Visual output limitations
How could you use it?
You want to illustrate an assessment, pretty up a presentation or get some ideas for your fine arts piece. Need to make a glitzy website or design a quirky logo for your new company. Trying to visualise a function or a system. Visual GenAI can do this for you.
What are visual outputs
GenAI visual outputs exist in several formats (e.g. .gif, .mp4, .jpg) but broadly speaking they are images and videos. They are generated by prompts that are text-based, image-based or a combination of both. Visual outputs can be on any topic and created in any style specified by the user.
Misuse, manipulated, modified or just plain bad
GenAI visual outputs are limited by training data sets and the underlying technology. Some outputs lack accuracy, such as generating unrealistic human hands or poor resolution. Many tools are unable to generate accurate symbols or images of text. Generated infographics often have meaningless text. Outputs may lack originality. Images can also be misused and the artificiality or the lack of originality is sometimes hard to detect.
Do you think you could spot an AI generated output? Test yourself by playing the game ‘Attention check: Authentic or fabricated'.
If that was too easy, take a look at these images and see if you can spot the signs that indicate they are AI generated content.
Specific risks and limitations
When using GenAI tools for visual output, be aware that the outputs may:
- Lack originality
- Breach copyright or privacy
- Mislead
- Be biased
- Inadvertently create offensive content
- Involve identify theft, and related to this, may lead to reputational damage
- Lack artistic integrity
- Lack journalistic integrity, and related to this, may undermine current affairs
Caution
Copyright ownership of AI generated images is also a contested space. Check the terms of the tools that you use to see what you can do with the images. Look at who owns what you’ve uploaded, who owns the output and what use is permitted of that output.
Explore risks and strategies
Explore risks, strategies and case studies about GenAI visual outputs by clicking on the plus (+) icons in the interactive below.
Stories - AI Design fails: When GenAI goes awry
We’ve all seen the not-quite right or uncanny images in our social feeds or when browsing random websites. You’ve probably also come across visually deceptive content whether that’s a deep fake or propaganda.
Sometimes the visual output just doesn’t make sense. Sometimes it’s been deliberately manipulated to deceive.
Below is a series of cautionary tales where the generation of visual output is questionable.
Remember and reflect
Key takeaway
Visual GenAI tools currently generate output that’s often questionable. Rapid developments in GenAI mean more convincing and polished images or video production are on the horizon. Regardless of the tool or technology advancements, your critical appraisal of the output remains key.
Consider
Although there are known risks to grapple with, there are also strategies to use that can bypass limitations or at least minimise any risks in generating visual content with AI.
Consider this strategies checklist to support your use of GenAI visual tools: