First Impressions On ChatGPT For Product Design
AI is a hot topic right now, not only in the design field. Every day, there seems to be an endless stream of hype messages about what AI can do now and how that is going to transform our lives.
I wanted to chime in, or at least better understand what the hype was all about. So I fired up ChatGPT, probably the most talked-about AI this year (so far, at least), and started to play around with it. How useful is this thing for doing product design work? Will it replace me as a designer? Do I have to be afraid of it or look forward to using it in my workflow?
What Is ChatGPT?
From the website of OpenAI, the makers behind ChatGPT:
“We’ve trained a model called ChatGPT which interacts in a conversational way. The dialogue format makes it possible for ChatGPT to answer follow-up questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.”
ChatGPT is a chatbot. That means I as a user can interact with it via text messages, like a chat. I can ask questions, but I can also ask it to do things (like write a recipe for me using eggs, ham, and butter).
The interface looks very simple, as it is a text chat. Now that we are here, why not ask ChatGPT to explain itself?
That's a thorough explanation, but it might be a bit too technical. As I quickly learned, ChatGPT is great at explaining things and you can control the level of detail.
Here I asked ChatGPT to explain itself again, this time to a five-year-old:
Completely different explanation than before! To sum it up, ChatGPT figures out the next word to say based on all the data it was fed and the feedback it got from real humans during its training. That is an important point to understand: It answers based on the probability of words. If, in the given context, the most likely word to follow is “dandelion”, this is what ChatGPT will write.
Using ChatGPT For Product Design
Let's start to try some design-specific things. I walked through my budget app experiment to see how ChatGPT could help during the product design process.
At the start of the process, there is a problem you think is worth solving. In my case, that problem was budgeting and not being happy with existing solutions. I pretended not to know anything more than that and asked ChatGPT to help me with budgeting my money. It came up with a detailed plan to follow for budgeting and mentioned budgeting tools and apps, so I asked for some recommendations. It provided me with a list of tools, including a short description of each.
Next, I got a bit more specific and explained my problem with the app I currently use. I asked ChatGPT for alternatives for me and my issues. At this point, I noticed that I was starting to treat ChatGPT more like a casual conversation with another human than a chatbot or a search engine.
This exercise was simple and took only a couple of minutes. I was quite happy with the experience because ChatGPT could provide me with alternative tools to look at in my context—something that would have taken a bit longer had I done this research myself using a search engine.
But, it is important to take all this with a grain of salt. The data ChatGPT is trained on ends somewhere in 2021 and as such it only has limited knowledge of all the things that came after that point. In my case, there could be an awesome budgeting app that launched in 2022 and ChatGPT would not know about it.
From Research To Design
Now I tell ChatGPT that I want to make a budgeting app myself and ask how it could help me with that:
It is also worth noting that you can get different answers deepening on how you frame your question or task. In this example, I provided much more context when asking to make an app for budgeting:
In general, ChatGPT seems to be a great tool for idea generation. Even if some things are pretty generic, you bounce ideas back and forth with it because ChatGPT makes use of the conversation history.
It also seems to be able to explain why it made certain suggestions. Here I asked it why it suggests having a monthly budget overview as part of the app. Going further, I asked it what such an overview would look like and it described a layout to me.
After playing around with ChatGPT for a couple of hours, I'm pretty impressed. It often feels like having a conversation and it can do a lot more than answer questions and explain things to you. ChatGPT can write emails for you, give feedback on written text, summarize entire books, and much more. It is easy to see how people get amazed by it.
I like the research and ideation capabilities, but you could also use it to write placeholder text (or even final copy) for your designs, learn about design methods to use for specific tasks, and more. It will speed up certain parts of the design process and can make your life as a designer easier.
On the other side, it just creates a series of words. It can't provide sources. It is trained on data that contains biases and prejudices and it will reproduce them.
I am also not sure how good its reasoning is. Sure, it could explain to me why it suggested including a specific feature in the app. That's enough to give you a starting point for thinking about it yourself. But that is an important part here: Think for yourself.
Even if you can get AI to do the whole design process for you, ask it (and then yourself) why the result is what it is. Why was it created in the first place? Why does it look like that? Why does it work like that? And why should it exist in the first place?
You can feed a lot of extra information to ChatGPT to get pretty specific results out of it. But ChatGPT can not talk to your potential users (at least not yet). And it can not listen, understand and interpret what they tell you. It can just create a series of words. But that is already impressive and useful as long as you keep thinking for yourself.