Google admits citing 4chan to spread fake Vegas shooter news
arstechnica.com
arstechnica.com
This is not some issue that you can fix. This is not some color being off, or some chat message appearing out of order. This is about AI in its continued infancy that’s now being actively and maliciously targeted.
It’s also about a huge responsibility of the worlds two greatest media outlets that both think they can have machines do a humans job and miserably failing at it.
Some users unable to log in? That’s an “issue” you “fix”.
Bringing the world closest it has been to the brink of a world war in the last 30 years, that’s the point where you step back and f'ing take responsibility for your actions and start rethinking your approach.
Sorry for the harsh words. I’m probably overreacting.
But generally, old jobs going away make room for new jobs coming in. And it seems to me like we have a whole mountain of new potential jobs in the content curation and moderation area. We see time and time again that this is something machines are not better at, but unfortunately, companies are very resistant to realizing their approach is inadequate.
I feel this is largely because they have no reason to. Due to a couple of laws, Internet platforms are largely not liable for these sorts of things, so they have no incentive to hire people and/or do a better job.
To "fix" this, we either need to reduce the allowed error rate, or enforce a disclaimer which states something like "This information is collected automatically by a program. No human reviews this, so please do your own research."
Using old terminology like "media outlet" does nothing to address the problem, all it does is frame the issue in a way that supports the narrative that these companies are mere aggregators that can bear no responsibility for the information they are disseminating.
Well, frankly, that's bullshit. You have to look at the actual effects in order to make any kind of reasonable judgement on the situation. Google and Facebook are a huge contributor to the fake news problem, and despite how difficult and maybe intractable the problem is, they are also uniquely positioned to be able to make a difference. They deserve to be held to a higher standard, and regardless of their protests, they at least have the resources to try harder.
I'm not absolving Google and Facebook. There are more trustworthy places to get your news and information. Why do you think people flock to less reliable sources? Unless that's solved, democracy and with it culture and progress as we know it is destined to be doomed.
Wisdom and knowledge do not come from technology. They come from wise and experienced educators working one-on-one with their students, not to tell them the truth, but to help them experience the truth. Nullius in verba!
Ed-tech is confidence game.
What technology offers is the productive capacity to support a world filled with actual human educators and actual human intelligence.
Indeed. If we do end up with some sort of "malicious AI apocalypse", then it will be because some humans wanted it to happen.
Based on the story linked, this event was the result of a situation occurring that google engineers did not anticipate. It's just an edge case that was not handled properly.
There's nothing "AI specific" about that. Clearly google is going to continue using algorithms to rank pages, and clearly this will sometimes return bad results.
The real issue is that any system which attempts to organize or curate untrusted information is subject to adversarial attacks.
AI up until this point, has been mostly trained on labeled, human curated data sets. But as it starts to get trained on real user input, it'll be subject to attacks by people trying to game it. AI needs to address this robustly, but it's in it's infancy as you say.
I mean, we're still working on bias issues like machines learning that "doctor" is "male" because you trained on a corpus of English text, and therefore, most of the input text referenced a doctor as male, and that's the AI getting poisoned by cultural biases, which wasn't even an organized attack.
> Something happens.
> /pol/, being /pol/, flipped the switch that generates semi-random information around a subject, seasoned with bias-of-the-day
> some posters on /pol/ decide to blame Geary Danley for the Vegas shooting
> google indexes /pol/
Given that:
> Searches for _words_ yield content related to _words_
> Searches for a name yield content related to a name
> Searches for "Geary Danley" yield content related to "Geary Danley"
This ensues:
> Media shitstorm because google is "citing 4chan to spread fake Vegas shooter news"
This only happens because:
1) We are expecting google to feed us only the truth? Otherwise we would say "someone took google seriously and arrived to the wrong conclusions"
2) We don't care the slightest what URL we follow from google? Otherwise we would say "someone who doesn't know what /pol/ is is taking it seriously"
3) We find it easier to blame some conspiracy than to take a step back and think? We ascribe blame to google.
4) All of the above?
I feel I am living in some bizarro world where everyone's feelings and expectations must be met, and any deviation from this will result in riots and name calling[0].
Few things exist to serve your purposes. Think before using any tool. You wouldn't use a blowtorch to trim your nails.
[0] https://phys.org/news/2011-11-poop-throwing-chimps-intellige...
It shouldn't be Google's job to vet articles. That should fall on the end user, and this push to hide "fake news" is just asking for more problems down the road.
Do we crucify Google for surfacing an expost facto false story?
It's not Google's job to verify sources of news. Could they algorithmically "guess" a credibility score and attach that to the newslink, sure, I suppose.
Then why bother with other sources? Just generate a random summary based on the search terms, rank it, and present it to the user if it is credible enough. Repeat as necessary.
Anyone including "4chan" in a list of media outlets is an idiot. (Or whoever enacts a system that automatically includes 4chan in a list of media outlets ... it's not important to users how they got to that result.)
Does anyone think 4chan articles belong there?
Yes, Google is a search engine. Everybody who's become accustomed to modern computing technology over the past ~20 years is very familiar with what it is and what it does. At least the more technically-inclined and aware do. To you, it's a cold, logical machine that provides best-efforts based upon their heavily researched proprietary algorithms. That's absolutely correct. To others, however, it's become something else.
To illustrate:
For many people, especially the most vulnerable to being persuaded by search engine results, Google as become a word as ubiquitous as Kleenex.
When your nose is running, you need a "Kleenex", even if you mean "tissue". You probably don't care ultimately if the tissue is Kleenex brand as long as it solves your problem.
When somebody says they want to "Google" something, they mean "google". They aren't concerned, necessarily (though bias lives even in pop culture over Bing), whether you use Yahoo!, Bing, Ask.com, or Lycos (it apparently still exists, at least in spirit).
To that second category of people I'm using to illustrate, the problem they want solved is that they want an answer. They've become accustomed to "googling" something delivering them an answer they can generally rely upon. It's a learned habit/behaviour and, I'd argue, hardly their fault.
Yes, North American culture could do well with become more critical in its thinking and that would probably solve this issue. Sorrily, getting there isn't so simple. In fact, it's probably a compound problem in light of who is aptly taking advantage of that system of habit and trust.
Funny enough, those exploiting this problem are the same who would prefer to remove even further the ability to think critically in US society (and others), by means of cutting back on public services like healthcare and education so that bare survival becomes the mode.
I'm starting to get off topic now, so I'll leave it there.
I don't have a direct solution. I'm not sure there is one. Google is as Google does, and yes people need to improve at thinking things through, especially what they read. But if you just resign the problem to that last point, I really don't think we'll get anywhere at all.
In the context of news truth is more relevant than lies, because news is an account of recent events, not fiction.
So if I go to Google News and it displays fiction as if it was news then something has clearly gone wrong, and I don't think Google denies that for a second.
To be fair, that is how Google presents things like the snippets. There's no disclaimer for the answers.
Normal non tech users wouldn't know if these were automated and highly fallible, or manually curated and somewhat reliable.
Here's my go-to example, where Google says a quarter is worth 50 cents. https://imgur.com/a/oibA7
No disclaimer or indication that this is at best, a bad guess. Also, it's been this way for months and months, even though I post it everywhere as an example. I also clicked the feedback button quite some time ago and reported it.
I understand this case wasn't a snippet. But it was labelled "Top Stories" and probably carousel featured. That sort of thing, like rich snippets, implies Google is presenting it as something more than just a search result.
The issue was that the data was being used in Google News, not Google Search. This doesn't just impact "what URL we follow", but how other pages get pulled in as "relevant". So we can be presented with "something terrible happened" from a news site, and "here's what we know about john smith" as relevant to the topic - without being told that the leap from 'something terrible' and 'john smith' was made by /pol/.
Yes, people should be capable of determining that 4chan isn't actually a news source, but that doesn't mean it's defensible for a Google feature intended to surface news stories to be so poor that it's trivially easy to bomb it into highlighting posts on the world's most infamous bulletin board as "top stories", even if the search term is really specific.
I wouldn't use a blowtorch to trim my nails. I would use Google with the expectation that an internet search company capable of writing software that drives cars and performs real time translation is also capable of not labelling a website which is both extremely well known and trivially identifiable as NSFW user-generated content as a source for a "top story"
Should google not index 4chan?
The whole "fake news" phenomenon is even more disheartening. That the public is getting news from their Facebook or Twitter feed and think that is sufficient speaks to a major failure of our educational system and a lack of development of critical thinking and analysis.
Andrew Ng interviewed Ian Goodfellow in the first course of his deep learning specialization on Coursera. Ian said that he's really interested in machine learning security, in part guarding against untrusted or adversarial inputs.
Whether they're poisoning your network in the first place, or your network is misclassifying them because they're too weird, it seems like you need to treat random inputs as potentially adversarial.
Even more reputable sites like Snopes tend to draw absolutist conclusions about things which I find are certainly not absolute.
Google and Facebook are culpable, but the news consumer is the problem. If you believe any information put in front of you, we've already lost.
EDIT: @mikeash: We don't give up. I just don't know what the solution is ¯\_(ツ)_/¯
@projekti: If "it doesn't scale", your only option is to meet with your representatives with a lobbying group, and introduce legislation to bring Google, Facebook, and other tech companies under US government oversight with legislation. They will not do this willingly, except possibly under the threat of regulation.
29% of LA Republicans blame Obama for the federal government's response to Katrina. http://www.nola.com/politics/index.ssf/2013/08/polls_show_lo...
It's fundamentally responsible for news media to take a subject like vaccines or "was Obama president during Hurricane Katrina?" and give a solid, well-sourced answer instead of "well one side says X and the other says Y".
The problem is beyond repair then.
> It's fundamentally responsible for news media to take a subject like vaccines or "was Obama president during Hurricane Katrina?" and give a solid, well-sourced answer instead of "well one side says X and the other says Y".
The first amendment has no such exceptions.
EDIT: (hn throttling limits kicked in)
> I said "it's fundamentally responsible", not "they should be legally compelled by the government".
If your policy doesn't exist in law, it doesn't exist; it's an unfunded mandate with no teeth. We might as well wish for unicorns that bring us gold.
That's silly. I said "it's fundamentally responsible", not "they should be legally compelled by the government".
One of two things will happen: 1) They want a summary, and denied that, will lose interest. 2) They will look for the summary that reaffirms their viewpoint and discard the rest (echo chamber/bias).
I mean, sounds to me like you're both saying roughly the same thing. The question is what is the solution to the problem of an electorate which is unable or unwilling to think critically?
I'm of the opinion that fact checks from reputable sources are beneficial when tackling moderate/undecided voters. No, hard-core "Sandy Hook was a false flag" folks are never going to be convinced, but they're not the audience for sane coverage anyways.
"Think of the children", etc.
A centrally controlled whitelist has as many (or more) risk than a pagerank algorithm. Moral panics about the gatekeeper's blacklist not being all-encompassing is also of questionable value... especially in regards to an event in the very early hours of a high-traffic news story which are generally full of incomplete information.
We have the right to a free press. Our responsibilities include vetting our information sources, responding to bad press, and providing better press should the existing press fail in its duties.
We have the right of free speech. Our responsibilities include countering dishonest speech, being mindful of our own speech, and defending the right to free speech.
We have the right to bear arms. Our responsibilities include keeping our arms reasonably repaired and secured, being educated in their safe use, and encouraging safe use from others.
The list goes on. To paraphrase a certain superhero, with great rights come great responsibilities. The system appears to fail when people ignore their responsibilities. It seems to fail when we rely on others to fulfill our duties. Every member of that society is obligated to those responsibilities, though shirking their duty has become the norm.
I don't have a solution and I am not a skilled writer, but hopefully this gibberish makes sense.
To say that the information distributors can publish whatever they want and the responsibility of information accuracy and quality must fall on the layuser, that's just poppycock. That's just a way for the giants to save money while we all get questionable data with which to live our lives. Unacceptable.
Better systems might require more work to mark a story as false (such as how reporting a story as false currently works in Facebook.) Over time, Facebook can see on their end which domains are pushing fake news and choose not to surface unreliable sources on people's news feeds.
I sort of think the radical idea is that a "news aggregator" is some new thing that can operate without editorial standards. People should be mocking Google for the shit job they do highlighting content.
I live in Sweden. How do I fact check what happens in a neighborhood in the USA? How do I know that a result of a clinical test is reliable?
Do I check it on another on-line source? Do I spend millions on R&D and do my own analysis? Or just few thousands going to Las Vegas and gathering my own data? Or do I need to wait for a Newspaper that actually has fact-checking?
> Even more reputable sites like Snopes tend to draw absolutist conclusions about things which I find are certainly not absolute.
Billions of citizens fact-checking every single news is not cost effective. It will be good to have a minimum of quality, and then you can apply any correction to the inherent bias of that medium. There is a difference between someway biased information and bullshit.
I question the utility of you having any opinion on US politics, especially at the local level. You are certainly entitled to one but I don't know what it gets you without a vote.
(Not to mention that a Swede can have a very legitimate, personal, vested interest in what's going on in the US, maybe the GP had family vacationing in Vegas. - I had family vacationing in France during the Paris bombings, my mother-in-law called me up frantically that day.)
https://travel.state.gov/content/passports/en/abroad/legal-m...
Maybe I have friends and family there.
Maybe I used to live and work there.
Maybe I have a long distance relationship with someone there.
There are millions of reasons why someone would be interesting in knowing what is happening in a place other than which they current live.
https://skeptics.stackexchange.com/questions/9423/did-charli...
https://skeptics.stackexchange.com/questions/2094/is-a-write...
If you want real fake news, Google any semi-obscure health issue. You'll be lucky to find one or two .gov's among sketchy sites peddling homeopoathic cures.
I believe that's only on /b/.
You can check facts the exact same way we've been doing for many years - find corroborating sources, view first hand evidence (and no, you don't need to do this in person), look for analysis from trusted sources etc etc.
Even if a newspaper claims to do fact-checking or is somehow legally obligated to do fact-checking, that's absolutely no guarantee that it's facts are accurate or free from bias.
To make this statement is akin to saying that the opinion of experts is useless. This is exactly what has lead to the flood of conspiracy theories, malicious mis-statements, and general degradation of the media commons that has put us in our present state.
I do not have the infinite amount of time necessary to fact check the flood of bullshit in social media, much less deliberate BS in politically motivated media. I rely on curators and fact checkers, experts and analysts to winnow the very narrow set of confirmable facts and conclusions from the vast sewer that you would prefer.
That's not at all what I'm saying - absolutely, look at the opinions of experts, just don't expect a news service to present all the ideal experts to you in one place. Go out and seek those experts on your own and verify who you think is credible and where their existing biases are.
This very quickly becomes more than a full-time job, much more even than a single full-time person can handle. Asking every person interested in world events to spend more than eight hours a day determining the veracity of everything they read, well, it's not a reasonably proposition.
If you expect some single organization or worse, crappy AI to put together the perfect list of experts for you, I'd be deeply concerned about your susceptibility to this kind of thing in general.
You simply cannot take this position in the comments section about a story in which Google very, very specifically fails quite badly to represent factual information.
The worst offenders were typically politically slanted (as would be expected in any human run system or ideological subject) but not only that as many academic subjects are full of internal debates with varying priorities depending on context, which made me question the utility of relying on 'trusted' 3rd-parties like Snopes for fact-checking.
If the fact-checking sources are as editorialized or based on the author's worldview on what is 'true' then it's just as bad as the root source. What then is the point?
The utility is that they offer a bit of extended research on specific claims, which larger articles or speeches often can't go into for the sake of brevity, lack of research, or occasionally (and likely less so) intentional deception. Their utility should not be fact-checking in any absolute terms.
4chan has had useful research on a number of occasions, that have gone beyond other journalistic sources, but more often the motivation is not the truth but entertainment and like any open forum has it's share of conspiracies. It should neither be entirely dismissed or given some inherent 'authoritative' credibility. This is why context matters, a context currently beyond the scope of machines, and too dangerous for a few 'trusted' humans. There's no absolute authorities on a general internet-wide scale.
Like what?
From the text : "These guys weren't ISIS, just opposing Assad and Russia."
[1] http://slatestarcodex.com/2017/02/22/repost-the-non-libertar...
> [insert internet company here] did nothing wrong! It's the reader's responsibility to think critically!
Obviously, there's a lot of truth to that. I'm sure everyone here broadly agrees with the concept of personal responsibility. I just don't see how making this blatantly obvious point gets us anywhere. It's like working with a bad colleague who management won't fire. Of course they're failing at their duties, but so what? You're stuck with them. It seems weirdly nihilistic to just accept the situation, instead of trying to improve it by making the colleague suck less.
Personally, I think it's OK to trust experts. We broadly trust doctors, engineers, researchers and lawyers to inform us about their specific fields. Journalism is different, in part because of the 1st amendment : the government can regulate who's qualified to give you legal advice, but not who's qualified to inform you about geopolitics. That's a feature, with some undesirable side effects. My contention is that curated fact checking is an acceptable, if imperfect, way to mitigate those side effects. Everyone's entitled to an opinion, but not every opinion is entitled to the front page of Google.
If they can't just rule out an image board famous for trolling, why should I buy all these stories about ai, machine learning and all the buzzwords.
So many fake news sites are using high quality stock images fo OG share, and we are more or less conditioned to believe a headline with a high quality image attached, right?