Upstream and downstream AI: Post-panel reflections
A couple of weeks ago I was kindly invited by talks organiser Altered State to join a panel on Artificial Intelligence (AI), considering whether it is “the end of the world as we know it”.
My fellow panellists were experts on the subject – Colin Gavaghan, Professor of Digital Futures at the University of Bristol, and Nigel Toon, CEO of Graphcore AI. The Hen and Chicken in Bristol has a spacious upstairs studio that made a great venue for the event.
It has been some time since I wrote Everything About You – a novel about someone whose AI assistant knows everything about her and can make better decisions about her life – so this was an excellent opportunity to brush up on my AI knowledge.
What strikes me the most is how far-reaching AI has become. We have no control over the way it is an upstream influence in our lives – in the way our data is managed, in aspects of society that include recruitment, healthcare, the way crime is predicted, policed and prosecuted, and the way new drugs are developed, to name but a few. Many of these have a positive effect – for instance using AI to sift through countless protein structures to find that elusive disease-curing molecule.
When I was originally researching this topic, it became clear that one of the key benefits of AI would be to make sense of the ‘big data’ age. Currently, almost unlimited information can be collected about every aspect of society – and the natural world – but the sheer quantity is so staggering, so terrifyingly vast it becomes almost nonsensical. Human intelligence can only cope with so much, but that is where AI takes over – neural networks chewing through endless figures (and photos, faces, every kind of data) and coming up with conclusions about where the next hurricane will strike, or – less usefully – what kind of advert will make you click ‘buy’.
At the downstream end, there is the sort of thing I wrote about in my novel – the ‘intelligence’ we see in Google Assistant and other virtual helpers, to which so-called Large Language Models (LLMs) such as ChatGPT can be added. These are the ones we are most likely to mistake for colleagues and friends, but they are really just data-fuelled prediction machines, or cobbling together the most likely combinations of words.
Here we get some choice in the matter. We can leave them well alone if we like, but as time goes on I sense that it will be harder to do so, and we’ll eventually have to find a well-judged comfort zone. There will be the option to let AIs plan not only our transport routes but also our meetings, workouts, haircuts, holidays and any other ‘life admin’. And if they do it better than us, and if we have more time for the good things in life, how important will it be to wrest back control?
The choice will be yours – grab the wheel, or sit back, relax, and be a passenger.
Photos: Freestyle Bristol
Check out upcoming Altered State talks
Check out the Citizen’s Guide to AI
