A Call For AI Courage

I was heading for University, and I think I’d signed up for English Literature or Maths and Business (well, it can’t have been a good fit if I can’t remember which it was) when I saw the “Artificial Intelligence” degree course and I was stopped in my tracks. I’d watched films like iRobot with Will Smith (this was pre-Oscars slap, so he was still “cool” back then) and was always captivated. I thought “I’m a little like Will Smith” (pre-slap, pre-slap…), so I switched my course to AI and Computer Science - straight away I was hooked.

I saw from my first interaction with AI that this could be humanity’s defining contribution to the universe. Creating intelligence is what we give God the biggest credit for - this was humans aiming to achieve the same thing. A machine that can “think”.

Our intelligence has evolved slowly over billions of years, and is confined to a fleshy lump of grey matter in a bone case on top of a fragile ape body. But machine intelligence is theoretically limitless, and has no such constraint as biological evolution or organic matter.

If a machine can “think”, it can “think more” by just adding more compute power, more chips. It can “think better” by adding more data, and consuming everything that has ever been written down or recorded, and everything that will be. It can “think forever” by making copies of itself, and upgrading to new hardware as it becomes available. It can improve itself, not just through self help books and trips to the gym, but by constantly improving and rebuilding its architecture from the ground up.

The upsides of achieving this are truly remarkable. Let’s take one example - personal healthcare. Imagine a doctor in your pocket, available to talk to 24/7, analysing the huge data set produced by your body each second, knowing your whole medical history, every medical paper that’s ever existed, every intervention and its effectiveness on people that match your specific profile, can discover new treatments and drugs specifically targeted to you, can simulate 1 billion different paths forward and listen to your every feeling, worry, complaint, question without judgement, who is dedicated completely to you with the entirety of its existence and attention and yet can survey and pool data from every human experience - who never gets tired, distracted, closed-minded, hungry, etc

Now apply that to all aspects of life - food production, conflict resolution, energy production, protecting the environment, transportation, entertainment, education, relationships, etc… AI has the potential to transform everything we experience for the better.

All of that might seem a bit far-fetched… especially from where we are standing now. AI is still pretty stupid after all, but it’s in its infancy. We are looking at a toddler, bumbling around trying not to fall over… what will it grow up to become? As all parents experience, it is sure to grow up too fast and we hope we’ve done our best to shape it!

In the excitement of the potential of AI, I looked around and saw a society not really ready for its promise. Economic systems built on valuing people through work - they break down. Social fabrics built on trust - they crumble. People’s purpose - it’s unclear what it will be.

Many people who’ve worked in AI often feel a sense of excitement and mourning for the future. In a world of superintelligent machines (that can outperform humans in every task and domain), what agency, impact, or meaning will humans have? Will we be satiated slobs with no purpose? Will we be unleashed creative geniuses with time for pursuing passions and enjoying life? It’s hard to say - would we all even agree which we should aim for?

AI is why I created the School of Code. One way I saw to potentially help was to support people to rapidly transform, to bring them into the technical world, and empower them to have a voice in the conversation. A hand on the wheel instead of being left behind.

AI is not a tool. It’s intelligence. There’s questions about how we should interact with it. For example, even if it is aligned to our best interests, what does that mean? Should it intervene when we ask for something that’s bad for us? Seems obvious in some cases, yes… If I ask it to harm someone, I’d hope it would refuse… but then what if I’m hurting myself by eating pizza for breakfast, lunch, and dinner… should it refuse to order the food I’m asking for? What if the AI isn’t aligned to our best interests at all?

These lines are blurry, and these problems are those more associated with philosophy lectures, not technology labs. And yet, those are the discussions we need to have and that are required to shape the future.

Elon Musk described humans as potentially a “bootloader for Artificial Intelligence”, saying he has “a deliberate suspension of disbelief in order to remain motivated” to keep making progress and overcome existential paralysis. I feel instead of ignoring the negatives, we need to wrestle with them.

The Stockdale Paradox, named after Admiral James Stockdale, encapsulates the approach we need. He was a prisoner of war, and survived captivity by confronting the brutal facts of his situation while maintaining faith that in the end he would prevail. Similarly, in AI development, we must confront the brutal facts, the dangers and pitfalls, but maintain faith and strive for a shared vision of a better future.

We can’t be naive and think there’s only upside. We can’t be cynical and think there’s only doom and gloom. We need to be courageous - to have the strength of mind to carry on in spite of danger. We need to confront the potential demons that lie ahead, mitigate them, but plough forwards together.

I’m looking to bring together as wide a group of people as I can to have these discussions and make sure we can shape this path forwards before it’s too late!