Deepfakes – Bringing more fake to fake news

With the increase in accessibility to Artificial Intelligence and Big Data tools, sophisticated audio, video and picture manipulation becomes a widespread phenomenon. Should you be worried, and how can you counter it?

Lenin’s speech – 1 May 1920

Picture, video, and audio manipulation are not a new phenomenon and have been around even before Photoshop and such came along. One of many examples of such early manipulations can be found around the 1920s, when Stalin rose to power. Back then, the politicians Leon Trotsky and Lev Kamenev, associates of Lenin and rivals of Stalin, were removed from a picture where they were shown attending a speech of Lenin.

 

 

What is new about deepfake is that it uses Artificial Intelligence to automate the process of producing fakes with increasingly better results. Video manipulation for example can be done relatively fast, compared to the old process of editing every frame by hand. Grafting a face onto a different person isn’t that hard anymore, as the video below demonstrates.

In addition, applications like Overdub by Descript let you create realistic profiles of voices, which can be used to “overdub” a video with new spoken audio. Combining new video and audio editing methods allows for creating fake speeches, like the one that the Massachusetts Institute of Technology created of Nixon, where he’s giving a speech he would have given, in the event of a failure of the Apollo 11 mission.

Fake news on steroids and how to counter them

It is already hard to discern fake from real news, requiring you to vet news articles, tweets and any news source you consume. With the ability to create realistic fake audio and video material, deepfakes add a new layer that requires even more thorough vetting by the reader, but also more direct action by the media platforms.

Various social media platforms like YouTube, Twitter and Facebook have announced their intentions to fight back against deepfakes with changes to their terms of services and by employing AI-driven countermeasures of their own. To that end, Facebook, in partnership with Microsoft and other companies, held the Deepfake Detection Challenge (DFDC) last year. The winning data models are being used to create AI-driven countermeasures, which should be deployed soon.
These countermeasures don’t mean that you can simply relax and consume your news. This is a constant arms race between fake news and those that try to counter it. As mentioned earlier, vetting your media sources and fact checking news is critical. Thankfully there are some resources and methods to help with this.

Harvard, for example, has four easy steps that help you with spotting fake news:

  1. Vet the publisher credibility.
    Would the publishing site meet academic citation standards?
    What is the domain name?
    What’s the publication’s point of view?
    Who is the author?
  2. Pay attention to quality and timeline.
    Are there spelling errors?
    Is the story current or recycled?
  3. Check the sources and citations.
    How did you find the article?
    Who is (or is not) quoted, and what do they say?
    Is the information available on other sites?
    Can you perform reverse searches for sources and images?
  4. Ask the Pros.
    Have you visited a fact-checking website?

Additionally, I’ve collected a few resources which will help you with step four.

  • Snopes is an excellent fact-checking site that fact checks all the latest news and rumors that crop up on the internet.
  • MediaWise is a great project with tips to enhance your media literacy (the ability to access, critically evaluate and create/manipulate media).
  • The Skeptics’ Guid to the Universe – A generally good book about critical thinking, which also takes a deep dive into logical fallacies (patterns of reasoning that are rendered invalid by a flaw in their logical structure) and how to spot/avoid them.

Robotic journalism

The widespread availability of Artificial Intelligence and Big Data tools also gave rise to a change in the realm of journalism. Many news outlets like Forbes or Reuters, are starting to use AI-driven tools to either support journalists in early drafts of their articles or to fully automate the writing of articles. The main goal, according to most companies, is to raise the quality of their articles and to produce more articles in general, but also to personalise news articles for each reader. Certainly these advances in technology will help reporters with their work, but should never replace a human reporter fully. The risk of accidentally creating fake news, if no screening is applied before publication is too great and publications like that could further weaken the standing of traditional media in the public.

Further tailoring news articles to the reader, as in changing the wording to fit, what the algorithm deems fitting for the reader, is a dangerous game. Such changes could further enclose people into an “an environment in which somebody encounters only opinions and beliefs similar to their own, and does not have to consider alternatives” also know as an echo chamber.

The shape of things to come

Even with all the measures legally and technologically employed to counter them, deepfakes are here, and they are here to stay. In the long run, we will need to learn to live with them and harden our legal, as well as our educational systems to limit the effects of their nefarious uses. Video, image and audio evidence needs to be scrutinised even more thoroughly before being admitted in a court case, and our schools need to teach kids about media literacy, critical thinking, and ethics, as these issues will always be with us.

In regards to short term solutions, there are multiple projects that aim to deal with the problem of deepfake and fake news. Microsoft for example has announced new steps to combat disinformation. Swiss scientists are also working on methods to spot deepfakes and stop the perpetrators.

So, should you be worried regarding deepfakes and fake news? Certainly. But deepfakes and fake news shouldn’t dissuade you from seeking out news and as outlined in this post, there are good methods to mitigate the effects deep fakes can have on our personal lives.

Beitrag teilen

Andreas Keiser

Andreas Keiser ist Business-Systems Coordinator bei der V-ZUG AG und bloggt aus dem Unterricht des CAS Big Data Analytics.

Alle Beiträge ansehen von Andreas Keiser →

Schreibe einen Kommentar