A new artificial intelligence model has been used to help predict the estimated time of death for six million people.
The Technical University of Denmark has created the model based on a number of factors and applied it to the entire population of the country.
It shows that if large amounts of data about people's lives are used to process language, they can systematically organise the data and predict what will happen in a person's life - and even estimate their time of death.
Sune Lehmann, professor at DTU and lead author of the article, told Moncrieff it is similar to a model used for Chat GPT.
"The machine that we built can predict more or less anything... but one of the things that we looked at was early death," he said.
"The idea is that we've seen this incredible development in language technology - you and your listeners have probably played with Chat GPT and found that it's quite amazing.
"What these things do is that they look at sentences and language as sequences of words.
"What we've done, and what's new here, is that instead of sequences of words we fed in sequences of all the strange and wonderful that happen in people lives and built a kind of similar technology".
'Events in people's lives'
Mr Lehmann said the data looks at stand-out events in people's lives.
"We learn in general about what events happened in human lives, how they're related to one another," he said.
"We built this algorithm that can predict all kinds of different things and we said one things that's been well studied and that people worked on for a lot of time is this idea of when will people die?
"Honestly we don't know; we're trying to interpret what the model is doing.
"The model is learning about how the events in human lives play into something like an outcome like death, and then we see how well it can be predicted".
Accuracy
In terms of accuracy, Mr Lehmann said the research looks the patterns from the data and then a subset is chosen where half the people die and the other half live.
"If you are just guessing at random it would be 50/50," he said.
"In that case the algorithm - which looks at data from 2008 to 2016 - can then say out of 100, it'll get 79 right essentially.
"It's much better than random but it's not incredibly accurate," he added.
Certain markers also point to a greater propensity of early death.
People in a leadership position or with a high income are more likely to survive, while being man, skilled or having a mental diagnosis is associated with a higher risk of dying.
Listen back here:
