What Ada did, based on all that data, aides said, was run 400...

>What Ada did, based on all that data, aides said, was run 400,000 simulations a day of what the race against Trump might look like.

Has this election revealed that the threat of an AI takeover can be prevented with memes?

washingtonpost.com/news/post-politics/wp/2016/11/09/clintons-data-driven-campaign-relied-heavily-on-an-algorithm-named-ada-what-didnt-she-see/

>Has this election revealed that the threat of an AI takeover can be prevented with memes?

>implying Tay didn't throw off the algorithms

As if Ada could ever stand up to Tay.

>what did she see
Rigged polls :^)
the dumb bastards fell for their own propaganda

Illuminati card game predicted this look up "Eliza" card

Lie long enough and you might just forgot the lie was a lie.

I mean, kind of

>Hillary wasted millions on an AI to tell her how to run her campaign

Fucking why?

clearly you forgot our covert agent

cheese brain!

JUST FUCK MY CAMPAIGN UP, AI FAMALAM

...

Ada isn't shit. It's just a garbage in, garbage out computer program. The polls were wrong, so the simulations were wrong. This is how it works

1. Input poll data
2. Figure out margin of error and weights for different demographic group like "white women with a college degree" and "Hispanic men without a college degree"
3. Run 40,000 Monte Carlo simulations
4. Print results (e.g. "we have a 70% chance of winning state X with these average voter results")
5. Interpret results in ways like "we should hold a rally with Beyonce in Ohio because Millennials and blacks love Beyonce and it's a close race there"

This is no different from what sports gambling websites have been doing for years.

So by rigging the polls, they ended up playing themselves?

When you see the American people as numbers, you're gonna lose. That's the difference between Trump and Clinton

They erroneously assumed that they would get 2 suburban/"moderate" republicans for every one working class democrat due to Trump's rhetoric and the fact that the media was busy telling everyone that he was "literally Hitler." Turns out that only works when a state has a sizeable amount of wealthy suburbanites to begin with.

And Sup Forums's ai gf still won!

>durr hurr le mysterious computer program with a name can tell all

it's a fucking program. It can only produce so much insight so far as you can programmatically provide it. So if you hire retarded developers, which I no doubt believe she did in all of her incompetency, you're going to be left with some basic ass statistics model that doesn't account for 98% of the variables involved in something as incredibly dynamic as a national presidential election.

I don't even know if the polls were consciously rigged. Independent pollsters have an interest in being accurate and their credibility took a big hit this election. From when I looked into most polls, the flawed assumptions were that minority turnout, primarily black turnout, would be the same as 2008/2012 levels and that female turnout would be significantly higher because women would want to vote for a female president. The reality is that black turnout reverted back to its historic mean and female turnout wasn't the tidal wave they expected.

The other big problem is that there do appear to have been shy Trump supporters and that "undecided" (many of whom were probably hidden Trump supporters) voters did break for Trump in the last week of the election. Ada's model probably projected most "undecided people to stay home. The most interesting story about polling this past year is that populist right wing movements like Trump and Brexit are significantly more popular when people are voting in secret and don't feel peer pressure to give a politically correct answer. Particularly college educated supporters are really afraid to say they support those causes. I'm "educated" and work in a liberal city, so I sure as fuck wasn't making my support for Trump public in the office around a bunch of cucks who would ostracize me. That's probably why Trump did better in automated and email polls compared with human pollsters as some people felt uncomfortable telling a person over the phone who they would vote for.

1. Polling is non-representative. Have you ever been polled? Has anyone you know ever been polled? Polling is notoriously bad

2. Weighting based on Obama turn outs. Sorry Hillary, but a 70 year old white woman is not gonna get the blacks out to vote like obama did.

I also have to imagine a large part of the bold polling was HIllary's people finding the results she wanted to hear, rather than giving her the truth.

...

Not even the most brilliant computer will give the correct answer when supplied with shitty data. Bad polls and incorrect assumptions about voter turnout resulted in incorrect predictions. Doesn't matter if you calculate it once or four hundred thousand times.

>what-didnt-she-see/

the data they didn't input.

what it says is that individually humanity cannot defeat AI, but as a group consciousness they can.

Polls usually are pretty good. As much shit as we give Nate Bronze, he was solid in 2012 because he looked at the polls (specifically the aggregate polls at the state level). Hillary also did win the popular vote like most polls projected. The final national result was within the margin of error. It's funny that Sup Forums's favorite poll, the LA Times/USC tracker, was easily the least accurate major poll in the election. It was good for measuring momentum, but damn were those results wrong.

The question for pollsters in the future is how to make sense of right wing populism that inspires massive voter enthusiasm. That's where they fucked up bigly in this election.