The Bittersweet Sweepstakes to Build an AI That Destroys Fake News

How cool is this?

Let's combine AI and Fake News in a single thread!!

Wired article - AI and Fake News

Here's snippet from the article. BTW - Wired is an *excellent* magazine and source of info:

A quarter-century after his self-driving car appeared in Byte magazine, Pomerleau is an adjunct professor at Carnegie Mellon, and last month, as so many lamented the role of fake news in the presidential election, he put a call out on Twitter, challenging the AI community to build an algorithm that could identify fake news and remove from it from online services like Twitter, Google, and Facebook. It was an open-ended bet, with Pomerleau putting down $1,000. And the community took him up on it.

Thoughts?

a reply to: Riffrafter

Maybe a large part of the problem is simply social media networks pushing news, fake or otherwise?

originally posted by: theantediluvian
a reply to: Riffrafter

Maybe a large part of the problem is simply social media networks pushing news, fake or otherwise?

What did you think of the Wired article?

How about Dean Pomerlau, adjunct professor at Carnegie Mellon? Interesting cat, n'est pas?

a reply to: Riffrafter

Sound impossible, unless the A.I will be using references to determine what is fake, which in turn could be derivatives of bias and false propaganda.

They're trying to sick AI on us before it can do it itself. With that as its foundation, Skynet nuking everyone might be a more preferable outcome to what would come from The Matrix.

originally posted by: omniEther
a reply to: Riffrafter

Sound impossible, unless the A.I will be using references to determine what is fake, which in turn could be derivatives of bias and false propaganda.

Ahhh....the GIGO tech acronym:

Garbage In - Garbage Out

And that is *always" the case when developing an AI Expert System. It's all about the data...

Now, if you want to try and use other types of AI like Fuzzy Logic or some version of Neural Network technology, that lessens the risk a little, but you still need to expose those systems to data (read: fake news) in order for it to learn. So once again you're back where you started.

Someone very unbiased and very smart needs to really vet the crap out of whatever is being used as an example of fake news as the system will only be as good as the info fed to it...

a reply to: Riffrafter

I agree with the author that they will fail. AI at this point is not up to the task of identifying deliberately fake news. I also agree that out of this something might arise that could be a tool to make fake news easier to identify. I like the idea of tracking viral clickbait hoaxes in databases like spam. I think spam filtering is probably the most closely related task.

I also like the paradigm of spam filtering in that it's of course opt-in, there's multiple competing vendors for solutions, etc.

a reply to: Riffrafter

what a joke

seriously how about research and not being a sheep to click bait articles.

Does this mean someone will feed it the "true version" who?

a reply to: jellyrev

The "true version?" "Pope Francis shocks world, endorses Donald Trump" doesn't have a "true version" per se as it's a hoax. There seems to be a lot of confusion about what "fake news" even is. I think this is a deliberate muddying of the waters. "Fakes news" isn't wrong opinion. It's not bias/spin/slant/etc or even getting a minor detail wrong. It's deliberte, outright, unabashed fabrication. In particular, what is being referred to by "fake news" are these sensational hoaxes used as clickbait by sites pretending to be news outlets. The absolute worst of these would be those that masquerade as legitimate media outlets like abcnews.com.co or foxnews.com.co.

It would be a very simple AI, it would just have to identify approved news articles originating from CNN or the AP, anything else would be wiped from the internet.

One good thing about this fake news matter, is that people are actually questioning the reality of propaganda, a very good start that I hope, leads to a general awakening.

What's wrong with people?

Just proves my views that we do need A.I. to function properly, because we're loosing those abilities ourselves.

The Weekly World News

Has been around since 1979 and in 37 years we somehow aren't able to understand what is real and what isn't?

I mean, even the news you get from Brian Williams isn't all true, so why would you believe EVERYTHING from CNN either without doing some kind of due diligence yourself?

I mean, I have to tell my kids that not everything on youtube is true, so why would it be any different in msm?

Instead of this Bullshat " fake news " concept, why can't we hold media accountable for pushing garbage?

No wonder people don't take accountability for their actions, they don't see anyone else being held to their own actions.

Fake News isn't new, it's called, don't be gulible and if something doesn't sound right or sounds too much, look into it, do the kind of research that would be required if you were in school.

a reply to: Riffrafter

Quite interesting!

I think that this will evolve into something similar to the spam email problem - in mathematical terms, a game of prey and predator. As such algorithms evolve, there are those who will attempt to game the system for their own purpose (social or economic, since fake news is designed to bring eyeballs/traffic to a site and hence money to the owner). This will result in tweaks to end the abuse ... and so on.

I would also like them to develop an anti-troll algorithm.

a reply to: Byrd

I certainly agree with you on the troll front. Some of the trolling is no different than a hate crime and should be handled as such. Freedom of speech is one thing, it's your right to have an opinion and to say it, however when your opinion is the constant belittling and written threats as well as deceptive actions to further that harassment, then you've moved into the criminal realm.

It's one thing for someone to say they hate a protesters shirt and opinion, it's another thing to create fake websites and social media profiles to further make fun of and slander that person.

The issue with fake news though is, it can easily be a matter of opinion and goes from being news to a editorial which isn't news but an opinion piece.

We used to have to include sources when writing reports and such when I was in school, but now with the internet, sources are no different than opinions and we can't say only stuff from mainstream guys is legit because we will miss out on the real news that the smaller guys get right.

Algos can be written but we need to have a method of either reputation based or a new system of people that can be paid as a 3rd party sort of like how pen testers get bounties when they show an exploit. We can provide bounties for people who provide the legitimate news sources or even facts backed up with more viable information, so if we don't believe it, we can read further into who everyone is, where the issue was 1st seen and who else reported it and other data that could help you either trust or not a article.

Just seems we need even more eyes looking into articles, a fact verifier.

originally posted by: Tranceopticalinclined
a reply to: Byrd
The issue with fake news though is, it can easily be a matter of opinion and goes from being news to a editorial which isn't news but an opinion piece.

They may need to develop standards for op-eds. More problematical are blogs; anyone can set up a blog (or could at one time) on sites like CNN or FOX and there was little oversight. Community newspapers can also acquire articles from freelancers - the problem being how to vet them since these freelancers aren't working under an editor and can't be held to standards by the threat of firing.

Algos can be written but we need to have a method of either reputation based

Agreed, but there has to be checking on that. In the game of Predator-and-Prey, a reputable source can be hacked. In fact (IMAGINARY SCENARIO) if the Chinese government decided to do a set up a news agent (to first become a trusted source and then start feeding propagandized fake news), it would be difficult to detect.

We can provide bounties for people who provide the legitimate news sources or even facts backed up with more viable information, so if we don't believe it, we can read further into who everyone is, where the issue was 1st seen and who else reported it and other data that could help you either trust or not a article.

Google has tried this recently with an add-on. I'm not sure that it's catching on, however.

originally posted by: JDmOKI
a reply to: Riffrafter

what a joke

seriously how about research and not being a sheep to click bait articles.

What are you talking about?

And who are you talking to?

And this *is* part of the research I'm doing re: fake news. I always research before posting. The fact that this article coupled the topic of fake news with AI was pretty interesting and I thought others may enjoy seeing this perspective.

Here's an idea: why not post - or at least allude to - your own research? Or even a little clarification on why you think something is a joke? From your post, it is not even clear what you consider a joke.

Take your own advice amigo...

a reply to: Byrd

I would also like them to develop an anti-troll algorithm.

From your lips/keyboard to God's ears as they say...

a reply to: Byrd

Very true, I work in the SEO ( Search Engine Optimization ) sector of the interwebs and have a good bit of experience with press relations and doing " guest posting ".

It's absolutely true that even CNN and other big names get their articles from joe schmo sometimes. It's too easy to do and really, Google is a big part of this issue.

It has to do with ranking your site in the search results, they know that if they include certain keywords and are backlinked enough by bigger names ( also not hard to do using blackhat tactics ).

All you need to rank high is lots of backlinks from bigger known sites, content that others will engage with or just not click the back button and traffic.

Google has dropped the ball a bit in terms of Search Engine Matters which easily relates to the FakeNews issues of today. That webspam was supposed to get buried on page 100 or worse.

Google needs to get better at burying these type of manipulations but hasn't found a good way to combat it yet. They recently removed basically a spam algorithm ( penguin ) that would penalize or even deindex sites that were found to be using spam tactics to garner lots of links back to their site which in turns tells Google this site is trusted by many others and will rank them higher on SERPs ( Search Engine Results Page )

Google is going to start loosing their ground to other engines if they can't combat this issue fast enough. Because while these tactics help FakeNews sites and others rank better, it's burying the small business websites because of their very nature that they are smaller and less established.

Our internet in general needs some more measures in place to help with the issues I've mentioned. Almost like search results need more categorization or another form of seeing through the manipulation, which is hard when people are making money learning the best ways to manipulate as well Google makes tons through those same buried small business sites paying loads into Adwords as their only means to been seen.

It's a really weird wild west situation where the larger entities are making money hand over fists.

a reply to: theantediluvian

Agreed completely!