News agency Reuters says it has worked with UK-based “synthetic media” company Synthesia to create a convincingly realistic, AI-generated sports presenter.
The deepfake-style programme is “intended purely as a proof of concept” Reuters said — although it said it may also create “products and services” around it in future.
The system uses an algorithm to combine real-time match photography with a minute-by-minute data feed of live reporting on matches. Combining these two databases Reuters can automatically create a presenter’s script for any match it is covering. Once the script is manufactured it is read out by a computer generated presenter.
The software was created by Synthesia: an AI and visual content creation firm based in London. The company has worked with many organisations to create manipulated media content, such as a series of adverts for the Malaria No More campaign that features David Beckham “speaking” in nine different languages.
Synthesia’s system essentially builds a digital copy of the presenter and then, using generative adversarial neural network models, the system is trained to accurately simulate new content of the video presenter using the sports data feed.
Reuters claims that the: “The prototype is intended purely as a proof of concept, designed to show the potential for future real-time news and information services presented using these new kinds of AI based presentation systems.”
The National Union of Journalists (NUJ) sounded a cautious note, telling Computer Business Review: “It is fascinating to see technology has progressed to the extent that robots can be TV presenters, but the personal touch and professional approach of a real journalist can’t be replaced. The technology should be used to enhance journalistic practices but should not be used as an opportunity to cut jobs.”
This is not the first computer generated TV presenter to be unveiled in recent years: China’s state news agency Xinhua rolled out an AI anchor in 2018. The news anchor Qiu Hao was created by the by Chinese search engine firm Sogou.
Deepfakes or synthetic media, is no longer an emerging technology: something evident from the slew of deepfake videos that are being produced for propaganda purposes.
Social media firms have begun to respond to the rise of deepfakes in various ways, this month Twitter instituting a new policy: if a Tweet is found to contain content that has been “significantly and deceptively” altered the site will label it and reduce its visibility. (The extent to which Twitter is capable of doing this efficiently remains an open question.)
Credit: TwitterAt the start of the year Facebook rolled out a similar policy of declining to actively delete deepfake content and stating that: “If we simply removed all manipulated videos flagged by fact-checkers as false, the videos would still be available elsewhere on the internet or social media ecosystem.
It added: “By leaving them up and labelling them as false, we’re providing people with important information and context.”
Synthesia said: “Synthetic media will significantly accelerate creative expression and lessen the gap between idea and content. It will bring with it new methods of communication and storytelling, enable unprecedented human-computer interfaces and challenge our perception of where the digital realm begins and ends.”
The potential for businesses to add dynamic and media-rich content to their digital platforms at low cost clearly will prove enticing and the opportunities manifold as the technology continues to evolve and costs of compute fall.