How Can Humans Best Use AI?

Often a little stress can sharpen the mind. A recent journey, by train, from Paris to Oxford was disrupted by first a cancelled train and then predictably, a delayed one. This complicated an otherwise pleasant day because I was supposed to be sitting in front of my laptop participating in the aperture 4X4 discussion forum on AI (artificial intelligence). Instead, I found myself nearly hanging out of the window of the train trying to get good phone reception as I spoke at the forum.

In order to compensate for the poor connection I felt obliged to say something colourful and interesting, and thus put forward the view that the best comparison for understanding how humanity can use AI is the tv programme ‘One Man and his Dog’.

One Man and his Dog

One Man and his Dog was a very popular, though quirky, BBC programme based on sheepdog trials across Great Britain and Ireland, which at its peak in the 1980’s had some 8 million viewers (still running on BBC Alba). In very simple terms it is a sheepdog trial, with farmers herding sheep with the help of their sheep dog, or in technical terms, humans performing a complex task, under pressure, with the aid of a trained, intelligent non-human.

While the comparison of AI with ‘One Man and his Dog’ was initially speculative, the more I think about it the more I consider it apt as a framework to understand how humans should use AI. I have not herded sheep, but imagine it can be as or more difficult as sorting data, as unlike data sheep have minds of their own. The combination of (wo)man and dog as a very productive team illustrates how the best uses of AI are beginning to emerge – by doctors, soldiers and scientists deploying AI to second guess and bolster their own decision making.

In addition, like AI, dogs can be trained to attack and defend, but while dogs make valuable companions I struggle to see how AI/robots can fulfil this function. There is a persuasive argument of how this could happen in book The LoveMakers, and in the behaviour of many people who find the metaverse an appealing place to ‘live’ (I am worried by the appearance of the LOVOT VOT family robot in Japan and by the growing use of the AI relationship app Replika).

While dogs can sense our emotions and perhaps intuit what we are thinking, the increasingly alarming aspect of artificial intelligence is that it can determine what we are thinking. A recent edition of Nature journal described how AI can be used to analyse human brain activity, and translate this accurately into words and images.

If the analogy of sheepdogs and AI is less eccentric than readers might have initially thought, it does I hope highlight the need for society to have frameworks and rules of thumb to parse the use and impact of AI.

Economically, AI is already leading to a repricing of the role of people, like software engineers, whom it can replace, but also to a reappraisal to the training and role of those who can use it to be more productive. I suspect that such is the rate of deployment of AI that in time many of the products and solutions it creates will quickly become commoditised.

Regulate AI

The practical aspects of this phenomenon are gathering speed – in the last week alone KPMG has announced a partnership with Microsoft MSFT to drive the use of AI in its businesses and the role of Palantir on the side of the Ukrainian military is becoming more clear. Also, actors and screenwriters in Hollywood are striking at the prospect that some of their work could be replaced by AI.

In addition in China, measures have been announced to control the use of generative AI by requiring firms producing these tools to be licensed by the government where these tools are targeted at ‘the general public’.

This move is consistent with my broad thesis that the US, EU and China will increasingly tackle new trends (notably technologies) in very different ways. America has the AI stock bubble (see Nvidia), the EU has its recent AI Act and now China is controlling the production points of generative AI.

In the US, regulators are beginning to catch up with international counterparts. The Federal Trade Commission is investigating whether OpenAI’s ChatGPT produces false information. More broadly, the OECD has warned of the negative effects of AI on labour markets.

My view is that if they want to see how humans and robots should work together, One Man and his Dog is a good place to start.