Myths about AI

Evans on machine learning an AI

Florian Sauter

2 minute read

What comes into your mind when thinking about artificial intelligence and machine learning? Assosiations of flying saucers? Or Google, Facebook and China ruling the world with their (evil?) empires of data? Or the conversations with family and friends that have read in the newspaper that AI will now take over and replace all of our jobs?

These are all common things you stumble upon in Ai discussions, but none of them are very helpful to understand the essence of what machine learning today really is and how it can help us. Benedict Evans wrote a fabulous blog article on this exact topic which I can highly recommend reading (here at full length)[].

What’s more helpful, and I fully agree with Evans here, is to think of it as a tool and an enabling technology that can help us a lot in automating things. And usually, this kind of automation is bound to a very narrow domain. I like Evans metaphor of today’s machine learning capabilities: It’s like you could create an infinite sized army of something between a dog and a ten year old and subtract their common sense (which no AI system today can imitate) and use them to process huge amounts of data.

As far as we can see today, this will lead us nowhere close to flying saucers and universally intelligent robots. Much more likely, this will bring us great gains in the way we can automate very specialized tasks and, as a technology platform, unlock things we just cannot do (or imagine doing) today.