DAM and AI Growing Pains: Ingestion or Indigestion?

VP for Nuxeo, Uri Kogan, has recently published the penultimate article in his 7 Common Beliefs in the DAM Market series. For this post, Uri turns his focus to that ubiquitous herald of the new age, Artificial Intelligence (AI) and more specifically, its potential uses within Digital Asset Management.

It’s undeniable that AI and machine-learning have earned their scout badges providing bots, virtual agents (“I can’t let you buy that Dave, but you might like one of these”), speech recognition and neural networks, but we’re not sure if we’d want to let this eager newcomer control, curate and deliver our digital assets for us just yet.

As an example, Uri lays out a typical workflow for us – an 8-step lifecycle of a digital media campaign from project creation, photo shoot and cataloguing, through to publication and eventual archiving, with a critical recall along the way.  He summarises these stages as “examples of AI and advanced analytics helping DAM users and administrators accelerate every step of the content lifecycle”.  However, after a cursory examination of these steps, it becomes apparent that AI is mostly being used as a synonym for ‘advanced analytics’, a common misinterpretation that’s gaining popularity as the bandwagon fills up with those hungry for the holy grail of digital reasoning.

The first two steps of the lifecycle perfectly labour this point:

1. Project creation: An intelligent system should take into account your historical performance on the steps of a creative process to suggest timeframes for deliverables. High performing people are usually also overconfident in their ability to hit deadlines – in fact, research shows your odds of hitting a deadline that you’re 90% confident you’ll hit are actually no better than even. Oops! Computers don’t suffer from cognitive bias, so they can help anchor our expectations to reality

2. Photoshoot: If you’re working with an agency or freelancer that’s uploading content to your system, why do you have to go to the trouble of getting IT to grant them credentials and set up a dropfolder for them? Your system should be smart enough to connect their contact information to the relevant projects so they can use a generic hot folder and the system can autoroute that content to the right place” [Read More]

The creative process undertaken by humans is (or should be) aware of its own cognitive bias, which is adaptive by nature and often goes against all logic. It’s true that computers don’t suffer from cognitive bias, but at this stage in AI’s evolution, it harbours something that’s potentially far more chaotic and inappropriate: the cognitive skills of a toddler that has no social, political or emotional perception or subtlety, and is thus prone to making monumental blunders.

The second point appears to be more admin than AI; the question here is whether we should trust an AI system to autonomously grant privileges based on its own assumptions.  Robust, mature and flexible authentication features are now baked into most systems and generally require more common sense and vigilance than intelligence.

Uri’s other examples of using intelligent agents are more valid, such as employing Google Vision to search for images that need to be taken down in a crisis situation.  Many of the scenarios, however, are concerned with selection, review and relevance – processes that rely on two crucial linchpins: the conscientious cataloguing and tagging of digital assets and the input of accurate, comprehensive metadata.  In our opinion, this is still the most meaningful way of breathing intelligence and awareness into any DAM or content-driven system and depends on  contextual awareness (i.e. what meaning the metadata is supposed to convey).  At this stage, while there are hints that AI can add value, it still requires significant contribution from human beings and at a level which is being radically underestimated by many in the DAM market.

Share this Article:

Leave a Reply

Your email address will not be published.