AI to replace stock photographers?
In a recently published paper (HERE) NVIDIA researchers seem to have moved a long way towards removing the need for photographers or indeed reality, following advancements in generative adversarial networks (GANs). GANs are a type of artificial intelligence algorithm that can generate photographs that look authentic to human observers. Introduced as a concept in 2014 by Ian Goodfellow et al., those results were not very flattering, but nevertheless, a system was in place to generate ‘realistic’ artificial faces:
Jump forward four years and things have evolved to a remarkable level, with the NVIDIA researcher’s work now able to copy the styles of source images onto destination faces, creating entirely ‘new’ people that do look like people:
The algorithm has also been applied to house interiors, cars and cats with mixed success, with none of the bedrooms below exist in reality:The NVIDIA researchers detail their progress in the following video:
The continued development and evolution of such technology will undoubtedly have many positive attributes, but does pose potential concerns for stock photographers, for example, in addition to numerous moral and legal conundrums.
I cannot help but feel that there should be regulatory guidance for the use of such images, particularly in advertising… however, IKEA has been using computer generated (CG) rooms in their catalogues since 2010 and they now have a bank of over 25,000 computer generated models and significantly more than three quarters of all of their product images are CG. With annual website visits over 1.6 billion and 220 million copies of their catalogue distributed, IKEA seems unconcerned by their significant move away from traditional photography – and most casual observers would be unaware that images such as the bedroom below have only existed in the virtual world.