Doom calculator claims it can predict when you will die with 78% accuracy

20 December 2023 , 12:39

789     0

Scientists have come up with an AI that can predict when you will die (Image: Getty Images/iStockphoto)

Predicting how you'll live and when you'll die has been a fascination for many people, but now scientists have developed technology that it claims can do just that.

The experts have built an algorithm that uses the story of a person's life to predict their future and it is accurate about 78 per cent of the time. But unlike other similar tech, it works like a chatbot, using existing details to predict what comes next. According to a new study, the model called "life2vec", is on a par with other models designed to predict similar life outcomes.

American and Danish scientists trained a machine-learning algorithm on a huge range of Danish data, feeding it numerous types of information on more than six million people, including income, profession, place of residence, injuries, and pregnancy history. This resulted in a model that can process plain language and generate predictions about a person's likelihood of dying early, or their income over their lifespan.

According to the study, factors that can lead to a shorter life include being male, having a mental-health diagnosis, or being in a skilled profession, while a longer life can come from having a higher income or being in a leadership role. The system, life2vec, looks at each part of your life as if they were words in a sentence then forecasts where that story will go based on what has been written so far.

The AI can predict a person's life earnings fairly accurately (Getty Images)

The model was trained on data stretching from 2008 to 2016 and similarly to ChatGPT, where users ask it to write a song, poem, or essay, scientists can ask life2vec simple questions such as "death within four years?" for any given person. Shockingly, based on the population data, it correctly predicted who had died by 2020 more than three-quarters of the time.

Greggs, Costa & Pret coffees have 'huge differences in caffeine', says report

Lead researcher Sune Lehmann said that to comply with GDPR and protect the personal information of the people whose data trained the system, it is not available to the general public or companies. He said: "We are actively working on ways to share some of the results more openly, but this requires further research to be done in a way that can guarantee the privacy of the people in the study."

The research appeared in Nature Computational Science and Lehmann, professor of networks and complex systems at Technical University of Denmark explained that in much the same way that ChatGPT and other large language models have been trained on vast expanses of existing written works, life2vec was taught by data from people's lives, written out as a series of href="https://www.mirror.co.uk/all-about/denmark" rel="Follow" target="_self">Danish kroner as a guard at a castle in Elsinore" or "During her third year at secondary boarding school, Hermione followed five elective classes." The scientists then gave each piece of information different tokens and these pieces of data were all mapped out in relation to each other.

Click to follow the Mirror US on Google News to stay up to date with all the latest news, sport and entertainment stories

These categories go across the whole range of human experiences with different codes, for instance, a forearm fracture is represented as S52, income is represented by 100 different digital tokens and "pospartum hemorrhage" is O72. Much of these relationships are intuitive – profession and income – certain jobs make more money.

It can also make predictions about people's personality. To do this, Lehmann and his team trained the algorithm to predict people's answers to questions on a personality test. But, he said: "It's important to note the data were all from Denmark, so these predictions may not hold true for people living in other places. The model opens up important positive and negative perspectives to discuss and address politically."

He warned that any discussion "needs to be part of the democratic conversation so that we consider where technology is taking us and whether this is a development we want". Even when the model is finally available to the public, Danish privacy laws would make it illegal to use life2vec to make decisions about individuals – such as influencing insurance policies or employers deciding who to employ.

Paul Donald