Placeholder Image

Subtitles section Play video

  • Let me tell you a story about artificial intelligence.

  • There's a building in Sydney at 1 Bligh Street.

  • It houses lots of government apartments

  • and busy people.

  • From the outside, it looks like something out of American science fiction:

  • all gleaming glass and curved lines,

  • and a piece of orange sculpture.

  • On the inside, it has excellent coffee on the ground floor

  • and my favorite lifts in Sydney.

  • They're beautiful;

  • they look almost alive.

  • And it turns out I'm fascinated with lifts.

  • For lots of reasons.

  • But because lifts are one of the places you can see the future.

  • In the 21st century, lifts are interesting

  • because they're one of the first places that AI will touch you

  • without you even knowing it happened.

  • In many buildings all around the world,

  • the lifts are running a set of algorithms.

  • A form of protoartificial intelligence.

  • That means before you even walk up to the lift to press the button,

  • it's anticipated you being there.

  • It's already rearranging all the carriages.

  • Always going down, to save energy,

  • and to know where the traffic is going to be.

  • By the time you've actually pressed the button,

  • you're already part of an entire system

  • that's making sense of people and the environment

  • and the building and the built world.

  • I know when we talk about AI, we often talk about a world of robots.

  • It's easy for our imaginations to be occupied with science fiction,

  • well, over the last 100 years.

  • I say AI and you think "The Terminator."

  • Somewhere, for us, making the connection between AI and the built world,

  • that's a harder story to tell.

  • But the reality is AI is already everywhere around us.

  • And in many places.

  • It's in buildings and in systems.

  • More than 200 years of industrialization

  • suggest that AI will find its way to systems-level scale relatively easily.

  • After all, one telling of that history

  • suggests that all you have to do is find a technology,

  • achieve scale and revolution will follow.

  • The story of mechanization, automation and digitization

  • all point to the role of technology and its importance.

  • Those stories of technological transformation

  • make scale seem, well, normal.

  • Or expected.

  • And stable.

  • And sometimes even predictable.

  • But it also puts the focus squarely on technology and technology change.

  • But I believe that scaling a technology and building a system

  • requires something more.

  • We founded the 3Ai Institute at the Australian National University

  • in September 2017.

  • It has one deceptively simple mission:

  • to establish a new branch of engineering

  • to take AI safely, sustainably and responsibly to scale.

  • But how do you build a new branch of engineering in the 21st century?

  • Well, we're teaching it into existence

  • through an experimental education program.

  • We're researching it into existence

  • with locations as diverse as Shakespeare's birthplace,

  • the Great Barrier Reef,

  • not to mention one of Australia's largest autonomous mines.

  • And we're theorizing it into existence,

  • paying attention to the complexities of cybernetic systems.

  • We're working to build something new and something useful.

  • Something to create the next generation of critical thinkers and critical doers.

  • And we're doing all of that

  • through a richer understanding of AI's many pasts and many stories.

  • And by working collaboratively and collectively

  • through teaching and research and engagement,

  • and by focusing as much on the framing of the questions

  • as the solving of the problems.

  • We're not making a single AI,

  • we're making the possibilities for many.

  • And we're actively working to decolonize our imaginations

  • and to build a curriculum and a pedagogy

  • that leaves room for a range of different conversations and possibilities.

  • We are making and remaking.

  • And I know we're always a work in progress.

  • But here's a little glimpse

  • into how we're approaching that problem of scaling a future.

  • We start by making sure we're grounded in our own history.

  • In December of 2018,

  • I took myself up to the town of Brewarrina

  • on the New South Wales-Queensland border.

  • This place was a meeting place for Aboriginal people,

  • for different groups,

  • to gather, have ceremonies, meet, to be together.

  • There, on the Barwon River, there's a set of fish weirs

  • that are one of the oldest and largest systems

  • of Aboriginal fish traps in Australia.

  • This system is comprised of 1.8 kilometers of stone walls

  • shaped like a series of fishnets

  • with the "Us" pointing down the river,

  • allowing fish to be trapped at different heights of the water.

  • They're also fish holding pens with different-height walls for storage,

  • designed to change the way the water moves

  • and to be able to store big fish and little fish

  • and to keep those fish in cool, clear running water.

  • This fish-trap system was a way to ensure that you could feed people

  • as they gathered there in a place that was both a meeting of rivers

  • and a meeting of cultures.

  • It isn't about the rocks or even the traps per se.

  • It is about the system that those traps created.

  • One that involves technical knowledge,

  • cultural knowledge

  • and ecological knowledge.

  • This system is old.

  • Some archaeologists think it's as old as 40,000 years.

  • The last time we have its recorded uses is in the nineteen teens.

  • It's had remarkable longevity and incredible scale.

  • And it's an inspiration to me.

  • And a photo of the weir is on our walls here at the Institute,

  • to remind us of the promise and the challenge

  • of building something meaningful.

  • And to remind us that we're building systems

  • in a place where people have built systems

  • and sustained those same systems for generations.

  • It isn't just our history,

  • it's our legacy as we seek to establish a new branch of engineering.

  • To build on that legacy and our sense of purpose,

  • I think we need a clear framework for asking questions about the future.

  • Questions for which there aren't ready or easy answers.

  • Here, the point is the asking of the questions.

  • We believe you need to go beyond the traditional approach

  • of problem-solving,

  • to the more complicated one of question asking

  • and question framing.

  • Because in so doing, you open up all kinds of new possibilities

  • and new challenges.

  • For me, right now,

  • there are six big questions that frame our approach

  • for taking AI safely, sustainably and responsibly to scale.

  • Questions about autonomy,

  • agency, assurance,

  • indicators, interfaces and intentionality.

  • The first question we ask is a simple one.

  • Is the system autonomous?

  • Think back to that lift on Bligh Street.

  • The reality is, one day, that lift may be autonomous.

  • Which is to say it will be able to act without being told to act.

  • But it isn't fully autonomous, right?

  • It can't leave that Bligh Street building

  • and wonder down to Circular Quay for a beer.

  • It goes up and down, that's all.

  • But it does it by itself.

  • It's autonomous in that sense.

  • The second question we ask:

  • does this system have agency?

  • Does this system have controls and limits that live somewhere

  • that prevent it from doing certain kinds of things under certain conditions.

  • The reality with lifts, that's absolutely the case.

  • Think of any lift you've been in.

  • There's a red keyslot in the elevator carriage

  • that an emergency services person can stick a key into

  • and override the whole system.

  • But what happens when that system is AI-driven?

  • Where does the key live?

  • Is it a physical key, is it a digital key?

  • Who gets to use it?

  • Is that the emergency services people?

  • And how would you know if that was happening?

  • How would all of that be manifested to you in the lift?

  • The third question we ask is how do we think about assurance.

  • How do we think about all of its pieces:

  • safety, security, trust, risk, liability, manageability,

  • explicability, ethics, public policy, law, regulation?

  • And how would we tell you that the system was safe and functioning?

  • The fourth question we ask

  • is what would be our interfaces with these AI-driven systems.

  • Will we talk to them?

  • Will they talk to us, will they talk to each other?

  • And what will it mean to have a series of technologies we've known,

  • for some of us, all our lives,

  • now suddenly behave in entirely different ways?

  • Lifts, cars, the electrical grid, traffic lights, things in your home.

  • The fifth question for these AI-driven systems:

  • What will the indicators be to show that they're working well?

  • Two hundred years of the industrial revolution

  • tells us that the two most important ways to think about a good system

  • are productivity and efficiency.

  • In the 21st century,

  • you might want to expand that just a little bit.

  • Is the system sustainable,

  • is it safe, is it responsible?

  • Who gets to judge those things for us?

  • Users of the systems would want to understand

  • how these things are regulated, managed and built.

  • And then there's the final, perhaps most critical question

  • that you need to ask of these new AI systems.

  • What's its intent?

  • What's the system designed to do

  • and who said that was a good idea?

  • Or put another way,

  • what is the world that this system is building,

  • how is that world imagined,

  • and what is its relationship to the world we live in today?

  • Who gets to be part of that conversation?

  • Who gets to articulate it?

  • How does it get framed and imagined?

  • There are no simple answers to these questions.

  • Instead, they frame what's possible

  • and what we need to imagine,

  • design, build, regulate and even decommission.

  • They point us in the right directions

  • and help us on a path to establish a new branch of engineering.

  • But critical questions aren't enough.

  • You also need a way of holding all those questions together.

  • For us at the Institute,

  • we're also really interested in how to think about AI as a system,

  • and where and how to draw the boundaries of that system.

  • And those feel like especially important things right now.

  • Here, we're influenced by the work that was started way back in the 1940s.

  • In 1944, along with anthropologists Gregory Bateson and Margaret Mead,

  • mathematician Norbert Wiener convened a series of conversations

  • that would become known as the Macy Conferences on Cybernetics.

  • Ultimately, between 1946 and 1953,

  • ten conferences were held under the banner of cybernetics.

  • As defined by Norbert Wiener,

  • cybernetics sought to "develop a language and techniques

  • that will enable us to indeed attack the problem of control and communication

  • in advanced computing technologies."

  • Cybernetics argued persuasively

  • that one had to think about the relationship

  • between humans, computers

  • and the broader ecological world.

  • You had to think about them as a holistic system.

  • Participants in the Macy Conferences were concerned with how the mind worked,

  • with ideas about intelligence and learning,

  • and about the role of technology in our future.

  • Sadly, the conversations that started with the Macy Conference

  • are often forgotten when the talk is about AI.

  • But for me, there's something really important to reclaim here

  • about the idea of a system that has to accommodate culture,

  • technology and the environment.

  • At the Institute, that sort of systems thinking is core to our work.

  • Over the last three years,

  • a whole collection of amazing people have joined me here

  • on this crazy journey to do this work.

  • Our staff includes anthropologists,

  • systems and environmental engineers, and computer scientists

  • as well as a nuclear physicist,

  • an award-winning photo journalist,

  • and at least one policy and standards expert.

  • It's a heady mix.

  • And the range of experience and expertise is powerful,