NAB
Send in the Clones: The rise of Synthetic Media
From virtual celebrities and MetaHumans to DeepFakes and voice clones, novel forms of synthetic media blur the distinction between physical and digital environments and will radically accelerate the process content creation and delivery. With it come important questions about privacy and ethical dilemmas.
article here
An article by future tech strategist and entrepreneur Mark
van Rijmenam runs through where synthetic media is today, where it might be
going and what pitfalls we must be aware of.
“We are entering a new age where more people will be exposed to synthetic media. It’s a mass social experiment, and we have no idea what the consequences of this medium might be. If we cannot predict or study its impact accurately, there is little hope of protecting ourselves against its dangers.”
But what is synthetic media? The definition given here is
virtual media produced with the help of artificial intelligence (AI). It is
characterized by a high degree of realism and immersiveness. Furthermore,
synthetic media tends to be indistinguishable from other real-world media,
making it very difficult for the user to tell apart from its artificial nature.
It is possible to generate faces and places that don’t exist and even create a
digital voice avatar that mimics human speech.
Examples that exist right now include virtual celebrities
like Lil Miquela, a persona that does not exist except online but who has
become one of the world’s most popular virtual influencers on Instagram, with 3
million followers.
“She (it?) and other virtual influencers are becoming
increasingly popular and will continue to do so,” says van Rijmenam.
The ability for anyone to create their own avatar to popular
the metaverse is being driven by Epic Games which has a program, called
MetaHuman, that does just that.
“MetaHuman Creator enables you to create fully rigged
photorealistic digital humans in minutes, in real-time, for use in video games,
virtual and augmented reality content, architectural visualizations, and more.”
Van Rijmenam himself says he is now ‘transitioning’ to a
MetaHuman character; a digital twin of himself.
That’s an interesting choice of word given today’s often
controversial gender IDs.
On the audio side, artificial voice technology, such as
text-to-speech and voice cloning, has become very popular. Companies here
include Resemble.ai that allows you to clone your voice to create digital
avatars and use them in movies; Respeecher and Voiseed. As described by van
Rijmenam, Voiseed makes audio content more human by creating a voice interface
that communicates in authentic, natural language using emotion and intellect.
Synthetic media tools allow for creating complex data
visualizations, or even videos, using only a spreadsheet. Analysts and
researchers often use these to present findings to a broader audience. Art
directors also use it to mock up ideas before they bring them to life in
development.
Synthetic music has the potential to generate sounds that
are indistinguishable from a human-produced track.
“AI generates ever-shifting soundscapes for relaxation and
focus, powers recommendation systems in streaming services, facilitates audio
mixing and mastering, and creates rights-free music.”
You can find AI-generated and copyright-free music using
platforms like Icons8 and Evoke Music. Synthetic images are already being used
for creating NFT art and generating realistic stock photos, while synthetic
Videos combine the worlds of photography and videography. They have taken on
many forms, but one of the most popular types is deepfakes. These are Face
Swaps, where one person’s face replaces another’s or face reenactment in which
the source actor controls the face of the target actor.
GANs
This is all made possible by advances in neural networks -
or Generative Adversarial Networks (GANs), to use the jargon. Since GAN outputs
look natural and indistinguishable from the original photos, they enable
synthetic media that is difficult to distinguish from real media, particularly
in computer vision and image processing applications.
“At the same time, advances in machine learning and deep
learning have made it possible to train computer vision algorithms on large
datasets of images,” he says. “As a result, today’s neural networks can see
things in photos that humans can’t even detect with our own eyes.”
What is synthetic media good for?
As with most AI technology there are pluses and minus. On
the one hand the sheer volume of content required to build the worlds of the
Metaverse is going to beyond all traditional computer assisted graphics
construction. AI-driven media can be created rapidly. One only has to look at
the impact of the fledgling text to picture AI Dalle-2 to realize the benefits
of this approach is going to have in kickstarting artistic creation in the 3D
internet.
Such technology is only going to get advance.
Text-to-picture is evolving into text-to-video and eventually will lead to the
creation of full scale feature films. There have been several experiments in
creating long form AI-composed content but what distinguishes the next
generation of synthetic media is that it will be indistinguishable from the
real thing.
“It can create an illusion of authenticity. This media type
allows companies to connect with their audiences without paying actors or
hiring professional photographers or videographers,” says van Rijmenam.
If certain jobs in media production are going to be automated
out of existence a new type of job will emerge to take its place.
“The primary focus is interacting with AI to help it become
more intelligent and capable,” says the futurist. “The skill of those who work
with AI is important. If employees do not stay updated on technological
advancements and improve their knowledge, they could be forced out of their
jobs — no matter how hard they try to avoid automation.”
Deepfakes and the law
As AI becomes ever more sophisticated, so do the ethical
challenges AI faces. The biggest concern is ensuring that the algorithm will
not engage in abusive or unethical practices towards humans and vice versa.
Text and video can be created to generate misleading, false, or non-existent
information, (fake news).
“When it comes to deepfake issues, journalism cannot escape
the fact that its old forms of reporting are under pressure due to the rise of
digital information. Therefore, we need media literacy and verification to
report on these deepfake videos and regarding worldwide disinformation or
propaganda.”
Personal rights and intellectual property laws will also be
challenged. The legality of AI-generated counterfeit content is often unclear,
making it difficult to know where your rights lie.
“Copyright law protects original intellectual property from
copying; however, in an era of exponential growth, it will soon be unable to
distinguish between ‘real’ and ‘fake’ text.”
Who should own the rights to a synthetic movie where all the
actors are created digitally? The studio or the creators of the algorithm that
generated the characters? These questions are still to be explored and framed
in law.
“Synthetic media will need to be regulated by law and
policy, so we’ll need new rules to determine ownership and licensing,” says van
Rijmenam.
No comments:
Post a Comment