A couple of months ago, I had a surprising conversation with someone I met at a networking event.
I was introducing myself, and mentioned that I’d been working in artificial intelligence and machine learning for nearly 20 years.
“That long?” she asked. “I thought AI was new.”
This really took me by surprise at the time, but I quickly realised it’s me who’s in a bubble here. For a lot of people who aren’t AI researchers, the slew of recent articles about generative AI (such as chatbots and image generation) in the mainstream press might be the first time they’re really seeing artificial intelligence talked about outside of science fiction.
So, I wanted to take this opportunity to highlight a small number of ways you’re probably already using AI-powered apps every day, even if you’ve never tried ChatGPT. In fact, I’ve specifically chosen examples that aren’t “generative AI” (the collective term for techniques that generate novel text, image, or video) to try and show the breadth of the field.
[Image shows two screenshots, one of the Met Office’s weather forecast app, and one showing a Google Maps route. Text is cream on a green background and reads “You’re already using AI every day…”]
When you check the weather forecast
Huge computers in the Met Office are constantly running models to try and predict tomorrow’s weather across the UK. These take in continuous inputs of meteorological data, such as wind speed, pressure fronts, humidity, and the current position of the clouds. Using all this data, the algorithms try to predict what will happen tomorrow, based on thousands of observations of what has previously happened under similar conditions.
Did you notice that weather forecasts got less reliable during the pandemic lockdowns? One of the inputs to weather forecasting is regular data from commercial flights, and when most planes were grounded, there was a lot less data being gathered to inform the models.
When you use satnav for directions
I’m someone who struggles with navigation, so I use Google Maps a lot. Every time you ask for directions from A to B, there are algorithms in the background which consider all the possible routes and generate a recommendation based on predicted speed and distance. These are calculated taking into account the current road conditions and traffic, so you might get a different route on a quiet Sunday afternoon compared to Monday morning rush hour.
Your mapping app also gathers your data (anonymously and aggregated with others) to feed back into the model — for example, when you’re stuck in a traffic jam with a number of other cars, each of which is moving slowly or not at all, that data is used to avoid sending other cars by the same route until the bottleneck clears.
When a streaming service recommends a show to you
While you’re watching Netflix, Netflix is also watching you. (Other streaming services are available!)
You can provide explicit “thumbs up” ratings, and add shows to your list to watch later, but the data you know you’re providing is only the tip of the data iceberg. Streaming services are also recording things like whether you finish a series and how quickly, whether you pause a movie and walk away (to make a cuppa, or for a few days?), which shows you binge addictively and which you rewatch. This data is all used to find a cluster of people like you, analyse what they liked, and push those shows in your direction (and vice versa). Brand new shows can be promoted to people who liked similar genres and themes, while background data processing checks whether recommended shows are subsequently watched, and enjoyed. It has even been spotted that the same show may be promoted with different thumbnails for different audiences, for example showing the female lead to people who tend to watch programs with more women, but highlighting the male lead to other audiences who prefer male-led narratives.
Similar recommendation engines sit behind many online stores, such as “people also bought…” on Amazon. And it’s not just online: supermarket loyalty schemes may send you vouchers for premium products which are regularly bought by the people who bought similar shopping to you, but which their data suggests you haven’t tried. (And don’t think you’re not being tracked just because you don’t have a loyalty card — spending on the same credit card can still be grouped, so unless you always pay in cash, you may as well opt in and get the vouchers.)
