Hot Topic: Who do you trust – man or machine?

Last month I went to Copenhagen for the LSX Nordic conference, which was great.

The metro ride from the airport to the city centre in a driverless train was a novelty for me, as tube trains on the London Underground all have drivers. I thought no more about it.

What if, though, the aeroplane in which I was due to fly had been pilot-less, and had been a drone flown purely by computer?

I’d certainly have thought twice about that – even though I know the technology exists, and rationally, it probably would have been just as safe.

So, what about medicine and pharma?

At the ASCO cancer conference in Chicago, researchers from Dell Medical School at The University of Texas at Austin have just presented an interesting study.

When cancer patients were asked who they’d trust most to answer a serious question about the future of their disease –  their cancer doctor or Chat GPT – 97.3% said their oncologist. None said Chat GPT. The other 2.7% were unsure.

You might say, given all those stories about AIs hallucinating, the result was unsurprising.

But I think it reveals something more fundamental than that. It’s about trust.

And right now, we are often reluctant to trust computers when it comes to some really important things.

Another example of this is a recent poll of clinical trial experts by The Pistoia Alliance, which found half of them cited trust and regulatory uncertainty as major barriers to AI being adopted.

Becky Upton, President of the Alliance, said: “Speed without control is not enough when patient safety is at stake.”

So, part of this is because we are unused to a computer carrying out a particular job, and are understandably nervous about it. AI is a bit like an untrained, untested teenager.

However, it is also because we recognise that people, especially highly trained professionals, have real-world experience that enables them to provide much-needed oversight, and to spot atypical situations.

People can also put things in context, and relate to other people, in a way that machines simply can’t.  Those qualities can be critical – they can be the difference between success or failure, and even between life and death.

Humans are not perfect of course. Pilot error can be fatal, as can a doctor’s mistake.

The balance is also changing: more and more people are happy to be driven by a driverless taxi, for instance.

And maybe “man or machine” is a false dichotomy. Today, a pilot’s job is – arguably – largely to oversee the work of a computer. Medicine is beginning to move in the same direction.  It’s man and machine, not man or machine.

But it’s worth reflecting that we are still a long way from outsourcing critical decisions entirely to machines, and maybe there are good reasons for that.