Ka hao te Rakatahi The dodgy world of deepfakes

Dec 22, 2019

Nā Nuku Tau

With 2019 drawing to a close, I thought I should set my sights on 2020. Why? Because two important events in the political world will take place next November, as both New Zealand and the United States go to the polls to elect their new political leaders. And as technology evolves at a pace many of us can barely comprehend, modern elections bring modern problems. I’m talking about “deepfakes”.

A big step up from simply “fake news”, a deepfake is a particular kind of manipulative media in the form of doctored videos. Deepfakes present real challenges to democracy and our electoral process, especially at the hands of those with the money to drive influence.

The term “deepfake” is a combination of the words “fake” and “deep”, as in “deep learning”, an Artificial Intelligence function. Using deep learning, a machine can analyse thousands of images and recordings, and use this to convincingly transpose one face over another in a recording. This creates increasingly hard-to-spot videos of deception. Increasingly, believable-looking videos are depicting people doing and saying things they have never said or done. It’s sometimes described as “Photoshop on steroids”.

Deepfake videos can portray utterly fabricated events as real. For example, there is a deepfake on the internet of US President John F. Kennedy in which he gives a speech that reality never witnessed. It was supposedly given in 1963, on the day he was assassinated. In juxtaposition with this, another particularly humourous example is the one of President Obama by comedian Jordan Peele in which Obama is depicted as saying black Republican Ben Carson is in the “sunken place”– a joke at Carson’s expense implying he is silent to black oppression. Other G-rated examples include clips of Nicholas Cage in movies he never starred in.

There are also resurrections of dead celebrities in films. There are endless (somewhat creepy) possibilities in the realms of film or media available to us.

Deepfakes can be used for a range of purposes, from political to commercial and recreational, and are potentially very dangerous. The video of President Obama, although humorous, could have had negative political effects if the spoof had not been admitted. On a much larger scale, there is potential for deepfake videos to display fake declarations of war by government officials. A New York University study identified deepfake videos as one of eight potential mis-information forms that may sway the 2020 US presidential elections.

In disgusting but relatively unsurprising news, the videos have already been used in pornographic and other nasty ways to discredit female activists and journalists. Sam Gregory, of the human rights non-profit group Witness, claims that up to 95 per cent of current deepfake images and videos on the internet are non-consensual pornographic images and videos, mostly containing celebrity women.

Versions of the technology are slowly making their way into the mainstream through apps and websites. As this technology becomes more accessible, it creates further risk of misuse.

Currently, there is no silver bullet for this growing problem. Readers might remember how annoying internet spam used to be in emails, or when entering dodgy websites. Now, due to filters and browser add-ons, this is largely a thing of the past. However, the speed at which deepfake technology is developing means no such cover-all solution yet exists.

At this stage, the creators of deepfakes have the advantage. Sam Gregory notes that around a year ago people believed that deepfake videos could not depict a person blinking, and that a way to recognise fake media was to spot this. Of course once this was publicised, deepfakes that blink suddenly appeared. Solutions are being worked on, but due to hackers and the general nature of the problem, these can’t be relied on as yet.

So, what is the relevance for Māori? As I have noted in previous columns, the media loves to run with the juiciest stories; usually at the expense of Māori – especially when the story is a negative one. An example previously discussed is the fake picture from a self-proclaimed scientist with no credible backing, who claimed the first New Zealand settlers were Welsh. The media ran with this picture and also asserted that the picture had come from the University of Edinburgh. When this was all proven false – silence. There is obvious potential for doctored
videos or images to create uproar at the hands of slack journalism.

So, what is the solution? Gregory notes that most of what is shared online is “true or true enough”, and that the solution is not to replace standard internet caution with a disregard for everything one may see on a screen. Even if there was some way to instantly identify deepfake media, a stamp over some video or image marking it as legitimate could still be easily misconstrued by context.

The approach suggested by “techxperts” in the field is a fairly commonsense one. We should be critical of internet content whenever it appears suspicious. Look for sources and corroborating evidence. These days we seem to have an inherent skepticism of many internet images anyway, and perhaps we now simply have to attempt to include video footage as part of our list of things to watch out for.

We are not in an age that marks an “end of truth”; but rather, in a time where information is more readily accessible than ever. We all just need to remember our media literacy skills.


Radio New Zealand, “3 Minutes Max”. 27 October 2019:

Grace Shao, CNBC. 13 October 2019. “What deepfakes are and how they may be dangerous’.

Tom Simonite, WIRED. 10 June 2019. “Prepare for the Deepfake Era of Web Video”.

Twenty-year-old Nuku Tau (Ngāi Tahu, Ngāi Tūāhuriri) has just completed his second year of a law degree at the University of Canterbury.