You would have to have been living under a rock not to have noticed the wave of enthusiasm for AI which is engulfing the image business. All the promos sound incredible – yet the reality is somewhat different – if you don’t know how to make the best use of it that is.
On a recent trip to meet clients in the UK I quizzed each one about how useful they thought AI was at present. Whilst a few talked with widened eyes about where it was going to take them in the future, they were much less enamoured with AI’s performance in the present.
At the forefront of the wishlist was the ability for AI to do keywording at zero cost. Yet, all who had seen it in action couldn’t get past the fact that it delivered very basic, obvious, literal keywords, often with mistakes and virtually no conceptual words. Man, woman, dog and house might be suitable as keywords for amateurs, but they were a long way short of what image libraries needed.
The confidence level in what AI produced was far too low to allow reliance on AI keywords without any human involvement. And that in a nutshell is the whole problem with less-than-perfect AI – you still have to pay people to check and verify, and that cost is often close to, or the same as having no AI at all. Similar problems have been reported with finding copyright infringing images on the internet (many are missed), and recognition of objects such as brands in images (objects missed and mistakes made).
The industry-wide error at this point seems to be an over-eagerness to find a complete AI solution, when as yet none exists. As a result, time and money is being wasted trying to get AI to do the whole job, rather than coming up with better ways to enable AI to assist humans to do their jobs better.
Picsell Media is currently working with a number of AI providers who are helping us enhance the capabilities and speed of our staff. We aren’t shooting for the stars, but we’re already landing on the moon.
In coming months we’ll be announcing a number of initiatives around celebrity identification, enhanced search systems and brand identification. In each case the technology is being tailor to improve the capabilities of people, not replace them entirely.
No doubt the “nirvana” of peopleless image libraries will arrive one day, but it is better to take advantage of what’s doable now.