Exponential Deception

Exponential progress is hard to understand for our brains

It is tremendously difficult for us humans to understand exponential progress. We’re just not built for it. A combination of cognitive, psychological, and environmental factors means we struggle to appreciate what “exponential” really means. We think linearly.

For example, here are 2 options - you can have £100 now, or 1p but I’d double it every day for a month. Instinctively, you’d grab the £100 - some people would say you’d be right to, a bird in the hand is worth two in the bush and all that. However, let’s think of this mathematically. On Day 2, your penny would be 2p. Day 3: 4p. Day 4: 8p. Day 5: 16p. After a week, you’d have 64p… that £100 still looks pretty good.

Day 10: £5.12. On Day 15, we’d exceed the £100 offer with £163.84. Nice! So we’ve proven it’s worth choosing the doubling penny in just 15 days. Halfway through the journey, we’ve got enough to rent a blade of grass in London.

Day 20: £5,242.88. Day 25: £167,772.16. Day 28: £1,342,177.28. Day 29: £2,684,354.56. Day 30: £5,368,709.12.

Be honest with yourself… it seemed impossible that 1p could turn into over £5,000,000. There might be a few reasons for that.

We were comparing to £100, we anchored to that value. It’s hard to see from £100 to £5 mil. We are used to linear patterns - if we walk at a steady pace, we know we will cover equal distances over equal intervals of time. The rate of change doesn’t increase dramatically - it can’t, our 2 legs can only move so fast. Even if we run twice as fast as we can walk, we can’t easily run twice as fast again, and again.

There’s also very little progress at the start of an exponential journey. Those penny-doubling calculations don’t get us excited for at least 2 weeks… a lack of early progress means we get complacent and lack urgency.

Maybe you knew automatically you’d be quids in, and your brain works differently. Congratulations - you are going to understand progress in technology, and in particular progress in Artificial Intelligence, better than the rest of us.

The race to market often obscures the rapid advancements in AI. The rush to be first leads to overhyped and underwhelming initial products, which can anchor our expectations and blind us to the underlying exponential progress.

Take image generation - in 2014 the first images of AI-generated faces were pixelated, grainy, greyscale images that wouldn’t be mistaken for images of real people. By 2019, we had faces generated that would make you double-take as to whether they were real or not. In 2024, AI-generated faces are more convincing to humans than real faces. In just a decade, we’ve leapt from blurry, and unrealistic to virtually indistinguishable from reality. In 10 years, we’ve gone from rubbish to more human than human. The same is true of the progress in voice generation, code generation, music generation, and many more. A few years ago, machines creating things of the quality and scale of what we see today would’ve been unthinkable by most people in the tech field, never mind the average person with very little exposure to this world.

This has real-world consequences. AI films have gone from joke or gimmick to becoming hard to ignore. Search for examples of OpenAI’s Sora model to find out how far we’ve come already. Film studios are now pondering whether it’s worth investing in spaces and people, or embracing AI into their production model.

With just 3 seconds of audio and a single headshot image, you can create a deepfake of you that could fool your nearest and dearest. What does that mean for scams? What does that mean for verification for things like banking? How do you know you’re talking to the person you think you are? What should we do about that?

“Well, it’s ok but there’s no way it can fool me”. You’re probably right, it’s often glitchy, with weird artefacts, and we can probably still with enough effort distinguish AI from reality. But remember our exponential progress story. That’s how AI has and will develop. The quote doing the rounds seems to be “what you see now is the worst this technology will be”.