The invention of AI-Tech raises red flags in all fields, including journalism, and synthetic media is making matters worse. This guide centers on the definition of synthetic media, techniques, and most shockingly, red flags to Journalism.
Why it matters:
The various forms of data that make-up news content are on the verge of duplicity, as synthetic media –an algorithm that can manipulate texts, images, and audiovisuals – is currently available to those who seek it.
With this AI-based model, ‘it’s possible to create ‘faces and places that don’t exist and even create a digital voice avatar that mimics human speech.’(Aldana Vales 2019)
Imagine a world, where it’s quite difficult to differentiate between fake and real news, since the disseminators of fake news can alter ‘evidence’ to suit their agenda. For instance; no one would cease to believe that World War III has begun, if videos of Trump, Putin, and Kim declaring war were globally circulated online. Though such news might be debunked by the governments involved, the psychological and economic panic it would cause might be greater than the effect of a missile.
Synthetic media can be created via three forms of generative artificial intelligence, namely; Generative Adversarial Networks (GAN), Variational Autoencoders, and Recurrent Neural Networks. These aforementioned G.A.I’s are used for photo, video, and text generation respectively. The word generation is used because most of the media contents created with these algorithms don’t exist; however, Synthetic media can also be used for duplication.
According to Aldana Vales, ‘Generative Adversarial Networks use two neural networks (a neural network is a computing system that can predict and model complex relationships and patterns) that compete against each other.’
The first and second networks act as a generator and a discriminator individually. The discriminator supervises the generator, ensuring no stone is left unturned. After some ‘to and fro’ revisions by the duo, the content produced would resemble the original.
Unlike Generative Adversarial Networks, neural networks in Variational Autoencoders are called encoder and decoder, since the technique involves compression and reconstruction of video content. ‘The decoder includes probability modeling that identifies likely differences between the two so it can reconstruct elements that would otherwise get lost through the encoding-decoding process.’ (Aldana Vales 2019)
Recurrent neural networks function by ‘recognizing the structure on a large set of text’. This is the method used in text autocorrect phone apk.
These techniques are applied in various projects such as; GauGAN, Face2Face, and GPT-2 model. The most recent application of Synthetic media can be found in Siri or Alexa. These virtual assistants now have the ability to ‘turn text into audio and mimic human speech’.
In a 2017 article, titled ‘AI-Assisted porn is here and we’re all fucked’, Vice exposed the circulation of a fake porn video, which isn’t a problem because most plots portrayed in porn movies are fake (LoL); except that the actor had the face of a popular non-pornographic actress, Gal Gadot (Wonder woman). Also, in 2018, ‘a video showing President Barack Obama talking about the risks of manipulated videos’ was circulated on Buzzfeed. The weird thing about this video is that the AI-Generated subject has Obama face and Jordan Peele’s voice, thanks to Synthetic media.
There’s an ongoing campaign against the potential harm of Synthetic media on news authenticity; however, ‘Beyond reporting…newsrooms are focusing on synthetic media detection and validating information. The Wall Street Journal, for instance, created a newsroom guide and committee to detect deepfakes. The New York Times recently announced that is exploring a blockchain-based system to fight misinformation online.’ (Aldana Vales 2019)
Synthetic media could help news agencies break language barrier seamlessly. Also, it could encourage the circulation of fake news. While It is impossible to stop giant tech companies from diving into AI-Tech research, journalists can learn how to control the damage posed by synthetic media.