Hey guys, Andy here. Generative AI took another leap last week with OpenAI including the works of renowned Japanese animation studio, Studio Ghibli, into the training data. This led to a deluge of Ghibli’d content which has re-ignited the debate around AI. I have talked about language and photography previously; today I can not help but look at the often heated arguments surrounding AI and think how the language being used in its defence does a lot of heavy lifting.
In regards to generative AI, I feel the language the tech industry uses is, at best, misleading. The phrase AI learns (like humans) feels instinctively wrong alongside the notion that copyright is controlling innovation and that AI is, merely, a tool that is democratising the creative sector. As a portrait photographer I offer a service and I find many who step in front of my camera have an intrinsic feel of what they want but lack the vocabulary to properly express themselves. In such instances I see myself as a guide, offering them the terms they require for us to reach the destination and goal of a meaningful and accurate portrait. Which is why I become suspicious when I come across inaccurate language.
From what I have been led to believe, humans learn through reasoning and criticism as well as adapting our knowledge and experiences to different contexts. Compared to this generative AI machine learning is surface level statistical pattern-prediction which can mimic the data it has been given to train on. In regards to Studio Ghibli, AI is able to offer content according to the studio style because of the patterns it has encoded from Studio Ghibli’s works. It is also possible for AI to regenerate the training data to such a result that it requires attribution or licensing, should the data be protected by copyright.
Thus AI has been described as a plagiarism program that commodifies content as it does not create original content, rather it offers content recombined according to the training data. It has not learned per se; it has copied, it has encoded, it has compressed and, when prompted, digitally regurgitates a version of something already in existence, which it can do at scale.
Words missing from the defence of AI are words such as skill, talent, and ability. When I look at, read or listen to a work, without the tools or talent there is simply no way for me to replicate what I have just experienced, let alone it to a reasonable standard. For example no matter how many times I may watch or try, I will neither be able to run the 100m in less than ten seconds nor paint a Picasso.
AI can cure cancer and AI needs to violate copyright otherwise innovation will be crippled. Really? Superficial statistical pattern-prediction can understand complex systems and it is in our geo- political interest to allow AI to scrap all creative content for free? Instead of copyright controlling innovation, how about it is protecting authors. If there is no protection, there is no incentive to create new output; instead we will see power consolidating in the hands of Big Tech.
AI is being described as a tool; but is it? As a photographer, I have used AI as a tool to help me complete a task. However perhaps AI is also a service which completes a task for you? Write an essay, compose music, draw a sketch; how much of the task are you undertaking yourself compared to it being done for you? Also, if AI is a service, then it becomes beholden to legal obligations which I am sure the tech companies would prefer to avoid.
Lastly there is the use of the word democracy. Oft used in the past decade, it has been claimed that camera phones have democratised the photography industry; this phraseology is now being applied to AI and art. The word is being used not in association with a democratic system or principles but rather that something, in this example a camera, is now accessible to everyone. I appreciate meanings of words can change and evolve over time but, to me, using democratisation instead of accessibility infers exaggerated grandiose aspirations of freedom and political connotations which are unsuitable to the conversation.
Language and language accuracy matters. The language that AI companies are using frames AI as a one stop, all encompassing solution to everything in our lives. That it is a bandwagon we must jump on and harness, now. It feels that the language being used is deliberate, to deliberately make AI appear benevolent and to misdirect a wider public as to the limitations of it.
andybarnham
I am a portrait photographer based in Cheltenham, UK. Born in Hong Kong to a Chinese mum and British dad, I had an international upbringing while I educated in the UK. I started photography as a hobby while serving as an officer in the British Army.
After my service I turned this passion into a career and became immersed in London's sartorial scene. I am now focusing my camera on portraiture and using this eye for detail which was refined over ten years. As a former Royal Artillery officer it is only fitting I shoot with a Canon camera.
