Nowadays, artificial intelligence is everywhere. It seems everyone wants to use this modern technology for any application. But what exactly is AI? What can it do? And more importantly, what can’t it do? In this blog post, we’ll explore the strengths and weaknesses of generative AI for website development.
What is generative AI?
The term AI has been around for a while, but today it’s primarily used to refer to generative artificial intelligence (GenAI). These are models trained on large amounts of data to learn to recognize patterns. The most well-known types of GenAI are language learning models (LLMs), such as ChatGPT, and text-to-image models, such as Stable Diffusion and Midjourney.
An LLM model is trained to recognise patterns in human language and can independently form coherent sentences based on a user’s input. For example, if you ask ChatGPT to write a short story, it will do so based on texts it has analysed.
A diffusion model, like Stable Diffusion, is trained to generate images based on textual input. For example, if you ask it to draw a puppy, it will try to recognize a pattern in the random noise that matches your description of the puppy. This noise gradually becomes more structured until you eventually get a random image that hopefully matches what you had in mind.
What generative AI is not
Many people overestimate the potential of AI. GenAI is often used in applications for which it is not suitable. AI is not a truth machine. The generated results are based on randomness and will often contain nonsense. It is sometimes said that AI “hallucinates.”
A good example of this is a lawyer in Utah who used ChatGPT to write legal documents. These documents referenced other legal cases that didn’t exist. (Source: The Guardian) The lawyer had used AI to save time and money, but this task wasn’t suitable for ChatGPT.
It’s important to remember that ChatGPT is a language model and therefore primarily attempts to form coherent sentences. ChatGPT has no concept of truth and will often generate grammatically correct nonsense.
“If droids could think, there’d be none of us here, would there?” -Obi-Wan Kenobi, “Star Wars Episode II: Attack of the Clones”
AI for the development of your website
With GenAI’s strengths and weaknesses in mind, we can examine some applications within website development. What can you use AI for in website development, and what can’t you?
AI is useful for:
- Time-consuming work that requires little creative insight. For example: sorting tables, detecting typos in your copy, or removing unnecessary tags from your code to improve performance.
- Inspiration. For example: Generating color patterns or suggesting fonts.
- Image generation. Can’t find good images to match your content? You can use Stable Diffusion to generate one. Be aware that generated images can often contain bizarre errors, so try not to overdo it. We generally prefer stock photos over generated images.
AI is not useful for:
- Complete designs. Even when AI generates a concept for a web design, you’ll still need to make adjustments to make the design more original and less generic.
- Modern designs. If you generate a design with a text-to-image model, you’re using a model that was previously trained on existing images. AI is therefore always one step behind. Training models also takes a long time and requires a lot of energy. Sometimes it’s better to look around and explore the latest trends yourself.
- Save time. Many people use AI because they think it will save them time. However, AI is only a first step. You need a human eye to review the designs. Then you have to iterate, adjust, and implement. So, using AI is certainly not a shortcut and can be less efficient in some cases.
The OpenArt blog features several mockups of AI-generated designs. While the designs on that page certainly look like mockups, they’re not really usable. They’re often generic or impractical. They’re also just images that serve as a starting point. There are already plenty of mockups or templates online created by professionals. In this case, AI isn’t solving a problem that hasn’t been solved yet.
Summarised
AI can be useful for generating inspiration or performing simple tasks. However, the human touch remains crucial. Contrary to popular belief, AI is not creative and will not generate new insights. A web design developed by AI is inherently dated and lacks originality.
AI can be very useful as long as it’s applied correctly. So, don’t ask yourself how AI can take over your job. It’s better to ask how AI can assist you.
