Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[MetalSlimeHunt]

I'm willing to do damn near anything to get people to vote. I see the list of electoral expansion ideas and it's just like "all of the above, now please". Mandatory voting, election holiday, employers have to schedule a voting block, college and high school active sites. Whatever you want. Execute people who write-in Mickey Mouse, I don't know. As long as it gets the turnout, because large-population democracies are killed by that turnout loss.

So why not write-in "instant runoff" or "Batman" or even this asshole.

This reminds me, after doing it on a whim for a few years I've decided it's going to be tradition that I will always write-in myself for Soil and Water Commissioner because nobody is ever on the ballot for it.

[Rolepgeek]

There's a very good reason for that, MSH.

We will never surrender to Persian rule.

[alway]

I frankly trust exactly zero reports on "paid/bused protestors" after the Charlotte protests. The police commissioner lied straight to everybody's faces (shocking) and claimed 80% weren't from North Carolina at all when in reality not only were most arrested people Charlotte residents, all of them were from the city and immediately surrounding areas.

The cops are liars. This is demonstrable fact. Any claims they make as to the identities of the people they've arrested is inherently untrustworthy, and there is nothing legally obligating them to tell the press the truth on these matters. Fuck the police and FTFE.

Ah.

The more I hear stuff like this, the more I think that, unless we employ AI's to manage our media in order to achieve re-convergence of society, we're pretty much screwed.

As someone in tech:
NONONONONONONOOOO BAD, NOPE, NUH UH, KEEP AI AWAY FROM STUFF LIKE THAT.

Basically, AI is just a regurgitation of training data, which is usually biased towards the most well-represented demographics among the creators. Bias in AI systems is a huge systemic issue, even in things as seemingly innocuous as speech-to-text software. Any sort of skew in the training data will result in poor optimizations for certain subsets. With a population-accurate sample of the US with 17% dark skinned people, an algorithm (ex: image recognition) trained on that data set will likely perform worse for that subset, since a small gain in accuracy for 83% is a bigger deal than a relatively large loss in accuracy for the 17%. A good deal worse if trained primarily on (and tested afterwards for accuracy with) typical Silicon Valley white male employees.

!!!Which, as it turns out, is the case for software *actually used to determine criminal sentencing*!!!

The best description I've heard for AI systems in situations like these is "Bias Laundering." Clearly (sarcasm) there's no racism or similar involved, because it's just math and complex systems! That just so happen to propagate existing racial stereotypes because that's what they were implicitly trained to do, using data sets taken from a system with heavy biases. You can't question it because the data sets are often corporate secrets and the models involved too complex to actually understand what's going on in the machine's head, and no one would think to anyway, since they aren't technical enough to understand just how bad AI systems are at doing this sort of thing in a socially responsible way.

Put them in charge of media, and we will long for the days of simple fake news stories propagated by Facebook's algorithms. Or rather we won't, because everything we read will make perfect sense to us despite being completely inaccurate and based on opaque systemic biases.

[Rolan7]

...  Did anyone else just think of Raiden screaming about "la li lu le lo" during naked cartwheels?
Oookay, that's enough Pepsi for one night.

(The reference is Metal Gear Solid 2 nanomachines son)

[wierd]

The differences between this and our current president are academic.
https://techcrunch.com/2016/03/24/microsoft-silences-its-new-a-i-bot-tay-after-twitter-users-teach-it-racism/

[MetalSlimeHunt]

...  Did anyone else just think of Raiden screaming about "la li lu le lo" during naked cartwheels?
Oookay, that's enough Pepsi for one night.

(The reference is Metal Gear Solid 2 nanomachines son)

Don't laugh, MGS2 is more relevant than ever. I recommend everybody go re/play it if you have the chance.

Solidus is a better Prez than Trump, that's for sure. We're just not going to talk about Armstrong verbatim saying "make America great again".

[Rolan7]

aren't we though (14 second clip) (2013 game)
https://www.youtube.com/watch?v=TEM5hW6BMJA

Nope.

Ah right.  So, technically down, but "only" half a percent...
We were seeing numbers about raw votes for Republican and Democrat, which did go down considerably, because the third parties apparently got 4.3% of the vote (up from 1.4%).

Really impressed that Stein got an even 1% O_o

[Frumple]

For what it's worth MSH, if it was intended as a reference it was probably aimed at reagan. Specific construction of words got coined back in the 80s, used by reagan, then... bill in '92. What.

I... hadn't cared enough about the phrase to check. And didn't remember. I'm not sure what exactly my reaction should be to the fact that apparently parts of GOP spent this election emblazoned with a clinton campaign slogan. Even if it was a reagan one, too.

[MetalSlimeHunt]

I know it's a Reagan reference, but if anybody else may wield the meme magic it'd be Kojima.

...Let's just pass an Amendment to make Kojima the Eternal Guardian of both Japan and America, it's preferable to this at least. We'll go bankrupt building giant robots, but whatever, Trump will probably approve.

[Sergarr]

I frankly trust exactly zero reports on "paid/bused protestors" after the Charlotte protests. The police commissioner lied straight to everybody's faces (shocking) and claimed 80% weren't from North Carolina at all when in reality not only were most arrested people Charlotte residents, all of them were from the city and immediately surrounding areas.

The cops are liars. This is demonstrable fact. Any claims they make as to the identities of the people they've arrested is inherently untrustworthy, and there is nothing legally obligating them to tell the press the truth on these matters. Fuck the police and FTFE.

Ah.

The more I hear stuff like this, the more I think that, unless we employ AI's to manage our media in order to achieve re-convergence of society, we're pretty much screwed.

As someone in tech:
NONONONONONONOOOO BAD, NOPE, NUH UH, KEEP AI AWAY FROM STUFF LIKE THAT.

Basically, AI is just a regurgitation of training data, which is usually biased towards the most well-represented demographics among the creators. Bias in AI systems is a huge systemic issue, even in things as seemingly innocuous as speech-to-text software. Any sort of skew in the training data will result in poor optimizations for certain subsets. With a population-accurate sample of the US with 17% dark skinned people, an algorithm (ex: image recognition) trained on that data set will likely perform worse for that subset, since a small gain in accuracy for 83% is a bigger deal than a relatively large loss in accuracy for the 17%. A good deal worse if trained primarily on (and tested afterwards for accuracy with) typical Silicon Valley white male employees.

!!!Which, as it turns out, is the case for software *actually used to determine criminal sentencing*!!!

The best description I've heard for AI systems in situations like these is "Bias Laundering." Clearly (sarcasm) there's no racism or similar involved, because it's just math and complex systems! That just so happen to propagate existing racial stereotypes because that's what they were implicitly trained to do, using data sets taken from a system with heavy biases. You can't question it because the data sets are often corporate secrets and the models involved too complex to actually understand what's going on in the machine's head, and no one would think to anyway, since they aren't technical enough to understand just how bad AI systems are at doing this sort of thing in a socially responsible way.

Put them in charge of media, and we will long for the days of simple fake news stories propagated by Facebook's algorithms. Or rather we won't, because everything we read will make perfect sense to us despite being completely inaccurate and based on opaque systemic biases.

Well, the relatively small size of training data compared to the problem space is certainly a problem, but hopefully with future advances in unsupervised learning from unlabeled datasets, which would allow the AIs to learn directly from the source, i.e. entire Internet, without the human bottleneck, it'll be overcome.

I mean, if people are able to identify it as an issue, then there theoretically should be nothing that would prevent AI technology from doing the same, too, right?

[ChairmanPoo]

Dont the dutch have mandatory voting?

[PanH]

Well, the relatively small size of training data compared to the problem space is certainly a problem, but hopefully with future advances in unsupervised learning from unlabeled datasets, which would allow the AIs to learn directly from the source, i.e. entire Internet, without the human bottleneck, it'll be overcome.

I mean, if people are able to identify it as an issue, then there theoretically should be nothing that would prevent AI technology from doing the same, too, right?

It is recognized, and often understood in the model that describe it. However, it's hard right now to improve on the situation, but I agree that it's similar to bad use of the model : even though it seeks to generalize as much as possible, you have to take into account its shortcomings. In the Google case, it's more of a deliberate choice, which even though is not entirely satisfactory is one of the best available (you could chose another ofc, but there'd be other shortcomings). In the other case it seems to be like bad use of the rankings, or an issue with the model/data training. It is entirely possible that the training data was not good enough.

[martinuzz]

Dont the dutch have mandatory voting?

Nope. The Belgians do.

[MetalSlimeHunt]

Dont the dutch have mandatory voting?

Nope. The Belgians do.

Yeah, about that. Due to recent ontology cuts, we can only continue to support the existence of a single Low Country following the termination of 2016. The decision seemed like it was going to be routine and along the lines of the whole Thailand/Taiwan thing, but it's gotten stuck in committee and it frankly doesn't look like they're going to resolve in a month and a half.

If they don't have a diplomatic miracle Sanction is probably just going to drop some amat on the whole area and call it done, so if you could just sort it all out amongst yourselves before that happens, that would be great.

[Sheb]

That's not a problem, give me a shovel and twenty years of sea raise and I'll get rids of the netherlands.