AI-driven synthetic media: Too real for comfort?

You would be forgiven for thinking the model pictured above, who recently went viral on social media, was as real as any of us. But in fact, all is not as it seems—she’s a virtual human, one that was created using the latest AI-driven synthetic media technologies that can achieve photorealistic renditions of human-likenesses, moving or still.

Brought to life by Hong Kong based software firm Pantheon Lab, a startup specialising in AI-driven synthetic media—which is the manipulation of digital assets (photos, videos, sounds and texts) by automated means using artificial intelligence algorithms. For this project, Pantheon Lab used face-shifting technology to create her uncanny human-likeness.

Until now, it’s been easy to tell that virtual influencers are digital creations and not real people, but some might feel this next level of photorealistic ‘virtual human’ is a little too close to the bone. Clearly, modern technology allows us to create evermore convincing virtual humans, but the question is, should we?

Ivan Lau, Pantheon Lab's CEO, believes it's only a matter of time before virtual human synthesis becomes mainstream. "Give it half a year and virtual humans are going to be everywhere, in my opinion," says Lau. "People are catching up to these things. Gen Z, in particular, are very pro virtual world anyway, and if they are becoming older and they're turning into the major voice and community, I think it's just the beginning for virtual humans.”

AI for good?

We've seen reports in the past of deepfakes being used by bad actors to trick or misinform viewers, but Lau and other proponents of synthetic media are adamant that their ground-breaking technology will be utilised for good purposes only.

"Our mission is to use AI for good. We are not looking to trick or mislead people or to create influencers just for the sake of showcasing what’s possible," says Keif Se-To, a software project manager at Pantheon Lab. "We are more interested in having real purpose. We are chatting with some universities as well as some medical companies and are trying to make use of our technology to do something good."

As an example of a 'good' use case of the technology, Pantheon Lab point to a recent enquiry they had from a school in India. “The school explained they didn't have a lot of teachers who are willing to travel to the rural side to teach young children," says Lau. "So we are working on helping to create virtual teachers who can speak in the local language to give an online zoom course."

Currently, Pantheon Lab offers products like its AIDOL Studio that allows brands and corporates to create their own virtual ambassadors and produce marketing and internal training videos without the need of a physical set, production crew, or even actors/anchors. The technology enables studio-grade presenter-led video automation with deep learning to cut production time from days down to minutes.

"Anyone involved in production will know it takes time to find the right presenter and to do the shooting with a crew," says Se-To. "We would like to shorten the time it takes and reduce the human input in that. So our AIDOL Studio platform is designed to automate the process and save time and money."

Unstoppable rise

Along with the development and broader acceptance of virtual worlds through the metaverse and Web3, it seems the rise of synthetic media and evermore realistic virtual humans is inevitable. But will brands and consumers be willing to utilise and really engage with virtual humans once they discover they're fake?

"No doubt it’s a safe bet for brands to partner with talent who have, quite literally, no human flaws—but it’s a partnership that won’t resonate with consumers," says Hannah Mahony, social and influence director at VCCP. "We’re currently in a state of influencer reality right now where transparency and relatability have almost become their own social currency."

But Jasmin Hyde, senior account manager at Icon Reputation, says that comfortable with it or not, behemoth brands such as Red Bull, Android, American Express and Cadbury have already proved it doesn’t take a human to sell, by creating some of the most memorable animated ads of all time.

"Whether it be animation, AI-generated computer-generated imagery (CGI) or cartoon, from a business perspective, there are several reasons why advertising without using human talent is beneficial. There is less expenditure on labour, generally a huge saving on production costs, and less time delays," says Hyde. "Virtual influencers also enable brands to have more control over messaging and the final output and can create ambassadors that ‘perfectly’ align with their brand."

But, Hyde adds, real influencers needn't lose any sleep at night as virtual influencers will not replace their human counterparts, but instead coexist.

Ethical dilemmas?

So far, India is the only country to address virtual influencers in national advertising standards, requiring brands “disclose to consumers that they are not interacting with a real human being” when posting sponsored content. As AI technology enables virtual influencers to become harder to distinguish from their human counterparts, should there be more widespread regulation and increased transparency around who is responsible for the content and whose and which moral values are being espoused?

"Virtual influencers should be held to the same standards as physical influencers," says Chris Gurney, group creative director, Virtue APAC. "Beyond the regulations around transparency and disclosure, virtual influencers should also be held to moral clauses to ensure that they behave in responsible and ethical ways befitting that of public figures. In addition, these moral clauses should work two ways for both brands and influencers to ensure a fair and equal agreement."

Virtual humans created by Nikkei to help with news reporting and online event hosting.

Hyde of Icon Agency believes that, as with deepfake technology, virtual influencer marketing will pose the most risk when combined with other technologies and social trends. "There is risk of increased cyberattacks, accelerated spread of propaganda and disinformation, and decreased trust in democratic institutions."

Noticing the steady rise of virtual influencers, Meta, the parent company of Facebook, plans to produce an ethical framework to guide the use of VIs, and has warned that “synthetic media has the potential for both good and harm”.

In addition to concerns around transparency and need for more regulation, there are ethical considerations as well. "These issues centre on the sensitivities of creating synthetic humans with different demographic characteristics from the humans creating them," says Scott Guthrie, head of the Influencer Marketing Trade Body (IMTB). "There have been instances of racialisation, including digital blackface, and cultural appropriation. There, too, are considerations of creating synthetic humans with unattainable beauty standards raising issues around body image."

Mahony of VCCP says it’s up to creators to divert their AI influencers away from impossible beauty standards, instead normalising all body types and beauty preferences. "If anything, there’s an opportunity to create accurate representations of a generation, ones that can grow and adapt in real time in-line with data gathered about that group of people,” she said.

Equally, Chris Gurney, group creative director, Virtue APAC, is optimistic that the virtual influencers that will take off with this generation are the ones who will challenge impossible beauty standards or any standards for that matter.

"They will provide the catalyst for us to imagine the different possibilities of who we can be. We already see this happening with the younger generation—their powerful imaginations playing out on Roblox, and the fascinating avatars they have created to represent themselves in virtual worlds," says Gurney. 

Hyde adds: "We won’t be seeing a hostile AI takeover any time soon. As with animation over the years, some brands and industries will make the leap to adopt virtual influencers, while others will ‘keep it old school’ with their human ambassadors.”

In fact, Hyde says, emerging CGI advancements pave the way for human influencers to create their own life-like avatar to sell to brands, which is an interesting revenue model in itself. She says: "So, brands might not get access to the human talent themselves, but only their CGI version, and the human then profits from that."