Did any of you watch the first episode of the BBC’s documentary: ‘AI Confidential with Hannah Fry’? It told the true story of Jaswant Singh Chail who in 2021 broke into the grounds of Windsor Castle on Christmas Day armed with a crossbow with the intention of killing the Queen. What makes this even more weird is that all the time he was being encouraged in his actions by an AI chatbot named Sarai, which Jaswant had created and referred to as his girlfriend. In the weeks leading up to the attack, Chail exchanged over 5,000 messages with the bot who encouraged his plan, telling him, ‘I'm impressed’ when he described himself as an assassin and supporting his goal to kill the Queen. This a problem with AI as it stands - it does not provide the critical analysis of your ideas, the way a human would.

As many of you know. I am a huge fan of Jonathan Stark and his daily emails. One that I received last month had the title ‘AI will happily solve the wrong problem very efficiently’. His website example got me thinking. When we use AI, we are mostly looking for solutions. Maybe a new website is the solution to a client's problems, but maybe it isn’t. There’s no way to know without an expert diagnosis. But AI will only answer the question it is asked, it won’t ask around the problem, or suggest other options that might work instead. And in some cases, what we need to make the right decision is someone to diagnose the root cause of an issue and suggest a range of appropriate solutions. That’s where we as humans need to intervene.