No way of knowing…

primary_care_training_NHS_GPs_Artificial_Intelligence

At the Consumer Electronics Show in Las Vegas, in January, the world’s big-brain-bobble-hats and corporations assembled to show-off their stuff.

This year the talk was all about Artificial Intelligence. Except most of it wasn’t AI.

The fact that you can tell Alexa to flush the loo, from a beach in Malibu isn’t AI. It’s a voice activated on-off switch.

Industry is keen to wrap itself in the AI flag.

Samsung have an AI, 98-inch TV with a 8k screen. Actually it’s a piece of software that doubles up from the normal 4k.

Nothing’s new. Adrienne Mayor in her book ‘Gods and Robots’; the concept of robots and automatons has been talked about as early as the Greeks. Creatures made, not born. I like the idea of Greek, geeks!

What is AI? A camera that takes a light reading, notes the film speed, sets the aperture and shutter speed and lets you press the button, is clever but not AI.

Most AI is machine learning, which is just jargon for being able to process a lot of data, quickly. A machine that looks at a scan and suggests a diagnosis… is what?

If you want a machine to tell you that a mass is likely to be a cancer you have to show the machine a lot of pictures of cancers.

A machine would be able to remember more images than a doctor might see in a lifetime.

Does that mean a machine diagnosis might be safer than medic? Ooooh, tricky.

I get the artificial bit… meaning it’s not human. But what about intelligence?

Remembering stuff is easily for a computer. Probably more reliable than the human memory? The rest will be good engineering. Because a computer can give you an answer doesn’t make it intelligent.

Google is very smart, overlooks your spelling errors, gets a feel for what you are looking for and gives you millions of answers in moments. Intelligent? No, it’s a gigantic data base.

Amazon remembers you bought a guide book to the West Indies and then sends you stuff about suncream, car hire and a mankini.

The Berkeley philosophy professor, Hubert Dreyfus, described three fallacies about AI.

First the biological assumption that neutrons are either on or off and therefore, can be replicated on a binary based computer system. Not really.

Second the psychological assumption that the mind works like a rule based machine. Err, no.

The third, that the machine can ‘know’ things. There’s a difference between knowledge and knowing.

So, none of this is true. Where do we go from here?

True artificial intelligence is elusive and it is the mystery that it creates that is part of its weakness.

Can we put our faith in a machine to make a diagnosis of our ills. Certainly. It is no different from a doctor learning what symptoms are, remembering them and matching them to what we tell them.

Remembering, that’s the key. Remembering when it’s right and remembering when it’s wrong… that’s the tricky bit.

Recently I flew Virgin, into a fogbound Gatwick. The Virgin fleet have software that is cleared to land in fog. No human involvement required, other than switching it on.

We landed faultlessly, and through customs and the baggage malarky in record time. The rest of the fleets were unable to land. Gatwick was all but deserted.

Was that AI, or machine learning, or a bit of software that knows speed, height, weight, crosswind and the other stuff pilots have to learn and remember.

All this is important because the workforce shortages the NHS is facing plus the gathering demand, means we will have to have a conversation about what computers with smart software can do, to replace humans in screening, diagnosis, image-analysis, prescribing, outpatients and goodness knows what else.

The bar for intelligence and what we can call smart is high.

AI can beat humans playing chess, recognising images, sort fruit, diagnose cancer and tell, by looking at your face, if you’re miserable.

Will it make mistakes? Yes, but fewer than most of us and that includes doctors.

How many mistakes will some of the best GPs make today? We have no way of knowing.

Contact Roy – please use this e-address – roy.lilley
Know something I don’t – email me in confidence.
Leaving the NHS, changing jobs – you don’t have to say goodbye to us! You can update your Email Address from the link you’ll find right at the bottom of the page, and we’ll keep mailing.
GDPR
We don’t sell or give access to your email address to any third parties.
You can unsubscribe at any time.
Disclaimer

Reproduced at TrainingPrimaryCare.com by kind permission of Roy Lilley.