Scientists have invented an artificial intelligence (AI) chatbot that can predict when you will die with 78% accuracy.
The AI, called Life2vec, can also determine how much money you’ll have when death finally catches up with you.
The model was created by scientists in Denmark and the US, who fed data from Danish health and demographic records for six million people into the model. Like ChatGPT, it learned to predict what would come next based on the words before it.
But unlike ChatGPT, this AI uses information such as income, profession and medical records to determine how long you’ll live, rather than to compose poetry.
Previous research has already determined that certain life factors can lead to longer or shorter lives. For instance, being male, a smoker or a poor mental health diagnosis can all contribute to shorter lifespans. Higher incomes and having a leadership role can lead to a longer life.
Each of these factors has a code in the Danish data set, such as S52 for a broken forearm, or IND4726 for working in a tobacco shop, which the team converted into words.
This enabled people’s lives to be summed up in data-rich sentences, including things like profession, income, injuries and pregnancy history.
‘In September 2012, Francisco received twenty thousand Danish kroner as a guard at a castle in Elsinore’ and ‘During her third year at secondary boarding school, Hermione followed five elective classes’ are examples of the stories written about subjects.
The AI, a large language model (LLM) like ChatGPT and Google’s Bard, then fits all of these words together like a jigsaw puzzle to determine how long the person will live.
The team tested Life2vec on a group of people aged between 35 and 65, half of whom died between 2016 and 2020.
Life2vec predicted who would die and who would live with startling accuracy – 78%.
It was also able to predict the results of personality tests more accurately than models specifically trained for the task.
Fortunately, or not, depending on your point of view, the chatbot is not currently available for use by the public.
The team, led by Professor Sune Lehmann Jørgensen at the Technical University of Denmark, noted that given the model was trained exclusively on data from Denmark, the results may not be as accurate for people from other countries.
They also stressed that the AI must not fall into the hands of big business.
‘Clearly, our model should not be used by an insurance company, because the whole idea of insurance is that, by sharing the lack of knowledge of who is going to be the unlucky person struck by some incident, or death, or losing your backpack, we can kind of share this this burden,’ said Professor Jørgensen.
However, he added that similar technology is already out there.
‘They’re likely being used on us already by big tech companies that have tonnes of data about us, and they’re using it to make predictions about us,’ he said.
And there is one benefit to knowing when you’ll die – postponing it.
Writing in the journal Nature Computational Science, the team adds: ‘Our framework allows researchers to discover potential mechanisms that impact life outcomes as well as the associated possibilities for personalised interventions.’
This article was first published on December 20, 2023
MORE : Scientists are feeding AI with chunks of human brains (sort of)
MORE : 12 months ago ChatGPT became a thing – but just how scared of AI should we be?
MORE : Half of university students are scared their jobs will be under threat from AI
Follow Metro across our social channels, on Facebook, Twitter and Instagram
Share your views in the comments below
