I just mean that the data that the models and systems are built upon is just as integral as the algorithms themselves, and that in order for the system to be completely transparent the data must be public as well. But such data has a number of privacy concerns that need to be addressed.

One interesting thing is that is currently being worked on is an interesting type of system called a GAN, a generative adversarial network. While they have been used mostly for image generation and modeling the latent space from probability distributions for visual data, research is now being done to applying GANs to generate data that can be used as a training set representing the same distributions as real data, but it is all generated thus minimizing or eliminating some sets of privacy concerns.

A sharp guy (Ian Goodfellow) came up the the idea for GAN's while boozing it up one night and the way they work is pretty awesome. Basically it is two networks competing against each other: one that is trying to discriminate between fake data and real data, and another that is trying to generate fake data that looks like real data and fool the discriminator. Similar to a bank trying to tell fake bills from real one vs an expert forger who is trying to make fake bills, they both get better and better at their job over time to the point that the fake bills are indistinguishable from the real. One of the tricks is there is a 'mole' at the bank who keeps telling the forger how to improve his fakes. The whole concept is pretty great and has turned out to very powerful, and there remains a lot of research and application of such methods.

Well, only if labor actually organizes and forms a syndicate, which does happen in some cases.

In most cases, labor will negotiate a deal with ownership where both parties can be happy. But when the laws are stacked against labor, ownership just takes more and more of the pie over time. This effects middle income people heavily, it's not just a factory worker's problem:

LOL awval

Bernie Sanders says 'every other major country' has family paid leave except United States
Politifact: Mostly True
http://www.politifact.com/truth-o-me...-country-has-/

When this was a radical progressive concept.

taxation is theft, yo

Property is theft

No one owns a theft

https://medium.com/@MattBruenig/the-...1-d3a49fad0580

I also think it's hypocritical of the Court to use the software without subjecting it to the standard rules in their state pertaining to the admissibility of expert testimony.

Agreed. That chart doesn't mean happiness.

Still, middle class is being sucked dry in our country. They pay a huge % in taxes and got very little back. The rich might spend more in taxes and even pay a higher % but they are still fine. Poor people get much more than they put into system. This used to be ok with middle-class job stability like unions when you started your career, you just assumed it would be 30 years. Now, you are lucky to make it 3 years at any single job.

Why not? Having capital isn't a skill.

1. For a simple decision framework, you don't need the entire dataset to see whether it is biased in ways that are socially problematic.

2. We are talking about criminal recidivism. The data sets *are* public. What privacy considerations?

The rest is you just getting excited about something irrelevant.

1. Wrong.

2. Wrong. Also I was talking about use of AI systems for judicial decisions going forward, but still wrong.

3. I was just referring to current research that is related to the issue (GAN's for data generation to alleviate privacy concerns), and decided to briefly explain what they are to any interested reader, as I find them interesting.

Wow, really compelling arguments there, Dan!

Take a look at how sentencing guidelines work today -- what does it even mean to ask for "the entire dataset"? And of course, sometimes the bias is obvious (look, for example, at historical disparities in types of drugs).

What predictions about recidivism would you find permissible (heck, even relevant) that didn't rely on public data?

Tonight's trip down the rabbit hole taught me that retro 80's uToob synthwave videos get far more views than the actual 80's tracks they're cribbing from. I'm not sure what that means.

Split the difference on labor and bosses? In Germany for companies with over 2000 employees workers elect almost half the board of directors. With 500-2000 employees they elect a third.

Well, I'm glad you're happy to give a ton extra to your employer, who also likely benefits from favorable tax status and charitable contributions.

I am entitled to everyone's labor, thanks in advance.

**** my cheap ass employer with their "non-profit" status.

The point I'm trying to make (to einbert) you're going to lose people who agree with you when you state "labor deserves all".

Obviously the investors who invested the capital to build the store, the factory, the hospital, deserve (some) return on their investment.

I thought you all would enjoy this.

Giving paid family and maternity leave would be the death of private businesses in America!!

this is about new york, but the incompetence of american rail/mass transit agencies is unbelievable https://www.nytimes.com/2017/05/01/n...y-signals.html

im not sure it's even a question of money. there are insane amounts spent on ****ty projects. it's mainly just about having no idea what theyre doing.