Predicting certain future outcomes can be highly beneficial in life, especially in the policy-making world, in mathematics, in business, among other scenarios.
So when a trio of social scientists from Princeton University in the U.S. set out to determine whether or not AI could also be used to predict the outcome of a child's future, they found out they couldn't even come close.
Using 15 years' worth of data, collaborating with 160 research teams, and using the latest AI tech was still not enough to see the future success of a child's life.
The study was published in the Proceedings of the National Academy of Sciences.
AI simply can't predict the future
AI can certainly predict trends and even offer useful insights to help industries in their decision-making process. However, determining whether a child's life will ultimately be successful is an entirely different kettle of fish that mathematics simply can't predict.
"We investigated this question with a scientific mass collaboration using the common task method; 160 teams built predictive models for six life outcomes using data from the Fragile Families and Child Wellbeing Study, a high-quality birth cohort study," explained the study.
"Despite using a rich dataset and applying machine-learning methods optimized for prediction, the best predictions were not very accurate and were only slightly better than those from a simple benchmark model."
The reason that AI can't predict this information is in big part because machine learning can reach conclusions but can't explain how it got there. In sales figures, this doesn't matter and usually works quite well with the data provided, but when it comes to social matters it's a different story.
When dealing with a person's future life and freedoms these become random guesses, and not very good ones at that according to the Princeton study. Despite having a treasure trove of data to work with from the "Fragile Families" study, the research team's system still could not correctly predict the outcome of each child's life.
As the research paper pointed out: "In other words, even though the Fragile Families data included thousands of variables collected to help scientists understand the lives of these families, participants were not able to make accurate predictions for the holdout cases."
"Further, the best submissions, which often used complex machine-learning methods and had access to thousands of predictor variables, were only somewhat better than the results from a simple benchmark model that used linear regression."