The integration of artificial intelligence (AI) into the design field has created new possibilities and transformed workflows. While some of the integrations have been subtle and have been integrated gradually, such as the neural filters in Photoshop, the latest versions of Adobe’s Creative Cloud programmes now feature generative AI.
These new, more powerful tools provide users with the ability to generate characters, patterns and whole scenes just using text prompts. However, while using AI for design offers a range of benefits, it also has plenty of limitations and potential drawbacks.
Four benefits of using AI in design
- Automation of repetitive tasks, which would previously have taken considerable time – such as removing elements from the background of a photo or extending a tightly cropped image – can now be done by AI, freeing up your good old human designer to focus on other important elements.
- Useability – neural filters in Photoshop can sometimes help resize that previously unusable too-small, too-dark photo supplied by a helpful colleague into a useable image, recolour black and white photos, and adjust the perspective of text and graphic elements to match the background.
- Exploring design ideas – generating variations on a theme or initial idea. This can be particularly helpful for brainstorming sessions or when working on complex design challenges – designers can use AI prompts to generate numerous iterations to test different options and help determine the best approach.
- Generating visualisations – AI can sometimes help mock-up a creative concept that can be tricky for a client to visualise. AI generated outputs (like the one at the top of the page), with their various quirks and small errors, whilst absolutely not good enough to use for the final artwork, can be helpful in establishing a visual style and composition, and improving understanding of the concept.
The drawbacks of using AI in design
For all its benefits, AI still has many flaws and drawbacks. As impressive as some of the generated images can be, AI tools tend to provide you with a specific item – in a specific composition, viewed from a certain angle and illustration style – and the results can be far from ideal.
While AI can easily create simple objects and even match an artistic style, generated graphics all too often have glaring errors which make them unusable. People look weirdly plastic and doll-like, hands and fingers are often either missing or overly abundant, and even seemingly simple, everyday items like cars can be missing vital elements like wheels.
Another area where AI struggles is in understanding context and nuance. Where a human designer will have an overarching understanding of the brand, target audience and desired outcome a piece of work is to have, AI generated output is dependent upon generalised knowledge that has been gathered from a wide range of sources. As such, results are often somewhat generic and clichéd and sometimes downright bizarre.
Feeding more detailed information into an AI such as Bard or ChatGPT to generate more nuanced prompts for a specialised AI image generator to achieve better outcomes is possible. However, this can be both time-consuming and expensive.
Most significantly, there are also questions regarding authorship and intellectual rights over ownership of the results, plus questions about unlicensed content used in training data.
There are currently several large outstanding lawsuits progressing against AI companies from a wide range of artists, writers, and image providers (including Getty, the image licensing giant, and the New York Times) for using their work without permission to ‘teach’ generative AI programmes.
Overall, AI tools can be a valuable asset in the design process, but they are not a magic wand or panacea – rather, they are a useful addition to the already wide-ranging tools at our disposal.
Successful design requires a human touch and deeper understanding to ensure it meets the project needs and delivers the desired results. Pretty is not the same as effective. If you’re looking for design that delivers, you still need a good team of people. Luckily for you, we happen to be experts.