The potential for AI is indisputable, but how much would you trust it in estimating when your death would be?
Though it has just been over a year since ChatGPT launched, over 100 million people use the site weekly; but how far could the technology go?
Listen to the episode of The Briefing here:
Could AI accurately predict the future? Some Danish researchers believe that it already can.
In a published study, they’ve taken the principles of large language models, like Chat GPT, and applied it to data of Denmark’s six million people, predicting with 78 per cent accuracy aspects such as personality and behaviour, and even the likelihood of death within a given timeframe.
In today’s afternoon episode of The Briefing, Bension Siebert speaks with lead author and researcher, Professor Sune Lehmann, about what the study found, and what it could mean for the future.
Professor Lehmann described the technology as a “foundation model” which meant it had the capability of predicting anything you have training data for.
“We had a lot of Danes who had filled out personality questionnaires so we could train on them and learn the patterns of their answers, and then predict… what they would answer on their questionnaires.
“We can do that quite well and better than sort of competing methods, even methods running on neural networks and then we looked at predicting early mortality.
“You can predict personality or early mortality; you could also predict probability of someone committing a crime.”
To hear the full chat with Professor Lehmann, listen to today’s episode of The Briefing.