Matt Webb: playfully inventing our way into the AI future

The future of AI might reside in a long-forgotten past of computing. Matt Webb leads us along the path not taken to a more useful form of AI.

Matt invents new products with startups and large companies. His studio, Acts Not Facts, specialises in early prototyping and strategy.


Watch the complete keynote


Matt explores what AI is by making software sketches. He made a software sketch of what it would look like if app icons showed how many people were actively using that app, in little circles on the icon. His Home Screen would look quiet, and his work screen busy. But that’s stressful. So, maybe he could make that info available, but after a long swipe: swipe-to-commute. He could find out how busy digital spaces were before he entered them – but only if he really wanted to.

We have entrance rooms in houses: a place to transition into the space, to take your shoes off, and hear how busy it is inside. Could Zoom have a space like that?

Let’s look at AI: ChatGPT has no visual affordances, no hints of what it could be useful for. Notion has AI menus, that gives you a sense of what it can do, but that’s still not proactive. You have to ask, it doesn’t volunteer to help.

GitHub’s autocomplete is more proactive, but it is not flexible. Robin Sloan trained it on 1950s sci-fi to help him with writing – but it doesn’t do editing, for example. There’s still lots to do – but we’re not in a hurry. We have time.

The AI capability overhang

We’re in a capability overhang – we have 10 years of work ahead to explore the possibilities, even with the AI tech we have now. It took 10 years for the web to figure out transactions, and 15 for it to hit on subscriptions for web services.

Right now, all AI paradigms are personal: it’s me and the AI. It comes from the basic conception of the personal computer. And that takes us to Vannevar Bush, who, in 1945, wrote a magazine article, published in The Atlantic and republished in Life, taking about how research scientists do research science in public. And he invented speculative machines to support that: like a head-mounted camera to observe as the scientists observe. But the one that had the most impact was the idea of the Memes – the whole of the world’s knowledge stored in desks via microfilm. You could create “trails” between information – and that inspired the World Wide Web.

A copy of the Life magazine ended up in a Red Cross hut in the Philippines, which inspired Douglas Engelbart to build a personal computer. And from there, they invented the mouse, and then, with Herman Miller, the separate keyboard and office furniture. Which leads us to Steve Jobs, the Mac and then Windows 95.

The forgotten road of communal computing

But could something else have happened? Was there a road not taken? There was: the Sage System, built to detect nuclear bombers, which ended up as a prototype of the air traffic control system, and inspiring the interest. There were 24 stations around the US, taking in radar signals and communicating. Each had two 250 tonne computers in those. Each was used by 100 people at the same time. They tagged information on the screen with a light gun.

It was a very human operation: people working together on the same computer in the same space. And their output went into “The Pit”, where people collaborate on interpreting the information to decide if the US was under attack. It’s a system remarkably like the brain. It’s clearly more sophisticated than open-plan offices and most business software.

What would have happened if we had spent the last few decades working on collaborative computers, not individual ones?

This isn’t about all being in one physical space, or in virtual reality. But we can benefit from architecture in other ways. The doorway effect is when your brain clears out working memory when you enter a new space, ready to deal with what’s there. And a door on the screen is enough to trigger that.

AIs as teammates

So… how does this apply to AI?

How about having a group of AIs giving you different perspectives on what you’re discussing? Or a room where different specialised AIs could work together? One of his sketches is an AI that will paint in stars, but not other shapes – but is always looking for stars to paint. Another is a poet. Another likes drawing spaces. Suddenly, you have AI teammates.

We all need different people at different times. We need teammates – and AIs could be some of them.