Earlier this week, I was in a meeting where someone was adamant they could spot “anything” produced by artificial intelligence. They were highly confident in their ability to detect material written by AI. The evidence, though, suggests otherwise. Several studies have investigated whether people can detect AI-generated material, and overall, we are nowhere near as good as we think we are. One study from Cornell University suggested our ability to detect AI material was as “good as a coin toss”. Another group of researchers found that the human ability to spot AI-generated material was “indistinguishable from chance”. And two years ago, in an analysis of 15m academic research papers, almost one in every seven involved AI, undetected, it seems, by the editors. Ultimately, you have no idea whether I am writing this or a chatbot is. (Honest, it is me…!)
So, why do people insist they can spot AI material? It turns out this is all about our fears of this technology. A survey conducted by the HR technology company GCheck found that almost two-thirds of people exaggerate their artificial intelligence abilities. The study showed that one of the key reasons for this level of “fibbing” is our concern that AI will take our jobs. When someone says to you, “I can spot anything produced by AI”, what they might be saying is, “I am not sure I can spot anything produced by AI, but if I admit that, I could lose my job to AI”.
A recent study published in the Harvard Business Review adds to this likelihood, noting that as AI use increases, anxiety about “surrendering” to artificial intelligence is rising. Indeed, a survey earlier this year by the renowned Pew Research Center found that only 10% of people are excited about AI and its potential. The rest are wary and concerned.
To compensate, we start to exaggerate. The GCheck research found that 40% of people speak confidently about AI during meetings simply to avoid appearing to be left behind. The study also revealed that a quarter of people have taken credit for work which AI did for them. And almost one in five actually admit to lying about their AI experience. People tell me all the time that they don’t trust AI, but this research suggests we shouldn’t trust humans.
Artificial intelligence is advancing at such a pace that we need to start being honest with one another about it. First, it is not going away. We need to accept it for what it is and not complain that it shouldn’t do what it is doing or that AI companies ought to be developing things differently. We are where we are. Let’s move on. The question is no longer whether AI should have arrived. The question is whether we are prepared to use it honestly and intelligently.
Next, we need to get to grips with the technology and avoid telling fibs about our understanding. A good place to start is Google’s AI Essentials course, available on Coursera. Microsoft has a similar course on its website. You can find out about other courses at the UK Government’s AI Skills Hub.
With more understanding about AI under your belt, it’s then essential to keep up with the constant change. The website “There’s an AI for that” has a daily newsletter which provides constant updates. Another daily newsletter is “The Deep View”, which is well worth getting. Or you can subscribe to the podcast. You might think “oh no, I don’t need daily newsletters”. However, this is such an advancing field that unless you keep up to date regularly, you can easily fall behind. And all that will do is cause you to start exaggerating your knowledge and understanding, as you feel further left behind.
Three weeks ago, the fledgling AI company Recursive Superintelligence raised $650m in funding and was valued at $4.65bn. Their idea is to produce artificial intelligence that can improve itself and create new AI software without human intervention. Just 20 days later, yesterday morning, Anthropic, the company behind the AI software Claude, announced that the idea of self-developing artificial intelligence was closer than we think. Within the space of a month, you have one company saying they want to investigate this and another saying it is almost a reality anyway. That’s how fast things are moving.
If you don’t keep up, you will be left behind. Otherwise, you will need to become very good at telling very believable fibs. And frankly, that sounds like harder work than learning how to use AI properly.
