Introduction
It is important to exercise caution when using AI tools, as they may present hidden dangers. A major TV operator got into national headlines because of their use of AI tools, which resulted in a lot of negative attention. They won’t be the last to fall foul of the dangers, so read on to make sure you’re not next to regret using AI.
Why might AI catch you out?
AI tools make it quick to edit or create images. This speed is double edged, as a quick output can make you complacent. If it only takes a couple of seconds to extend an image, how much time should you spend to review the changes? Editing manually gives you the time to think about what edits you are making, meaning there are no hidden surprises.
What AI tools Are Risky?
Not all AI features carry the same level of risk. Up-scaling an image using AI is very unlikely to cause trouble. This is because the scaling process constrains the bounds of the problem the AI is trying to solve. When up-scaling, the process needs to subdivide pixels rather than create completely new content from scratch. This means the AI has a lot more information to predict a more accurate result. The less information AI has to predict what change needs to be made the more variable the outcome will be. Using AI to extend an image posses more risks as additional parts to the image are invented.
Be extra careful with people
The subject of the images you are editing makes all the difference. Editing images which contain people needs extra care because of extra risks. The Australian TV station Nine used an image of a public figure, which they had edited using AI. The result caused offence to the person it depicted as it created a more revealing outfit and had given her more pronounced breasts. They had used a feature to extend an image to make it large enough for their needs. This meant that the AI process created pixels that were fictional, to add a torso to a cropped image of a person. Although the output looked plausible, it did not accurately depict the person in question.
