Designing for Artificial Intelligence

AI is starting to power our technology, and for the most part, it’s been flying under the radar — integrating with the tools we are already using. But after spending some time working with open.AI’s GPT-3, I believe there’s going to be a massive demand for products that enable people to interact directly with AI. This shift provides a unique design challenge as AI is unlike any other tech we’ve created.

What is GPT-3

GPT-3 is a natural language processing AI that has near-human level reading comprehension and writing capabilities. That alone is cool, but what makes GPT-3 incredible is that unlike you and me, it has actually read War and Peace and all the other classics, and most of the internet. This gives it an unbelievable amount of knowledge to pull from when creating new content.

Ways AI surprised me

I’ve been using GPT-3 to create branding content, product ideas, and landing pages — and of course, playing with it to do random stuff like write plotlines and create riddles. Here are some ways AI has impressed me:

It is creative in a new way

AI is an interesting type of intelligence because it is truly alien to us in design; it works a lot differently than we do. However, it was trained on our work so it has a lot of relatability. What surprised me in using it is that its creative work is analogous to our own but feels different in a pretty cool way.

A good example of this is DALL-E, an AI that generates images from text:

Source: DALL·E: Creating Images from Text January 5, 2021

It is clever

One of the primary things I’ve been using GPT-3 for is to create landing pages and marketing copy. GPT-3 is a great tool for this, but what really surprised me is that it also creates puns and wordplay. In other words, it’s the best copywriter I’ve ever worked with.

It is flexible

I have yet to put together a language-based idea that it cannot handle. GPT-3 can do just about anything.

It is fun

Working with GPT-3 is just entertaining and fun! Especially when you’re working with other people on it. It is an idea machine and I love seeing what it comes up with.

Design for AI Interactions

From what I’ve seen so far, working with AI is way different than working with any other type of software. It’s like having another person to brainstorm with, and you take on more the role of a curator of ideas. It’s kind of like the difference between working with a calculator and working with a world-class mathematician. Technically both things do the same work, but one of them is a lot more helpful and your job shifts in response.

AI transitions the user from a commander of the software to a coach/supervisor. You really are working with another intelligence and this presents a new design challenge for us — how do you create a tool that helps people work with AI well?

Encourage understanding of AI

Just like traditional design principles state, we need to do our best to make sure users know what is happening so that they can better engage problems they come across. This is actually a lot trickier when it comes to AI because, well… we don’t really understand it. But what we CAN do is provide a framework to help people know what AI generally does.

Embrace play

As part of developing people’s understanding of AI we also need to help them see that it is actually fun. Working with AI has incredible moments where it can make you laugh, impress you and others, and can be really entertaining, especially if you’re working with it and another person. Sharing the experience of playing with AI is simply an enjoyable pastime.

When we’re creating tools for AI, this is something that I think we should celebrate and encourage, much like startup culture embraced the fun side of work, AI can embrace the fun side of software.

Enable iteration

There are definitely a lot of different ways to employ AI as a software tool, so far I’ve experienced it as a creativity generator. In that role, being able to iterate, save work, and try new things is pretty key and I believe that systems that do this well will win out in the end.

Promote problem solving

There are times when the responses you get back from AI don’t make sense, sometimes you just need to try again, but sometimes it’s because your inputs are causing the problem. In design, it’s imperative that the inputs are exposed to the user so they can see the problems and go and fix them.

Why design matters

The UI for AI has massive implications for all of us, AI is likely to be the greatest and most impactful invention of our lifetime, and the ability to use it well will determine the opportunities people have in life. If our designs are accessible and are able to take these incredibly complex tools and make them easy to use, then more people will benefit from the power that AI is creating.

We are entering a strange new world, for the first time the power behind human creativity won’t be just the human mind but something alien to us as well. Those of us who are tasked with developing and designing AI applications will need to understand that intelligence and be prepared to translate it through our work into human oriented design.

Originally posted on Jeremykolb.com