AI Can Now Create Artificial People - What Does That Mean Humans?

• Read original article here

DataGrid, Inc. announced it successfully developed an AI system capable of generating high-quality photorealistic Japanese faces, it was impressive. But now the company has gone even further. Its artificial intelligence (AI) system can now create not only faces and hair from a variety of ethnicities, but bodies that can move and wear any outfit. While these images are fictitious, they are incredibly photorealistic.

Using GAN to Help AI’s Creativity

The Kyoto-based startup focuses on creative applications of artificial intelligence. DataGrid usesgenerative adversarial networks (GANs), so the AI can "learn" from a database of existing images and then from there it can generate its own versions.

GANs are one of the latest developments in the ever-evolving field of artificial intelligence. GANs were first proposed in 2014 and essentially pair two AI systems together that battle it out where one system creates while the other critiques the outcome, learn from the experience and adjust to improve the quality of results. GANs can create "new" information from following the existing rules and are a very exciting development.

While DataGrid's artificial faces weren't the first to be generated, with the help of GANs the "humans" they created are the most believable, yet plus the bodies have the capability of movement. The company's website explained that this involved two lines of research and development, including "whole body generation" and "motion generation." In the past, GAN-generated images would often have tell-tale flaws that allowed you to see they are were artificial such as asymmetry in the eyes or ears and even backgrounds blending into the faces. DataGrid minimizes these flaws because they use a non-descript white background and realistic light shining down on the computer-generated models.

How will this technology be used?

DataGrid plans to license the technology to advertising agencies and clothing companies so they can use it to create artificial models that are photogenic and the right shape and size for any marketing campaign.

The Swedish fashion chainH&M admitted to using computer-generated models on its website after it was confronted and challenged about "uncanny similarities" with the models. In this case, the heads of real models were superimposed on the same body. While the company emphasized that their process simplified photo shoots and would allow customers to focus on the clothes rather than the models, there was a backlash by others saying these computer-generated images set an unrealistic body image.

There is also concern that this technology could be used for more nefarious reasons. Imagine what can happen when entire bodies of people can be generated, and it's challenging if not impossible to determine what's real and what's fake.

DataGrid suggested they would offer this generative technology to other fields without going into any specifics. While it’s only speculation what those fields might be it’s plausible that instead of television news anchors or sports broadcasters, a cable or local news station might “create” its talent using this technology instead of hiring human journalists. Artificial humans could be used in any marketing endeavor to replace human actors or spokespeople. Speaking of actors, would the ability to create artificial humans reduce and eventually eliminate the need for human actors on television and films?

Now that the technology has advanced to the point, it's challenging to determine real from fake, what expectations should we have about disclosures regarding what is an "artificial human" versus a real person? At this stage, the focus is on creating entirely unique "artificial humans" rather than using an existing person's likeness, but the same technology could create artificial versions of very real humans—such as already been done through deepfakes. There could be serious ramifications if an artificial human was believed to be the real thing (i.e. a political leader) and those listening to the message believed the message was being delivered by the actual person. Rumors and conflicts could easily be started based on what a notable person’s AI likeness says.

As one of the latest developments in artificial intelligence, it’s understandable why there’s hype around GANs and the artificial human beings it can create. As the hype gives way to real applications of the technology to create images, video, artificial human beings and more, it will be intriguing to watch it unfold.

WhenDataGrid, Inc. announced it successfully developed an AI system capable of generating high-quality photorealistic Japanese faces, it was impressive. But now the company has gone even further. Its artificial intelligence (AI) system can now create not only faces and hair from a variety of ethnicities, but bodies that can move and wear any outfit. While these images are fictitious, they are incredibly photorealistic.

Using GAN to Help AI’s Creativity

The Kyoto-based startup focuses on creative applications of artificial intelligence. DataGrid usesgenerative adversarial networks (GANs), so the AI can "learn" from a database of existing images and then from there it can generate its own versions.

GANs are one of the latest developments in the ever-evolving field of artificial intelligence. GANs were first proposed in 2014 and essentially pair two AI systems together that battle it out where one system creates while the other critiques the outcome, learn from the experience and adjust to improve the quality of results. GANs can create "new" information from following the existing rules and are a very exciting development.

While DataGrid's artificial faces weren't the first to be generated, with the help of GANs the "humans" they created are the most believable, yet plus the bodies have the capability of movement. The company's website explained that this involved two lines of research and development, including "whole body generation" and "motion generation." In the past, GAN-generated images would often have tell-tale flaws that allowed you to see they are were artificial such as asymmetry in the eyes or ears and even backgrounds blending into the faces. DataGrid minimizes these flaws because they use a non-descript white background and realistic light shining down on the computer-generated models.

How will this technology be used?

DataGrid plans to license the technology to advertising agencies and clothing companies so they can use it to create artificial models that are photogenic and the right shape and size for any marketing campaign.

The Swedish fashion chainH&M admitted to using computer-generated models on its website after it was confronted and challenged about "uncanny similarities" with the models. In this case, the heads of real models were superimposed on the same body. While the company emphasized that their process simplified photo shoots and would allow customers to focus on the clothes rather than the models, there was a backlash by others saying these computer-generated images set an unrealistic body image.

There is also concern that this technology could be used for more nefarious reasons. Imagine what can happen when entire bodies of people can be generated, and it's challenging if not impossible to determine what's real and what's fake.

DataGrid suggested they would offer this generative technology to other fields without going into any specifics. While it’s only speculation what those fields might be it’s plausible that instead of television news anchors or sports broadcasters, a cable or local news station might “create” its talent using this technology instead of hiring human journalists. Artificial humans could be used in any marketing endeavor to replace human actors or spokespeople. Speaking of actors, would the ability to create artificial humans reduce and eventually eliminate the need for human actors on television and films?

Now that the technology has advanced to the point, it's challenging to determine real from fake, what expectations should we have about disclosures regarding what is an "artificial human" versus a real person? At this stage, the focus is on creating entirely unique "artificial humans" rather than using an existing person's likeness, but the same technology could create artificial versions of very real humans—such as already been done through deepfakes. There could be serious ramifications if an artificial human was believed to be the real thing (i.e. a political leader) and those listening to the message believed the message was being delivered by the actual person. Rumors and conflicts could easily be started based on what a notable person’s AI likeness says.

As one of the latest developments in artificial intelligence, it’s understandable why there’s hype around GANs and the artificial human beings it can create. As the hype gives way to real applications of the technology to create images, video, artificial human beings and more, it will be intriguing to watch it unfold.