Copy goes here

© Nick Dunmur

AI, TDM and IP...?! What are all these acronyms, and why should I care?! 

Acronym – (noun) A word formed from, or based on, the initial letters or syllables of other words. 
The world is full of them, seemingly nowhere more so than in the land of the creative industries. We have acronyms for everything, but there are some that you need to be aware of and, dare we say it, take an interest in. The following is an attempt to unpick where we, as image-makers, stand in regard to the development of Artificial Intelligence (there’s your first acronym… ‘AI’). 

You might already know the other two: 
‘TDM’ = Text and Data Mining. Essentially, this is a process of analysing machine-readable content (also known as ‘scraping’) with a view to extracting the components in order to do something else with them. Mostly, this takes place with digital assets, but of course one could add in a step to scan physical ones. 
‘IP’ = Intellectual Property. This, of course, includes copyright, the form of IP we are most likely to care about, seeing as we licence the use of it (or should do!) to generate income for our businesses. 

It’s important to have an understanding of this, because since people started making stuff, humans have largely been unchallenged as the source of creative work. But things are changing – AI platforms are present and developing, which can, on the basis of a prompt of a few well-chosen words, generate a unique image representing whatever those words were. You may have already played with an AI platform (like DALL-E 2 or Midjourney) to see what kind of fantastical image can be realised through the combination of the words, (for example), ‘badger’, ‘planet’, shirt, ‘high jump’, all in the style of Georgia O’Keefe. (I haven’t, before you ask!). 

These AI platforms need to have raw material on which to be ‘trained’, before they can start generating their own type of unique material. Guess what makes appealing training material?… Image-text pairs, otherwise known as photographs (or other visual works, of course) complete with metadata (in the form of keywords or captions)! As photographers, we have shedloads of that valuable IP, both online and visible. Prime real estate for AI firms looking for training material. All of which brings us back to TDM, the first step in the process of training those AI platforms. 

Some of you may be aware that we already have an exception to copyright protection, for the purposes of non-commercial TDM. The words in italics are important because this qualifier specifically excludes the activity that many AI platforms are embarking on; that of identifying, copying, extracting and mining data from our work for commercial use. We should be paid for that commercial use and if the IPO (sorry, another acronym… the Intellectual Property Office) have their way, the current exception will be broadened to allow this commercial activity, with no recourse for the creator to be able to opt out or easily restrict access to our work, unless we are prepared to place our websites behind some sort of subscription page or log-in, and who wants that?! We want (nay, need) our work to be seen. 

As you know, we are doing as much as we can to push back against this – we have already talked about that work in previous newsletters, but this piece is more about giving you the underlying nuts and bolts of how it all fits together and why it’s important. 
It’s why I wrote to our members recently, suggesting we all amend our website Terms of Use to include a clause to prohibit this type of activity, in the event that should we need to prove the contractual basis on which someone has unlawfully accessed and used our work, we can point to that clause as evidence. 

Of course, we are well aware that much ‘scraping’ (the word sounds as awful as the impact) of our websites has already taken place and that we are in uncharted waters as far as the rate of development of AI is concerned. We can take some comfort from the fact that we are not alone in being concerned and that the concept of ethical AI development is alive and well. Were this to become the mainstream, the notion of getting paid for the use of our work for the purposes of AI training might become more of a reality. No-one is going to licence the use of a few hundred photographs and accompanying metadata directly with a licensor – it’s simply not viable – but they may well be able to through secondary licensing via a collective management organisation like DACS or PICSEL. Many of us already get a share of income via DACS Payback, for the use of our work in certain periodicals, on TV, etc. (more on DACS Payback here) and if you haven’t thought about this as an extra revenue stream, perhaps you should. 

I have tried to keep the technical language out of this piece as far as possible, but it is complex and the world we face as creators is not getting any easier which forces us to engage with this stuff more than we might otherwise wish to. 

 

Nick Dunmur, 31 Oct 2022 

AOP Photography Awards

The AOP Photography Awards are known as the ‘Oscars’ of the photography world. They celebrate excellence in the creative photography and image-making industry.
This is your chance to be seen by leading commissioners and names within the photographic industry.
Enter our 39th AOP Awards…