- Life2vec predicts future outcomes
- Utilizes chatbot-like functionality
- Achieves 78% accuracy rate
Scientists have devised an algorithm that forecasts an individual’s future existence and death based on the narrative of their life.
A recent study reveals that the ‘life2vec’ model achieves an accuracy rate of approximately 78%, placing it on par with alternative algorithms specifically engineered to forecast analogous life outcomes.
In contrast to other models, however, it functions as a chatbot by predicting future events based on current information.
It was developed by scientists in Denmark and the United States who fed a machine learning algorithm a vast collection of Danish data, including income, occupation, place of residence, injuries, and pregnancy history, among other information about over six million actual individuals.
Their final product was a model capable of generating predictions regarding a person’s lifetime income or the probability that they will pass away prematurely using simple language.
Certain risk factors may contribute to an increased premature mortality rate. These include masculine gender, mental health conditions, and occupation of a skilled nature. Contributing factors to an extended lifespan encompass higher socioeconomic status and assuming a leadership position.
life2vec prophesies the future of your life by analyzing each aspect of it as if it were a single word in a sentence, utilizing the information that has been accumulated thus far.
In the same way that ChatGPT users instruct it to compose an essay, poem, or song. Scientists can ask life2vec straightforward queries such as “Will you die within the next four years?”
Training of the model spanned the years 2008 to 2016.
Over three-quarters of the time, it accurately predicted who had passed away by 2020 using their population data.
However, to safeguard the privacy of the individuals whose data was utilized in the system’s training process, neither the general public nor businesses have access to it, Sune Lehmann, the system’s chief researcher, said.
“Your path to wealth begins here – don’t wait, get your free Webull shares.”
Academic Lehmann, a professor of networks and complex systems at the Technical University of Denmark, stated, “We are actively working on ways to share some of the results more openly; however, this requires additional research to ensure the confidentiality of the study participants.”
Danish privacy rules restrict life2vec from making hiring or insurance policy decisions about individuals. This restriction applies even when the model is eventually made public.
Similar to how ChatGPT and other extensive language models have been trained using vast collections of pre-existing written material, life2vec was instructed using data derived from individuals’ daily lives, represented as a sequence of sentences abundant in relevant information.
These encompass statements such as “Francois obtained 20,000 Danish kroner in September 2012 for his service as a guard at a castle in Elsinore” or “Hermione attended five elective courses during her third year of secondary boarding school.”
Using unique tokens, Lehmann and his team mapped information.
The categories present in individuals’ life stories encompass a broad spectrum of human experiences. For instance, an IND4726 code denotes working in a tobacco establishment, while a ‘postpartum hemorrhage’ is designated as category O72. Income is represented by one hundred unique digital tokens.
Many of these relationships, such as that between occupation and income (certain careers pay more) are intuitive.
However, life2vec maps the vast constellation of components that comprise an individual’s life, enabling a user to request a prediction that takes into account the lives of millions of other individuals and countless other factors.
It is also capable of predicting the personalities of individuals.
Lehmann and his team accomplished this by training the model to predict the responses of individuals on a personality test.
Participants are requested to assess ten statements according to their level of agreement. Examples of such statements include, “When I first arrive at a new location, I immediately make friends,” and “I seldom voice my opinions in group discussions.”
Lehmann highlighted that the data were exclusively from Denmark, therefore these projections may not apply to other countries. Furthermore, it is highly improbable that the majority of people are interested in their mortality date.
Lehmann told Newswise, “The model introduces significant positive and negative perspectives for political discourse and resolution.”
Presently, tech companies employ comparable technologies to forecast human behavior and life events. For instance, these firms track our social media activity, create exact profiles of us, and use them to influence us.
“This discourse must be incorporated into the democratic dialogue. We may ponder the future of technology and determine if it’s a desired development.”
Doctors label government ‘irrational,’ open to talks on strikes
