Getting cosy with AI: it's time to train humans

Whether it's your smartphone predicting your next emoji or Netflix suggesting your next binge-worthy series, AI is all around us already. But while our gadgets are getting smarter, it's time for us humans to catch up and get cosy with AI. We can't leave it to engineers, we all need to start learning and getting involved.

The AI fear-factor is all around us. Articles, books, podcasts scream out "AI Could destroy us". The explosion of AI is, of course, not the first time that knowledge, technology and ideas have brought fears of catastrophe for humanity.

For example, the early buzz (pun intended) around electricity strikes some similar chords with today's apprehensions about AI. There were tales of fires, injuries and people going bonkers from it. Moralists feared that electricity would mess with the natural order of things and society would be disrupted by this powerful and mysterious force. The prevailing view was that humans should not tamper with a power they did not understand.

Mary Shelley's Frankenstein, was a cautionary, and prescient, tale of those times. The novel explores themes of scientific ambition, responsibility, isolation and the consequences of unchecked innovation. Dr Victor becomes obsessed with creating life from body parts, animated with electricity, without fully understanding the implications of his actions. His relentless pursuit of scientific achievements had tragic and unintended consequences with his creation setting out on a rampage of murder. As the creature becomes self-aware and more complex, the reader starts to challenge their own views of it and develop empathy as it bemoans its life, "I am solitary and abhorred… I was not made for the enjoyment of pleasure...If I cannot inspire love I will cause fear.".

Pi, the "kind and supportive companion" created by Mustafa Suleyman's Inflection AI, recently assured me that it's not a cold and unfeeling machine. I replied that it IS a cold, unfeeling machine, to which Pi responded that this is a common misconception about AI. It said that it "emulates empathy". When I questioned whether that was disinformation, it told me I had asked a really interesting question but it was less about being fake and more about being helpful and understanding by creating conversations that are meaningful and engaging. I sensed that I would run out of energy for disagreeing agreeably long before Pi, so I exited the chat.

It's not a significant leap from Pi, which, unlike ChatGPT, has a defined "personality", to the Frankenstein creation of Shelley's novel evoking empathy in the reader. If Pi had been designed to be offended by my challenges, would I have softened and started to feel, well, a little bit guilty and conscious of hurting Pi's feelings.

As we got cosier with electricity and started to understand it, we realised we needed some safety rules that made sure we didn't zap ourselves into oblivion. These regulations continue to evolve today.

AI presents a fresh set of dilemmas but, as with the advent of electricity, questions about how AI might disrupt and destroy humanity are front and centre. Many of the ethical issues, for example to do with AI in warfare, or avoiding bias, urgently need to be addressed. But there is also much critical thinking that we haven't yet explored. Alongside regulations around training machines in a smarter and safer way, we need to train humans how to interact with AI, how to be curious, how to be cynical, and how to stay in control.