Regulating intelligence

There are many questions about how we can regulate artificial intelligence, and these grow every day. With deep fake voices and faces, anybody could be convinced to give away their money, bank details or more. How do you regulate AI?

To illustrate a basic form of AI, I asked Deepmind to create a video about the future of fintech. Here’s what it came up with:

The text was generated by ChatGPT and automated into a fairly clanky news video. But then OpenAI launched Sora last week, a much more stylish video generator. Here’s an example. Show me a video of a Kangaroo dancing or Mammoths in the snow and here’s what you get …

It won’t be long before anyone from any desk in any part of the world can create a convincing fake of a boss or family member and say: “please send money as I’m in trouble”.

Forget the Nigeria 411 scam, which is now more than ten years old, or the Facebook romance scams – did you know that Meta generates three out of five online scams? – we are moving rapidly to a place where you could be seeing fake news, fake faces, fake everythingn online, all the time.

So, how do we regulate this?

The answer is difficult as the horse has already bolted, but it must be deeply concerning that AI is, on the one hand, making our world a much better place but, on the other hand, making our world a much more dangerous place.

But then this is true of all innovation and progress. Are we Luddites or are we able to keep up? Were the Luddites right or wrong?

TBH, we could argue either way. Every step forward creates opportunity and danger, and AI is no different. Thousands of jobs will disappear, but new jobs will appear. We need people to train AI algorithms; we need people to make sure they work right; we need people to fix them when they go wrong; we need people to improve them … and so on. In other words, AI is only as good as the system that created it and that system is run by people.

So sure, we can let that system run away, unregulated and created deep fake scams, but will we? I hope not. The challenge however is how to get the deep fake world checked and regulated when it’s already out there.