Researchers Develop AI That May Be Able to Predict How You Die

Amaze Lab

Artificial intelligence could change the way we live, but now a new study suggests it just might be able to predict how we die. Experts say it’s a lot like how text prediction models anticipate language, but this one anticipates life events instead.

Transcript

00:00 (upbeat music)

00:02 Artificial intelligence could change the way we live.

00:06 But now a new study suggests

00:08 it might also be able to predict how we die.

00:11 Experts say this sort of thing shouldn't really worry you.

00:13 After all, doctors have been making health predictions

00:15 for years based on an individual's life habits.

00:18 Now an AI might be able to do the same,

00:20 though it can do so much faster

00:22 and likely better than its human counterparts.

00:24 So how does it work?

00:25 The study was conducted by researchers

00:27 at Technical University Denmark

00:29 using data from some six million residents.

00:32 The researchers found that by implementing

00:33 the same techniques as language predicting AI

00:35 like chat GPT, they could also predict

00:38 what might happen to individuals throughout their lives.

00:40 Finding that rather than deciphering

00:42 which word comes next in a sentence,

00:44 their model figures out what events come next in one's life

00:47 given its preceding actions and events.

00:49 And the researchers discovered that their AI model

00:52 called Life2Vec is actually better at predicting

00:54 a cause of death than any other current methods.

00:57 What's more, it was also able to predict

00:59 certain personality features such as extroversion

01:02 based on those actions as well.

01:03 With one of the study's data scientists,

01:05 Sunay Leeman saying about their findings,

01:07 "The whole story of a human life in a way

01:10 can also be thought of as a giant long sentence

01:12 of the many things that can happen to a person."

01:15 (upbeat music)

01:18 (upbeat music)

01:20 (upbeat music)

Category

Transcript

Recommended