AI robots can do a great job in managing digital assets, but may need a human looking over their shoulder.
That's the message from Creative Folks co-founder and chief executive Andrew Lomas, who says that while AI is a powerful tool in DAM systems, "it is not without its faults".
In a recent blog, he details DAM applications including image recognition and categorisation; AI-empowered video and transcripts; machine learning and more.
One primary use is the creation of creating additional metadata. "AI can transform the searchability and value of the brand assets in the DAM much faster than a human can, or wants to for that matter," he says.
But while AI has the absolute speed advantage, it can't replace input from a human or integrated business system altogether, as services such as Clarifia or Rekognition need to use models as a reference point.
"Though it has improved over the years, AI can and does submit false positives into searchable metadata," says Lomas. "This means that AI, though it is a powerful tool, will require oversight to ensure it creates the right metadata."
DAM solutions need to be able to organise AI metadata separately from human entered (or business) metadata, so 'false positives' are easy to track and remove, and users can identify AI-generated metadata and rank results.
"Despite this need for oversight, AI is critical to efficient and powerful digital image management tools," he says.
With the use of video on the increase, Lomas says finding a valuable snippet in a big "blob" of a file can be time-consuming, and AI can help with transcribing videos and assigning proper metadata to them.
"Video transcription is one of the most useful things you can leverage AI for," he says. "For AI systems, sound is much easier to understand than images as there has been more work done on this over the years, which leads to accurate transcriptions that can later be searched and used to understand what is valuable inside videos."
He says transcribing video is a mature process for AI, with several microservices available, which can be tapped into by a DAM system. Marketing teams can then gather the same metadata for videos as they do for images, making these assets much more searchable and manageable.
AI leverages machine learning (ML), with the ability to develop rules that ensure it does not make the same mistake again. But Lomas says though ML is a powerful and massive asset to a DAM service, "there are still limitations to how effective an AI can be.
"AI will never be perfect, as there will occasionally be false positives and negatives no matter how much data it has examined," he says, urging the investment of time.
Loading assets into a DAM without adding appropriate metadata is the main issue that AI integration within DAM systems are attempting to solve, and machines will still require human input to keep assets organised. Using automation to ingest the assets will reduce the gaps in the rich metadata needed when manually uploading.
"While AI promises everything a business needs without the need for a DAM service, DAM adds the context that AI needs to organise and interact with assets," he says.