AI will be a theory until it isn't

For decades we've been promised a future where computers can think, but the reality has not only taken longer, the results are not entirely what the public expected. Recently AI powered website builders have been promoted heavily, but the results are less-than-appealing. AI might be able to write code and build out a webpage, but it so far lacks creativity. It's following rules.

It was at Google I/O in 2018 that I heard about the latest advancements in AI, and specifically from Google, and how computer scientists didn't quite know in the early days how long it would take for AI to be viable. It came down to computing power, and decades of theories and experiments to arrive at machine learning and the deep learning subset.

Google was touting a whole new assistant that would call businesses and verify hours, a feature we've barely heard from since. Meanwhile the worry that AI will take over and end jobs raises every time a new feature is added to ChatGPT.

We should be taking an approach to AI that Steve Jobs promoted back when the Mac first came out. The Mac is a bicycle for the mind, meaning it enhances what we can do beyond our own capabilities, just as the bicycle allowed humans to go faster and cover greater distances with a machine. The Mac unleashed the desktop publishing revolution and graphic designers were now capable of doing more in the same amount of time.

In the mid-2010s I had design colleagues working on using generative processes to perform iterative design, particularly applying copy in multiple languages to a design. Such processes are in service of helping remove the mundane tasks so designers can either be in more meetings or being creative. AI is currently a lot like basing a website design on conversion metrics—you'll know what worked but not why.

For now, AI website builders and generative artwork apps will continue to mimic what is considered best practise, even if that includes a sixth finger or an h1 overlapping h2 text. ChatGPT can help you write, assuming you know how to first write a good prompt, but the underlying understanding is not there yet. AI tools in design will for now help with iteration, inspiration and composition, but the human still gets to decide what looks right and works.