Disinformation 2020 – US elections, AI, deepfakes, manipulation
An Oxford study of the 2018 US elections showed most of social media fake news and election interference came from domestic US sources.
Regardless, as 2020 US election frenzy begins, prepare for an ever louder cacophony of howls on Russian meddling.
Almost everything we experience, beyond what we directly encounter in our immediate surroundings, is a narrative – we have to rely on someone else to provide inputs and updates. And that opens the door to manipulation and disinformation.
Artificial Intelligence (AI) is beginning to proliferate in the generation of inexpensive deepfake videos and realistic fake news that can fool viewers and readers.
Deep-learning fake video, or deepfake video, is a powerful AI disinformation tool. Deepfake videos are false, synthesized videos generated by artificial intelligence. Very realistic fake videos can now be created of politicians saying or doing things they never did in real life.
For example, US House Speaker Nancy Pelosi was shown slurring her speech as though drunk. This was a crude example of deepfake. Things have gotten much ‘better’ and much more sophisticated since then.
Anything is possible in this new realm of ultra-realistic fake videos. And if we are unable to detect fake videos, we may lose trust in videos and in everything we see and hear from others. That could be the end of perceived truth in all events we do not experience first-hand. We may simply stop believing.
Our conception of reality itself may soon be under deep threat.
Not much good can come of this, but the genie is already out of the bottle. Can this be the collapse of world order and the nascence of some bleak anti-utopia? Unfortunately, only time will tell.
So how did all this suddenly come to pass?
Deepfakes rely on computer systems called neural networks, which try to mimic the human brain in order to recognize patterns.
Neural networks classify and group together inputs to the system. Sample data is ‘learned’ by the system and new input data that has similarities to one of these data samples is grouped with that particular data sample, creating clusters of data with similarities to each other.
Thus, neural networks use machine perception to label and classify sensor inputs. They can classify fresh data after they have trained on a sample dataset. Neural networks are components of machine learning.
This has progressed to the extent that inexpensive systems are now available for general use, finding applications in generation of deepfakes.
Deepfakes use two machine learning models competing against each other: one model, the ‘forger’ trains on a dataset and begins generating fakes. The second machine learning model, the ‘detector’ tries to detect fakes generated by the first one.
After each round, or iteration, the forger tries to learn why it was unable to fool the detector, and uses that learning to create the next fake. This process is repeated, usually with increasingly more realistic fakes, till the forger fools the detector. Such systems are called generative adversarial networks (GANs).
The larger the data set for the forger to train on, the easier it is to create a high quality deepfake that will fool the detector. This makes deepfakes particularly dangerous for public figures such as politicians, business leaders, movie stars and so on, with a lot more video available of them, which means a lot more training data for the forger.
Deepfakes are very difficult to detect. Any technique that can be used to detect deepfakes can be used to train deepfake GANs, which can then beat those forensic techniques. The deepfake machines learn all the time.
For example, lack of eye-blinking, lighting mismatches, or differences in voice characteristics (when compared to the original), were used to detect early deepfakes. The systems have learned and current deepfakes easily pass those old forensic tests.
In addition, there is a lot more effort going into developing cheap and effective deepfake GANs, than there is in developing deepfake detection systems. Individual users without knowledge of neural networks or GANs can now buy inexpensive software that can create high quality deepfakes. Besides, some people advertise and are willing to create deepfakes for others, for just a few dollars per video.
Another powerful disinformation tool that is just now being developed is AI generated fake news. Synthesized fake news, very credible to humans, can be generated by several AI based systems. As with deepfake videos, the danger is that AI generated deepfake news may make us simply stop believing what we see and read.
Again, as with deepfake videos, when there is a lot of data from which the AI fake news system can learn, the fake news generated by the systems becomes more credible and plausible to humans.
In February 2019, a San Francisco based organization called OpenAI launched an AI fake news system called GP-2. This software learns from the vast amount of available real news data, from millions of web pages, to generate fake news that reads like real news.
GP-2 works so well that most humans found fake news generated by it to be as credible as real news.
In June 2019, researchers at the University of Washington released an AI based fake news generator called Grover, that could publish as well as detect fake news by analyzing text patterns.
Ever more capable AI fake news generators are rapidly coming on-line, learning from larger and larger data-bases. As these get bigger, fake news created by them will only get more credible and become more proficient at fooling humans.
The 2020 elections will most likely see an explosion of deepfakes, with the first large-scale attempts at AI fake news, along with domestic disinformation and fact manipulation.
Fact manipulation in the west begins with ‘Freedom of the press’, a supposed freedom repeated so often by the western press that it is taken for granted by many consumers of western media products. Freedom of the press is just a well-marketed hoax that masks journalists’ all too numerous forays into fiction and disinformation.
Press freedom is a myth, a sanctimonious cover for media pandering to their commercial interests, and indulging in their political biases. Even news journalists are increasingly shrill activists and editorializers, not reporters of fact. There is no freedom of the press – which bows to the wishes of advertisers, owners and other corporate interests.
In the US, faith in the press seems to be split per party lines. According to polling by US based organizations, nearly 60% of Republicans feel journalists create a large amount of made-up and fake news, compared to 20% of Democrats who feel that way.
Republicans’ deep mistrust of journalists, from many decades of left-wing biased reporting in the mainstream media – where three times as many Republicans (compared to Democrats) feel journalists publish fake-news – is a damning indictment of the mainstream western media.
Many western journalists are left-wing sympathizers. Over the years, their reporting has become less fact and more leftist activist propaganda. This type of reporting reinforces democrats’ beliefs, so democrats are more likely to trust mainstream media reports.
There have been many studies done in the past three or four years on election meddling. Most of these studies focus on social media and the internet while analyzing fake news. The studies grant automatic and blanket credibility, veracity and integrity to established media outlets.
These studies, for the most part left-leaning or biased in favor of the US Democrats, conveniently place the mainstream media on a pedestal of truth, and use that reporting to categorize as fake news all accounts which contravene mainstream reporting. This when the mainstream media have been shown to spread disinformation over and over again.
Instead, the media have gone on a propagandized witch-hunt against Trump, knowing their democratic base would lap up all the falsehoods that were reported, increasing viewership and readers. For western media, disinformation is all about sensationalization, with corresponding increases in ratings and revenues.
For example, here are some major types of disinformation. Given the western mainstream media’s self-righteous, sanctimonious, and self-serving claims of reporting the ‘truth’ and of ‘being a free press’, examples of obvious and overt propaganda and disinformation are restricted to the mainstream western media:
Accusing an individual or organization of something you do – Hillary Clinton’s and the Democrats’ Russia collusion hoax is a well-publicized example of this type of disinformation. The media, with its left-wing bias, spread a false narrative accusing Donald Trump of colluding with foreigners – in this case the Russians. However, numerous enquiries showed no collusion between foreigners and Trump and his campaign. On the contrary, evidence surfaced revealing Ms. Clinton’s funding of and colluding with a foreign espionage agent from the UK, Christopher Steele, to create a false narrative on Trump.
Evidence also surfaced that Democrats and Hillary Clinton’s team colluded with the then authorities in Ukraine to obtain damaging information on one of the Trump campaign staff, from years before he joined the campaign.
Presenting a narrative in such an emotional way that facts lose meaning – The recent grandstanding on the environment, and the primal screams of an ill-informed Swedish teenager with serious mental issues, is a typical example of this propaganda technique.
Deepfake videos – Coming soon to a screen near you.
Ridicule – This is marginalizing people through mockery in order to discredit them, instead of sticking to facts. In the lead-up to the US 2016 presidential elections, western media played up Trump so he would become the Republican nominee, thinking he would be the weakest candidate among the Republican primary contenders, therefore expecting Hillary Clinton to beat him easily.
Once he was nominated as the candidate, the media switched to ridiculing him, to aid Ms. Clinton. In this case, however, things did not go according to plan, due to a number of problems with the Clinton campaign, not least of which was Clinton herself, and Trump was elected despite western media’s ridicule based propaganda.
Using key phrases and words to spread fake narratives – For example, consistently adding the phrase ‘with no proof’ or ‘unconfirmed reports’ when reporting details provided by Russia on the war in Syria, while consistently repeating verbatim all propaganda from the Syrian Observatory (based in London and far from the action), without mentioning that not only was there no proof for the Observatory’s claims, but that the Syrian Observatory was in fact a propaganda arm for the rebels.
Using terms such as ‘poisoning’ instead of ‘unexplained sickness’ to describe the Skripal case, or implying ‘murder’ instead of ‘suicide’ when reporting the death of the oligarch Berezovsky, are also typical examples of this form of disinformation.
Fiction masquerading as fact, dismissed as a ‘joke’ when caught – Adam Schiff, chairman of the Intelligence committee of the US congress, is facing censure for propagandizing a telephone call from President Trump to Ukrainian President Zelensky. Schiff read into congressional record a rambling, nonsensical passage of Schiff’s own creation, implying it was from President Trump’s phone call, pretending he was reading from an official transcript.
In his fake presentation, Schiff tried to make it appear as though Trump had pushed Zelensky over and over again to start official enquiries in Ukraine of Trump’s political opponents and also pretended that Trump threatened Zelensky.
When caught out, Schiff tried to dismiss his outrageous and unprofessional behavior as an attempt at ‘parody’, in the middle of a very serious impeachment effort by the Democrats.
Fake narratives without proof – There were myriad inconsistencies in the Dutch enquiry and media narratives on Malaysia Airlines MH17, starting with photographs of BUK SAMs taken from social media and presented without context, along with parroting of Ukrainian government press releases from Kiev masquerading as actual reporting from the war zone.
This was accompanied by condemnation of ‘Donbass rebels’ who purportedly shot the airliner down. This was then presented as proof of Russian guilt by association with the Donbass freedom fighters, and is a typical case of this type of propaganda – made up narratives without proof.
False facts – Here are a number of examples of western mainstream media stating false facts in order to sensationalize a false narrative, reporting falsehoods that are nevertheless interesting to their readers, to further politics that the media espouses, and to improve sales (yellow journalism at its worst):
Washington Post. The Democrats claim that a member of the CIA was shocked by the content of a phone call between President Trump and Ukrainian President Zelensky and decided to file a ‘whistleblower’ complaint with his superiors. The acting Director of National Intelligence, retired Vice-Admiral and ex-Navy SEAL Joseph Maguire was asked to testify before congress about the matter.
On September 25, 2019, the Washington Post published a sensational, made-up, fake story that DNI Maguire was under pressure from the Trump White House to keep the facts of the case from congress, and had threatened to resign. When DNI Maguire emphatically said that was simply not true (adding words to the effect that he had not quit anything in his life and was not about to start now!!), the Washington Post later quietly changed their reporting to reflect DNI Harris’ denial, while loudly proclaiming that they stood by their story.
Washington Post. On December 30, 2016, the Post published an entirely fake story with a spectacular headline – ‘Russian operation hacked a Vermont utility, showing risk to US electrical grid security, officials say’. The Telegraph, of the UK, followed-on with: ’Systemic, relentless, predatory’ Russian cyber threat to US power grid exposed, as malware found on major electricity company computer.
Panicked US politicians began threatening Russia with a serious response.
However, the Washington Post kept changing their story in subsequent reports, until finally being forced to admit the story was inaccurate and that the ‘malware’ found was on a laptop not connected to the US electric grid, had nothing to do with the electric grid, and had no connection to Russia.
Fortune. On June 12, 2017, Fortune published a fabricated, false article that RT had hacked and taken over C-SPAN, a channel that televises the affairs of the U.S. Congress. Fortune even claimed C-SPAN “confirmed” it had been hacked by Russia. Based on their technical analysis, C-SPAN reported that it had NOT been hacked and that they had suffered an internal routing error in their network.
CNN. On June 22, 2017, CNN faked a news story that Trump aide Anthony Scaramucci was involved in the Russian Direct Investment Fund, which was being investigated by the US Senate. Three journalists at CNN were fired after the subsequent scandal, when the report turned out to be completely fabricated.
New York Times. On August 7, 2017, The New York Times published a fake news report falsely claiming that the Trump administration had suppressed and hidden a government climate report. In fact, the study had been published seven months earlier by the Trump administration, and had even been reported on by the Washington Post.
NBC. On Sept 11, 2017, NBC News confidently spread propagandized misinformation that Russia was behind numerous attacks on US embassy staffers in Cuba, causing brain injuries in some cases, using a Russian super-secret sonic weapon. Biologists found that the strange sounds heard by US diplomats were emitted by an insect during mating season. Neurologists found no brain injuries among US embassy staff, instead determining the diplomats suffered a collective psychosomatic hallucination.
CNN / MSNBC. On December 9, 2017, CNN falsely reported that Donald Trump Jr. had had advanced access to Democrats’ emails published by WikiLeaks, before those emails were made public. MSNBC chimed in an hour later, claiming to have independently confirmed this fake news sold by CNN and MSNBC as a bombshell news scoop of collusion between Trump and WikiLeaks. CNN and MSNBC are yet to apologize for this debacle.
While many of these diverse examples are from 2017, most of 2018 was about fabricated stories and supposed leaks from the Mueller investigation, along with false narratives on Russian interference in the 2018 election. Mueller’s 2019 report on the Russia collusion hoax showed there was nothing there, and that it was a fabrication by the Democrats and their allies in the mainstream media. Since then, 2019 has been about Ukraine, impeachment, and fake news on phone calls, transcripts and whistle blowers.
And so it goes, on and on.
American Democrats and Republicans have become extremely polarized and will not budge from their views. Pity the poor Independents, usually the deciding group in US elections, who by year-end 2020 will surely have their brains turned to mush by the inevitable cacophony of election-year claims and counter-claims – truth, fake, and everything in-between.
Let the games begin.