We’re a group of creative technologists who understand that the development and proliferation of synthetic media must proceed within a framework of community-led stewardship. We are committed to advancing the technology and its creative potential by safeguarding the rights of individuals, promoting a healthy and sustainable internet, and taking time to critically examine the evolving ethical implications of synthetic media.

We love to create.

Our team is made up of some of the world’s leading AI artists and synthetic media creators. We love to create fun content and aim to delight audiences around the world with digital experiences that offer a glimpse of a hyperreal future. We are our own best customers and beta testers for our products. But we also believe it is imperative to create content that is used to educate viewers and raise awareness about the underlying technologies in order to disrupt the negative impact of unethical uses of synthetic media.

Earlier this year, we released @deeptomcruise, a series of parody videos on TikTok that quickly garnered more than 100 million views and coverage from hundreds of media outlets around the world. In the videos, a near perfect likeness of Cruise chews gum, plays golf, hides in a turtleneck, and performs magic tricks. These activities are mundane, but the intense global interest from viewers is a natural reaction to both the immense celebrity of the real Tom Cruise and the technical marvel of a hyperreal synthetic Tom Cruise.

We believe that @deeptomcruise is a milestone for synthetic media that marks the first time that an AI-generated video has presented a synthetic reality that is indistinguishable from actual reality for millions of people around the world. We hope that @deeptomcruise has a positive impact on social discourse around synthetic media and deepfakes. It is heartening that in the ~150k user comments on our TikTok videos we see a large number of users sending love to what they believe is the real Tom Cruise and an equal number of users dutifully informing them that it is actually a synthetic version of Tom Cruise.

In addition to creating our own content, we’re building a platform that will make synthetic media production more accessible to brands, celebrities, and creators. Today, the tools used to produce hyperrealistic content require a thorough understanding of image processing, deep learning implementations, and VFX techniques. The technology is in its infancy and there are currently only a handful of people who can create hyperrealistic content of the caliber of @deeptomcruise. But things are moving fast!

Synthetic media > deepfakes

The media landscape is rapidly changing. Over the past few years, there’s been a foundational shift toward the use of artificial intelligence to assist in the creation of content. This process lowers the cost of creating many forms of content and, in the future, practically anyone will be able to create rich, customized virtual experiences featuring hyperreal content and advanced special effects.

We think of synthetic media as the catchall term for content created with assistance from AI tools. While nomenclature and definitions are still evolving, ‘synthetic media’ is a good lens through which to think about the burden of labour in creating content. On one hand, we have traditional media that is the product of human creators and their labour, even though it is often created with technical tools. On the other hand, synthetic media relies on complex algorithms trained on reams of data to create at least some elements of the content without human intervention. It’s a rapidly growing field populated by a diverse group of amateur and professional creators and the content ranges from virtual avatars to photorealistic objects in video games.

The term ‘deepfake’ was coined by an anonymous Redditor to describe pornographic videos featuring synthetic versions of celebrities in 2017. These crude videos were created using generative adversarial networks (GANs), a machine learning technique that pits two algorithms against each other in an iterative contest where one algorithm attempts to generate fake content that will trick the other algorithm into thinking it’s real. In the case of these early deepfakes, the creators were using these algorithms to map a celebrity’s face onto a porn performer’s body without consent from either party. Reddit soon banned these types of videos, but not before various developers released their source code and techniques into the world.

As a result of these deepfakes, most people are familiar with the darker side of synthetic media, and unfairly generalize some of the negative attributes of malicious deepfakes to synthetic media as a whole. But malicious deepfakes are only a small subset of synthetic media that are defined by their intention to deceive or cause harm to the people they feature. In short, while malicious deepfakes are synthetic media, most synthetic media is not intended to cause harm.

Community

The @deeptomcruise videos were created by Chris Umé, co-founder of Metaphysic and one of the world’s leading synthetic media creators and AI artists. Chris has a VFX background and has spent years working with AI models in order to produce hyperrealistic content like @deeptomcruise. But he couldn’t have done it on his own. Many of the techniques and tools used by Chris are developed by a community of academics, developers and artists who recognize the vast creative potential latent in code. In many ways, @deeptomcruise stands on the shoulders of giants, and we are very grateful to the community for that.

A few years ago, Chris and a small community of researchers, technologists, and artists quickly recognized that the technology underlying deepfakes had a much greater potential than non-consensual deepfake pornography. They began experimenting with the technology to create parodies and educational material meant to raise awareness about synthetic media. In 2018, Chris joined a private Discord server with a handful of other leading synthetic media creators. The group swapped techniques for improving their craft while committing themselves to ethical applications of the technology. This was bottom-up, community-driven coordination around a set of informal ethical standards and, today, the same group of creators are at the heart of the community and the leading exponents of their craft.

As the technology improved and the community of creators grew, we saw an opportunity to create an organization that both propels the technology forward and advocates for its ethical development. And so Metaphysic was born. In the future, we hope that our community can move from ethical aspirations to more formal codes of conduct and standards that are safeguarded by responsible industry leaders.

Metaphysic builds software to help creators make incredible content with the help of artificial intelligence. Find out more: www.metaphysic.ai

For more info please contact info@metaphysic.ai or press@metaphysic.ai

About the author: in a galaxy far away, I was a lawyer turned internet/society researcher. In the 7 years before co-founding Metaphysic, I built tech companies in SF and London. I have always been obsessed with computational photography and computer vision, so it a thrill to work alongside amazing people on the next evolution in how we build and perceive reality — one pixel at a time.

© Thomas Graham & Metaphysic Limited 2021.