A few months ago, I wrote about some predictions for the next 20 years on how technology will be creating disruptions / opportunities in all walks of life. In particular, how we may capitalise on the upcoming changes in our respective stock businesses.
I’ve recently come across an interesting development linked to the mid 2010s phenomenon of fake-news. Now add onto the mix – fake-models.
Creating AI faces
Nvidia, an American technology company based in Santa Clara, California, recently developed an algorithm, after researchers fed the software thousands of real celebrity photos.
The system then improves, over a period of days or weeks, to create a more realistic and desired portrait results. Face-morphing has been around for decades but never to this level of sophistication.
Watch their YouTube video, and see if you can spot any real people among the morphing synthesized faces. Some celebrities include: Beyonce, Ellen DeGeneres, Jennifer Aniston, Channing Tatum, Elisabeth Moss, Kim Kardashian, and Ryan Reynolds.
There’s Still Room for Improvement
As with any breakthrough technology, the first results are poor. Nvidia researchers have admitted that their AI-spawned faces still leave “a lot to be desired,” but confidently pointed out that “we feel that convincing realism may now be within reach.”
One of their biggest hurdles is generating high-resolution images larger than 1,024-pixel resolution. At such small resolutions, the commercial applications are limited.
Useful Applications in Stock Photography
The most useful application are in the fields of advertising, video-games and stock photography.
For us stock photography creators, the possibility to generate realistic humans to be inserted into stock images can be highly profitable. I’ve written about how adding a person within a frame of a stock image automatically increases its “premium factor”.
Imagine instead of spending time and a small fortune hiring professional models (or at the very least asking friends/family), in a few years time, we may be able to create a multitude of different avatars featuring people of all ages, sexes, ethnicity, etc. These characters could be modified to include different expressions to fit into a multitude of different scenarios, such as a stressful corporate meeting. Think of it like the Sims but much more sophisticated.
How would human reviewers be able to tell the difference between an AI human (no model released needed) and a real one (requires a model release)? It’s too early to tell and I’m probably getting ahead of myself.
An important question to ask is whether AI presents an immediate threat to the entertainment industry, in the scale of the soon-to-be-unemployed taxi/bus drivers of the developed world – what do you think?
More information on Nvidia’s Research
To read up on this fascinating field, see Nvidia’s research page.
All faces in this blog are AI-generated.