Twitter to label deepfakes and other deceptive media
reuters.com
reuters.com
Per their blog, it appears they have several tiers of action but what standard label. I imagine that the first time people see "false" next to a clip that's real (and flattering to their opinions, perhaps), they're going to discount the warning pretty sharply.
This is exactly why the policy is needed.
If a baby is born with its spinal column outside of its body, a quivering mass of braindead flesh, unsustainable, destined for a slow and lingering death-- it should be terminated.
That is infanticide. Humane infanticide.
People going around yelling about "Ralph Northam endorsing infanticide" don't care about that. They don't care about the baby, and they CERTAINLY don't care about the mother.
They just want to stir up shit and get people yelling.
And because they assume that all people are as amoral (or immoral) as they are, people shouting about "infanticide" think that women are rushing to the doctor late in their third trimester to abort healthy babies for no good reason.
Many people find the flippant nature of his response disturbing. I'm not "anti-abortion" but I found the Northam situation squicky.
YMMV, and that's the beauty of living in a free and liberal society. Your comment would probably win hearts and minds if it was more about educating folks about the medical reality some parents are faced with instead of castigating and denigrating "the other side" (which doesn't exist, it's a spectrum).
"The baby is going to die, so this isn't murder, and it's going to suffer, so this is mercy" is not necessarily a background assumption. For some people, those claims are the debate. Deontologists actually exist, and some of them sincerely believe in souls, and eternal damnation or paradise. To them, the doctor is committing a mortal sin, and for the infant any life long enough for baptism can mean the difference between limbo and heaven. Whatever Northam's intent, they would genuinely find his position morally similar to any other infanticide.
I don't intend to argue for that position, but I think it's important to realize how quickly "simple" rulings on misinformation or malice can become judgements on entire ideologies. If Bentham, Kant, and the Pope would all disagree on whether a choice is ethical and humane, perhaps it's not a matter to be settled by Twitter moderation rules?
> People going around yelling about "Ralph Northam endorsing infanticide" don't care about that. They don't care about the baby, and they CERTAINLY don't care about the mother.
OK, but the problem here is that Ralph Northam, in the video, does endorse infanticide. There is nothing untrue about that analysis. Sure, there may be nothing wrong with that from my or your point of view, but you are tacking on your own narrative to the video all the same. Because you and I don't care about the mother, either, no matter what you may say to the contrary.
That viral video clip is just the VA governor saying some stuff during an interview on a radio program. You and the OP both attached your own critical analysis to that video, and that's great. Platforms which actively discourage this have alterior motives beyond sharing, beyond the truth, and beyond kindness.
You know who cares the least about the mother, out of all of us? Twitter.
I disagree. "Man endorses infanticide under x conditions" is the truth, and "Man endorses infanticide" isn't. Half of the truth isn't the truth.
> "If you saw someone about to murder a child, would you punch them to stop it?"
> "Yes I'd punch that person"
> This just in: AndrewUnmuted advocating violent assault and battery.
We would never stand for a journalist making that kind of report, why?
Because it's a contextomy of the worst kind that totally twists what was said, and is tantamount to a lie even if "technically" it's true.
Why should Twitter encourage the spread of that lie?
- You would still technically be committing the legal definition of those crimes, you just have a valid defense, under the rights afforded by defense of others
- You could aways weasel out by saying you're referring to the physical acts, not the legal definitions of those terms
Once you start arbitrarily shedding context, it becomes easier and easier to flat out lie, using the "truth", then weasel your way out once held to task for it.
They don't need the excuse or the train of action:
we don't like that => but it's true => label as lie => removed as lie.
They're allowed to go:
we don't like that => removed.
Their ability to remove a post has nothing to do with morality.
Morally they have no obligation to support hate speech and that alone already excludes them from supporting free speech.
But we can all play that game by picking very selective definitions despite knowing that is not what the other person means.
>They don't care about the baby, and they CERTAINLY don't care about the mother.
How do you know the intentions of the people making such statements? Is this any better than those who make similar assumptions about those who fight for pro-choice positions?
>They just want to stir up shit and get people yelling.
Also known as engaging voters to increase voter turn out. A valuable strategy to increase voter participation when my team does it, a despicable tactic to push partisan laws when the other team does it.
>And because they assume that all people are as amoral (or immoral) as they are
And now you are labeling them as amoral and hinting at possibly being immoral? This goes pretty well beyond engaging in honest discourse, no?
Because there's a historically consistent set of beliefs that goes along with being pro-life that includes de-funding medicare, public health, and SNAP thus letting new and expectant mothers fend for themselves.
Edit: There's a reason the US has the highest infant mortality rate in the developed world[1].
[1] https://www.healthsystemtracker.org/chart-collection/infant-...
Yes, some people are against abortion and many programs that stop the need for abortions. But even their reasoning is often based on their own beliefs that they are acting in the best interest of the child.
Also, there is a distinction between supporting laws preventing harmful actions and supporting laws causing supportive actions. For example, most everyone here would support banning killing of 10 year old children regardless of where in the world they live by citizens from their home country, but generally few here would support SNAP, WIC, and medicare, or their country equivalents, being expanded as is to cover all those same children. Does this mean they don't care to protect those children? Such views would quickly lead to the conclusion that either you fully dedicate yourself 100% to fixing every possible wrong or else you don't care about anyone.
It's euthanasia.
https://en.wikipedia.org/wiki/Adnan_Hajj_photographs_controv...
The recent Harry/Megan fiasco has provided a great set of examples on how the same action can be reported as a positive or negative depending on the paper's preferred spin: https://www.buzzfeednews.com/article/ellievhall/meghan-markl...
Several papers run opposite stories locally: DMG Media (owners of the Mail) notoriously did this with the HPV vaccine. Irish Mail were in favour, UK Daily Mail were against, on the same day.
The Sun vs Scottish Sun is another repeat offender.
I can't recall which paper (in the UK), but there was one that run both stories (whichever consumable it was both caused and prevented whichever disease it was) on the same day, same publication!
Because it's not even about spin, it's just 'content' getting written and published and read. I'm not sure what the (actual) newspaper equivalent phrase is for 'generating clicks', but that's all it is.
I see why they're in Vancouver.
The same person, diametric takes on the same subject. The only difference is I guess her preferred candidate didn’t do as well...
A less political example: there was a time when E3 was the most important week for the video game industry. It's far less important now.
I don't know much about the Iowa caucus, but it's entirely possible for something to be very important historically and also be irrelevant today.
Most of this looks like the twenty-four hour news cycle wanting a lock on credibility and deciding who gets to sit at the table. It's a shame, because that would have been a good discussion to have, but they've made their reputations so ... poor.
https://www.reddit.com/r/newscontradictions/top/?sort=top&t=...
Its fake, just because I said it, doesn't make the content true.
If you tweet that you are Man of the Year 2020, you are fake, a phony. The tweet, on the other hand, is simply a lie.
I'm not trying to be difficult, but is it not valid in this context to correlate "lie" with "fake" ?
Your example, a correctly attributed but false announcement about yourself, is not itself "fake" in any way - it's just a lie. There's no question over what it is - an assertion made by you.
But the original example:
"You just need to take a picture of X candidate, stick a made up quote beneath them, and put it on Facebook...
...creeps into "fake" territory because the image is an object with an implicit lie about its origins and motivation. The implicit lie - "I was created in order to raise awareness of a disturbing truth about candidate X" - is distinct from the explicit lie (candidate X said <thing>). Hence the superficially superfluous act of putting the text on a picture, instead of just tweeting "candidate X said <thing>".
The distinction is even more stark with the followup addition: "the fake-ness is even more effective, if the said candidate is the originator and controls the narrative." Here the explicit lie remains identical, but the implicit lie deepens considerably ("look at the fibs the crazy opposition is spreading around").
The problem with these kinds of censorship actions is that there is literally no end to it, and it will never solve the real root cause of the problem. It will just suck up resources and people/institutions will keep complaining that they got into trouble because someone took a tweet as a truth...
This is so true! I was on twitter recently and one of the recommended tweets was a factual statement without any source. I was shocked that only maybe 1 or 2 out of hundreds of replies said anything about not being able to confirm the statement. The vast majority of replies were about what the statement did or didn't prove.
If you know anybody who's deeply into facts, evidence, and rational discourse, please point them toward my latest project, https://en.howtruthful.com/ and have them watch the introduction video linked at the bottom of the page. After looking at twitter I'm thinking it will be hard for me to find people like that.
They achieved a reasonable degree of success, so I suppose lulz were gained.
Alright, fine. What if the attribution is just plain mistaken in a misleading way? What if you put an inspirational quote next to someone it applies to, instead of the speaker? If the speaker couldn't possibly have said it, or the quote is famously from someone else, so the goal is comedy? Or the countless pictures misattributing that "...fire inside me..." quote which escaped from Fallout: New Vegas? (Is the deceptiveness different when it's shared as a mistake vs meme vs prank? Can your algorithm tell?)
Twitter's put at least some thought into that, thankfully. They talk about "significantly" altered content and whether it's "shared in a deceptive manner". But good lord is that a fuzzy thing to decide or automate.
Basically what most consider the famous words ever said by X person would fall under these rules.
I always chuckled at the idea that deepfakes are going to be the problem which a simple jpg already works today and is more effective than video because people will actually click it.
I think it would be a better idea to educate people about these things, just like when "Photoshop" and other similar software started to pick up, a lot of people started creating "fake images", nowadays I feel like most people are aware that you cannot trust a picture, because they can be so easily edited.
This argument is nonsense an innefective. Try and explain to your elders about what a deep fake is, and how to identify one.
The movie : "The Irishman" used a lot of the technology successfully to take years off the cast.
Applying labels to known false media is not the same as an opinion based 'fact' or alternative facts.
So don't come with this absolute strawman point of : "Ooo where is the line." - The line is at fake. Always has been. Whats been blurred is the idea that Facts are not Facts.
They are. An opinion is not a fact. A deepFake is Fake. Hence the name. And its about time people grew up and used their intelligence to realise whats a FACT and whats not a FACT. Whats real and whats fake.
It's a shame its up to companies like Twitter to do this for everyone, but clearly too many people believe everything they see & hear without question.
Ok where's the line on deceptive media? A deepfake obviously has concrete criteria to identify it. What about a little spin? Cropping out a piece of a picture or a few seconds of a video? Muting a crucial bit of audio? What about that silly scenario where a reporter goes out in a kayak to show flooded streets and a cheeky passerby walks behind them in knee-high galoshes? Those all seem deceptive. I don't trust Twitter to have my best interests at heart when deciding. Then again, I don't use Twitter so I don't really care.
If the 'spin' consists of a lie, then how could you even consider it true ? 10% false and 90% truth does not result in a diluted truth or half truth:
'Half True' is an oxymoron. The content is take as a whole.
If I said to you : "You need glasses with a light on them because at night its always dark" - Thats spin.
If I said : "The sun is gone forever so you need glasses with a light on them" - Thats bullshit.
Use your common sense. And if its a 'grey' area then you can't call it fake can you ?
https://media.tenor.com/images/c731ffc70a71d8a89b5e3581953f9...
Yeah man, we should just go on like that, remove more and more responsibility from people until they are tame enough that you can just give orders and they follow.
Lets say Politician A implies something without explicitly stating it. If Politician B said that A said what he only implied is B then lying? What happens if A then says B is lying about him saying it.
Is A lying when he said he didn't say it? Is B lying when he said A said something?
I am extremely dubious stuff like this could actually be done objectively. Sometimes A would be accused of lying and sometimes B would be. The rules would not be applied fairly and equally.
Personally, I did not find the use of the "technology" to be "successful." Like so many other ridiculous gimmicks used in movies to attract the attention of the press and the Academy, it may have at times achieved its very singular visual goals, but that doesn't matter one bit. Because you can't make Al Pacino and Robert De Niro - both in their 70s now - act like 30 year-olds.
Without the massive press push about this technology, the film would have likely suffered a lot more ridicule about this "de-aging" tactic. Hollywood will not be able to convince me that they have actually achieved something good here, until they are able to prove this is not just technology meant to expand the lifetime value generated from their beloved aging actors, for whom they have no true planned replacements.
Just pick a few pictures where A1 was holding A2's hand, and where B1 was not holding B2's hand. Now you have a story.
Guys who get labeled as fake who didn't deserve it and guys who should be labeled not getting it. All the false positive/negatives will have a terrible time.
You said well the consequences:
- False positives - the wrongfully labelled ones ;
- """hackers""" - because there's not enough amount of '"' to make it sound different but bare with me, it's just those who learn/are lucky to manipulate/circumvent the system;
Even if this are small percentage cases, it will be in the order of hundreds of thousands, if not millions of users.
How do you give support to this people? Is it going to be another Youtube fuck-fest where you end up talking to bots? Because this shouldn't be allowed - at all.
If you want to make automated control measures, then people should have the right to talk directly to a human who can actually solve their problems. Yet that seems to be impossible, because of the scale.
Twitter didn't label it, so it must be real!
> Twitter under its new policy will similarly apply a “false”
> warning label to any photos or videos that have been
> “significantly and deceptively altered or fabricated,”
> although it will not differentiate between the technologies
> used to manipulate a piece of media.
Depends if the praise they get for "at least they're doing something/not pretending the problem doesn't exist" outweighs the criticism they get for false positives, false negatives, inconsistency/bias, arbitrary judgements, response times, or labelling so many things fake that it becomes meaningless.
Way back in 1994, Time magazine published a photo of OJ Simpson that was photoshopped to make his skin darker [1] - but having some colour-correction of magazine cover photos is inevitable. The line between regular and deceptive colour correction is blurry, and some people will be upset no matter where you try to draw it.
[1] https://people.southwestern.edu/~bednarb/su_netWorks/project...
Meanwhile, there was an a very convincing deceptively-edited video circulating on Twitter which clamed to show parts had been sped up by overlaying the White House version on the original - it was very vivid, with the overlay seeming to jump ahead in fast-moving parts. This was an illusion created by running the videos 7 frames or so out of sync the whole way through and abusing the fact the slight differences weren't noticable in slow-moving parts during normal playback. No-one noticed. Youtube fighter of misinformation Captain Disillusion even helped spread it and gave his approval.
My general reckoning is that Twitter would get much more praise for labelling the non-tampered video as misleadingly edited and endorsing the accuracy of the video that was misleadingly edited to make the other one appear that way, than the other way around.
Attempting to label these so people aren't manipulated by it is something that deserves praise.
It's not a free pass of course, it does seem insanely hard, and it better materializes into something concrete with a real impact that does solve the issues at hand.
it's pretty incredible how we went from "give people the benefit of the doubt" to "people are manipulated".
I did not imply that everyone was de-facto manipulated by those. Only that it can already happen.
Except for all the purposefully manipulative things they allow on their platform and don't label or attempt to debunk.
You could say they're targeting the biggest problem first, but twitter isn't going to touch politician's tweets or the mainstream news en mass. If you're worried about the average twitter users' critical thinking skills, don't you think those important-sounding sources of information should be THE key targets of a anti-misleading or anti-manipulative campaigns?
It amazes me that people who consider themselves leftists are ok with massive corporations deciding what information they're able to see.
I used to be amazed, but not anymore. I've come to accept that people will always practice tribalism and will defy any principles to achieve their end.
Most of the time it ends up with Truth = Deep Pockets, i.e CNN, NY Times, Fox News, MSNBC, LA Times, etc are all "True" even if they do spread false info.
and Random YouTuber's or Blogs are labeled as "Fake" if if the info is true.
When deepfakes actually become easy to produce, and pervasive, then humans will stop trusting videos in the same way most folks do not blindly trust still images (an image of Trump's head perched on top of a Mr Olympia body would not fool anyone). At that point I doubt browsers will need to take any explicit censory action.
The fires of cognitive dissonance and a fixation on hate will be ended. Valid interpretations will be set by a handful of tech companies regardless of the value lost. This is the Thanos snap that has been coming, two halves of the populace into one. Hate is dead, long live the heartless.
Link: https://blog.twitter.com/en_us/topics/company/2020/new-appro...
The other 50% will be obvious jokes.
Out of curiosity, I browsed throw tons of (political) clips posted by BBC and did not see a similar warning (regarding UK goverment) or anything close to it.
And I just checked, the CBC gives a similar warning as well.
It's reasonable that Twitter should begin to be less permissive on their platform too, to get ahead of controversies that might occur on their platform in the upcoming election. Twitter political threads are still lawless wastelands full of incendiary comments from suspiciously bot-like accounts, so clearly something has to be done.
[1] https://www.telegraph.co.uk/technology/2019/12/17/google-fac...
Is the end result of this just going to be that deepfake generators face natural selection pressure to improve beyond the exclusion threshold of this feature?
What kind of verification is used to ensure that there are not false positives, or misuse for nefarious purposes?
You know twitter is a company, it has a CEO, a board of directors, and is beholden to the wishes of the shareholders. What happens if shareholders want certain types of content flagged in a way itself is deceptive or is conducive to false conclusions about the status or quality of content? What happens when governments apply pressure? What kinds of safeguards in place to prevent this kind of thing?
How do you secure such a system from becoming a tool of manipulation?
Hey, another thing in this topic: It would be very cool if a social network could somehow indicate if a shared/retweeted link had not been read for a while in advance. Imagine if a network would also have this system implemented for a good while, like a year, to then retroactively show this information on all links. I wonder what it would look like...
Your comment must be sarcastic as well. Please tell me that you don't actually believe that the government should hold a monopoly on truth...
FTFY.
2. have we completely given up on critical thinking and education?
3. free speech vs not so free speech. i guess we're choosing "not so free"?
2. False dichotomy - Twitter attempting to label deepfakes and deceptive media has no negative effect on critical thinking or education, just as Twitter not doing has no positive effect on either. Twitter is not setting education policy for any society or government, here.
3. There is no government or society in the world which does not place some limits on speech in some form, even the US has laws against terrorist threats, slander, libel, lying under oath, misleading labels on food and drugs, misleading contracts, etc ... demonstrating that one common and widely accepted basis for limiting speech is deception and misleading information.
Also, merely classifying data is not preventing its spread, and is not, therefore, a limit on free speech even in theory. If that were the case, then every tag on Twitter would also somehow be "not free speech."
Also, Twitter is a privately owned platform, not a government or a public square.
we have had this for the past 400 years (since the birth of the modern press). why do we need it now? how did you, for example, simply decide you can't rely on other people's good judgement? what triggered this change?
We've also had experts and arbiters of truth for 400 years. Why do we need to abandon the premise that truth has value in civil society?
>how did you, for example, simply decide you can't rely on other people's good judgement? what triggered this change?
I personally have never made any such decision.
Trust is unavoidable either way. Twitter wants people to trust their good judgement in determining what is and isn't a deepfake, fabricated of false information. If you choose not to trust Twitter, you inevitably choose to trust the source, or the community, or your own bias (created and fed from third party sources.) No one is able to independently and objectively verify every claim they encounter.
The only (as yet undetermined) relevant factor here is whether algorithms are any better at determining truthiness than the web zeitgeist. If so, then I see no harm done. If not, the feature will probably be abandoned.
I don't particularly buy into the implied third option that Twitter and other platforms will be using features like this to intentionally mislead the public in order to further some radical political agenda of misinformation and thought control.
The very same people that are for these types of rules have abandoned it. With things like "there is no one truth, only my truth" believing that there is more than 1 "truth"
Most of the time that type of narrative is being pushed by the same people that desire the tech companies to suppress certain kinds of speech
>>If you choose not to trust Twitter, you inevitably choose to trust the source, or the community, or your own bias
It has never been the case that you should trust anyone. No Twitter, or the new agencies or anyone. This is why people should be encouraged to look at original sources and multiple sources. This is also why in the age of everyone having only "anonymous sources said" is a huge problem.
The other problem is most news is just regulations of another outlet or the "news media" takes a single random tweet from some random person and runs 100 nearly duplicated stories sensationalized based on that single account with no verification.
>>whether algorithms are any better at determining truthiness than the web zeitgeist.
This has been tried many many times before, it always fails and this too will fail
Who are you quoting?
>It has never been the case that you should trust anyone. No Twitter, or the new agencies or anyone. This is why people should be encouraged to look at original sources and multiple sources. This is also why in the age of everyone having only "anonymous sources said" is a huge problem.
So what you're saying is, people should trust "original" sources and "multiple" sources.
But original sources can be misleading, deceptive or have ulterior motives, so how do you verify their authenticity without trusting a third party?
And if no single source can be trusted, how can you trust multiple sources? Surely all sources are equally untrustworthy, even if they all report the same events. How do you verify that a commonly reported narrative is not simply well orchestrated propaganda?
It has always been the case that trust is necessary. Before the internet, people trusted the news. Before television, people trusted newspapers, radio and newsreels. People trust Einstein, Bohr, Newton, Darwin, Euclid without personally doing the necessary experiments to rediscover the entire corpus of modern scientific thought from first principles. Before the printing press and the Enlightenment, people trusted the Clergy and town criers.
Subjective truth is always based on implicit trust. Anyone who believes otherwise about themselves is simply unaware of the systems in which they place their faith and from which they form their biases.
>The other problem is most news is just regulations of another outlet or the "news media" takes a single random tweet from some random person and runs 100 nearly duplicated stories sensationalized based on that single account with no verification.
Absurd hyperbole. "most news" is decidedly not unverified claims from random twitter accounts. I'm guessing you've got some specific instance in mind that you're trying to conflate into a general practice, but if you look at the headlines and news from any mainstream source (CNN, Fox, BBC, AlJazeera, what have you) the vast majority are not the sort of sensationalized, unverified, Twitter-based stories you're describing.
>This has been tried many many times before, it always fails and this too will fail
Do you have a whitepaper proving that an algorithmic solution to this problem is impossible? If so, let us see it. Otherwise, you're just posturing.
> how did you, for example, simply decide you can't rely on other people's good judgement?
I'm not sure what this means - I find I have to rely on the judgement of others simply in order to get on, as there isn't enough time to evaluate everything myself and it would be irresponsible to conclude that it's all false. That line of reasoning leads to concluding that Chicago doesn't exist because I've never seen it.
But I need to be incresingly careful about whose judgement I trust and in what subjects.
This is a terrible argument and people need to stop, at least you did not all use the trite "can't yell fire" non-sense but "they are a private company" is right up there with that.
>>Also, merely classifying data is not preventing its spread,
It is highly unlikely that this stops at just classification and tagging. I am sure there is some element of Suppression as well
Well, to be fair, it's all free speech.
The guy who uploaded the deep fake exercised his right to free expression to do so, and the private company that labeled it a deep fake also exercised their right to free expression. Technically speaking, this is an example of the radically open and unregulated nature of free speech in the private sector. Other people labeling your content as "lies" or whatever is a natural consequence of that radically open environment.
It would be "not so free" if the government sent people around to question you whenever you upload a deep fake. Like they do when people mouth off while in, say, a CounterStrike session, about what they're going to do to the other kids at school.
It's unsurprising that your argument isn't swaying any of the various corporate counsels. I mean, it's an argument that's purely theoretical in a world where the government has said they will hold you accountable for fake news and government agents siphon the communication feeds from your servers full time anyway.
Think of it this way, a threat to chop off someone's arms, is only effective if you have not already chopped off their arms.
Point being, they're going to be held accountable in any case. That's resulted in just an almost shotgun approach to political stuff to the point where they sometimes even shut down legit political candidate accounts. In addition, they're being radically transparent with the government agencies in the form of data feeds so they can never be said to have let election interference content through. (Because they gave that content to the government agents and the agents didn't evaluate it as election interference either.)
So they are no longer masters in their own homes in any case.
No idea why they don't compel 4chan to do the same? Politics maybe? That's just a guess. I'm thinking maybe the politicians have already run all the numbers and they know that the people who frequent 4chan are typically not the 65 year olds who swing elections. So who cares what the nugget heads do? Again though, that's just a guess.
But who knows really?
My take on the DC hearings is it was just a smoke screen because they're scared to death of a trend materializing that non-career politicians can get elected and actually do the bidding of those that elected them. That and their afraid of losing control of the narrative.
As for Twitter, I'm not sold on the damned if we do damned if we don't picture you're painting. I think Twitter has a definite agenda and they'd be policing speech and de-platforming those they don't agree with regardless of the politicians.
Twitter wants it both ways, they want to editorialize but they don’t want to be bound by the regulations in place for news publishers.
Right now we're going through a rough patch because the effort required to forge images/video is dropping and the old heuristic of "if it's a video it's probably legit" is going out the window. Simultaneously the internet is re-arranging who gets to gate keep mass communication and to what extent. Both these things are going to cause strife.
I think it's useful to remember what the printing press did to Christianity in Europe in the 1500s and what mass literacy did to monarchies in the late 18th through 19th century. Some blood was spilled but I think we can agree the printing press and mass literacy have been massive net positives to civilization.
I won't be losing any sleep over whether TwatBookGram does or doesn't censor anything. It will all work itself out in the end.