We need to redefine our relationship to AI
Have you ever met someone who you originally thought was going to be a jerk and is today one of your best friends?
Well, that’s our story. We met in 1999, had nothing in common at first. Different backgrounds, different interests, different upbringings, ambitions – and so on. Our story should have ended after graduation. Instead, however, one of us hired the other. Then, the other hired the first one. One pushed the other into video games. The other pushed the first one into AI.
Two decades later – we’ve worked together in 4 different companies, won multiple awards for our projects, founded a company as partners, and have become uncles to each other’s children.
This is the relationship story we want to build for humans and AI: seemingly so different at first, yet able to bond, share common experiences, and grow together.
You might think that at this point in history, this makes little sense, but AIs are already more than tools. They learn by themselves. They surprise us. In that regard, they have more in common with how our toddlers learn than how common computer programs improve. Still, AIs aren’t humans. They’re a new form of intelligence. In its infancy, often limited, but constantly growing and improving faster and faster. From tools to intelligence today, and very possibly from simple intelligence to a newly dominant one tomorrow, for the very first time in the history of our species.
So how do we build a relationship with something that is as alien to us? Certainly not by training it in isolation. Or biasing it with bad data. By not interacting with it. By meaning for it to replace us, rather than augment us.
A paradigm shift is needed. We have to redefine how we’re building our relationship to AI.
To do this, we first need the right team. The right kind of people, with the right mindset. People who are natural bridges between diverse disciplines, skill sets, and areas of expertise that are seemingly unrelated; think about how the people who birthed computer graphics were much like Ed Catmull, combining computer science and hand-drawn animation. Similarly, we believe people with hybrid experience in both AI and human/machine interactions, in both tech and humanities, will connect the dots.
Then, we need to provide the AI community at large with the right toolset. Similarly to how OpenAI Gym popularized access to Reinforcement Learning, the same is needed for what we call “Human – AI interaction loop training”. Training AIs directly on interactions humans have with their environment, with each other, and with the AI agents themselves.
Finally, and perhaps the most complex step, is building the right common experiences for training AIs with us. Through shared simulations initially, we can explore a full loop of bi-directional learning, where AIs will not only learn directly from us, but we’ll learn from them as well. We can then show that the coming together of humans and AIs amounts to more than just the sum of their parts.
You don’t expect to ever be close to an AI?
Well… we didn’t think we’d be close to each other 20 years ago! Yet here we are today, on our biggest shared adventure yet: building our relationship with the next form of intelligence that will share the road ahead with us.
Join us. We can collectively craft this new relationship for the betterment of all, and show that humans and AIs can elevate each other through it.
Dorian & Fabrice, AIR co-founders