The potential for mapping out someone’s entire life upfront is each exciting and scary. A brand new artificial intelligence trained on the private data of every body in Denmark can do exactly that.
Today’s deep learning-based AI systems are prediction machines. They work by ingesting vast amounts of knowledge and using it to select statistical patterns that may be used to make informed guesses about previously unseen data.
Despite the uncannily fluent linguistic capabilities of AI chatbots, they operate in much the identical way. They learn from huge amounts of text data after which try and predict what word comes next in a string of text.
What allowed the breakthrough in capabilities that we’ve seen in the previous couple of years was a brand new deep learning architecture, generally known as a transformer, that may train on much more data than previous algorithms. It seems that when you possibly can train models on almost everything of the web, their predictions turn out to be very sophisticated.
Now researchers have shown that they’ll use the identical form of techniques to coach a model on an enormous database of health, social, and economic information collected by the Danish government. The resulting AI was capable of make highly accurate predictions about people’s lives, including how likely they’re to die in a given time window and their personality traits.
“The model opens up essential positive and negative perspectives to debate and address politically,” Sune Lehmann from the Technical University of Denmark, who led the study, said in a press release. “Similar technologies for predicting life events and human behavior are already used today inside tech firms that, for instance, track our behavior on social networks, profile us extremely accurately, and use these profiles to predict our behavior and influence us.”
The dataset the researchers used spans from 2008 to 2020 and includes all six million Danes. It features information on their income, job, social advantages, visits to healthcare providers, and disease diagnoses, amongst other things.
Getting the info right into a format a transformer can understand took some work though. They restructured all the knowledge within the database into what they call “life sequences,” with all of the events related to each individual organized in chronological order. This makes it possible to do next-event prediction in much the identical way an AI chatbot does next-word prediction.
When trained on large numbers of those life sequences, the model can start to select patterns that connect disparate events in someone’s life and help it make predictions concerning the future. The researchers trained their model on the life sequences of individuals aged 25 to 70 between the years of 2008 and 2016 after which used it to make predictions concerning the next 4 years.
After they asked it to guess the likelihood of somebody dying in that period, it outperformed the present state-of-the-art by 11 percent. In addition they got the model to make predictions about how people scored on a personality test, and the outcomes outperformed models specifically trained for that task.
While the performance on those two tasks is impressive, in a paper describing the research in Nature Computational Science, the team points out that what’s really exciting concerning the model is the very fact it may possibly potentially be used to make every kind of predictions about people’s lives. Previously, AI has normally been trained to reply specific questions on people’s health or social trajectories.
Obviously, this sort of research raises some thorny questions on privacy and human agency. However the researchers indicate that non-public firms are almost definitely doing similar things with their very own data, so it’s useful to know what these sorts of techniques make possible.
And given the rapidly advancing capabilities of AI, it can be essential to have public debates about what form of AI-powered predictions we allow in each the private and public sphere, says Lehmann.
“I don’t have those answers,” he said in a press release. “Nevertheless it’s high time we start the conversation because what we all know is that detailed prediction about human lives is already happening and immediately there isn’t a conversation and it’s happening behind closed doors.”
