Sending a fax from the beach

AI can do amazing things; what an incredible time to be a designer

In 1992, AT&T made some very public predictions about the future of technology. You should watch the ads—they’re fantastic and were directed by the brilliant David Fincher—but if you can’t, they’re provocations that highlight where technology is headed. In a movie-guy voiceover, you’re asked “Have you ever borrowed a book, from thousands of miles away?” or (my personal favorite) “sent someone a fax, from the beach?”. The voiceover continues “You Will!” and that’s because of AT&T.

I’m not here to hate on how they got the predictions wrong. In fact, I love how accurate they are. They’re so close. Video chats from a phone booth? Well, kinda. Concert tickets from a cash machine? Yeah I suppose so. The whole series is like the uncanny valley of predictions. It reminded me of my most recent beach holiday, where a particularly annoying guest was sitting on his sun lounger with his headset, yabbering away on his conference calls. I would have much preferred it if he were only able to send a fax from the beach…but I digress.

I’m sharing these old AT&T ads because I think we’re at a similar inflection point with AI and conversational interfaces. Everyone’s trying to guess what’s next, and we’re all circling the same design questions.

Translating human intent

Julie Zhou wrote a brilliant article on the limitations of conversational interfaces and I absolutely loved this line:

“The entire design discipline can be distilled into the craft of translating a creator’s intent into a user experience that fulfills the desired intent.”

It got me thinking. Wondering if my intent is the same as everyone else’s when I approach my AI. When I think of Gemini and how it might be able to help me. I then had a typically millennial experience with AI—I went and asked it to do something for me.

I asked it to write itself a prompt. I wanted it to go and do some deep research into how different generations are using AI and conversational interfaces.

It then returned with a comprehensive report, and offered to code an interactive web page to explore the results. I said yes, and hosted it on my website so you can dig around in the results. It took Gemini 20 minutes from me asking the question to having the site live. Impressive.

Open-ended vs. Closed

But here’s the thing. The report points out the main difference between generations using AI—and Sam Altman has also spoken about this—is that their intent seems to chart over a spectrum. This might be an over-simplification, but the way I see it, the younger generations are more comfortable using AI to navigate open-ended problems. Their intent might be unclear, even to themselves. They might want advice on a big life decision. They might be weighing the pros and cons of something. Expanding their individual understanding of a topic. Whereas older generations (like me) often have a closed problem to solve. A specific task that needs to be done. A request to make of the system.

Now this might be more to do with the situation than anything else. Folks in their 40s are busy with work and other commitments and use tools like ChatGPT and Gemini to get shit done. Whereas 18 year olds are focused on learning and generally expanding their minds—so why not use these tools to do just that?

But what if the difference is driven by something else? A level of comfort younger people have with sharing. Sharing their thoughts, sharing their hopes, dreams, desires. When I was 18 (and admittedly, I was in the UK where everyone is emotionally repressed), it would have been really odd to be as openly hopeful or ambitious as many 20 year olds are today. I love it. I’m here for it. But it has design implications.

How should we design for these open-ended intentions? And if younger users are using tools like Gemini and ChatGPT as operating systems, why aren’t we designing them to work that way? In my own experience of writing this article I had a very closed-task approach to using Gemini. I asked for some research. Gemini did that research to a high standard. It then offered to turn it into an interactive website. That’s nothing short of amazing—it was incredible to see Gemini code the interactive report in real-time. But I don’t think Gemini truly understood what I was trying to do. Sure, I could have asked it for charts or quotes to drop into this Substack piece. I could have nudged it to help me formulate the narrative of the article. And it wasn’t short of ideas. Gemini even offered to turn the report into an audio summary for you folks who prefer to listen to Substack instead of read it.

The thing is, I can’t shake the feeling Gemini just sent me a fax from the beach.

What an incredible time to be a designer.