demotivated to do anything, just to be safe.

It forces people to worry about fitting the profile of someone who would be more likely than others to do something wrong. Maybe because you live in a certain neighborhood, or recently started biking to work, or have parents who are migrants from Eastern Europe, that thing moves the needle from a 4% chance to a 5% mean chance of committing a list of crimes and triggers police interest.

Not doing anything wrong ceases to be good enough.

Maybe they could index the civil liberties in neighborhoods or states by an index like that. Move them from stop-and-frisk through no-knock and so on to martial law.

Reply on news.ycombinator.com