It also doesn’t matter how much data we feed the AI.

A study attempted to predict the future of children. Social scientists from Princeton ran an experiment with 160 teams to find out if a method existed to forecast how a child would do in later life.

What would be the child’s grade point average score? Could the family be evicted from their home?

Each team had access to 15 years of data and was free to use any predictive tool they thought fit. Result: No team turned out even remotely successful. (TNW)

Princeton study: AI can predict trends, not the future

Advanced math, neural networks, machine learning, superfast computers, cutting edge technology, and highly qualified researchers. None were able to predict how a child’s future would turn out.

“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,” said the research paper by the Princeton researchers.

The limitations of AI immediately become apparent. It has no crystal ball, or predictive abilities, though it can extrapolate data to produce likely trends and gather insights into a business that would otherwise not be possible for humans to glean.

Is predictive AI so much hocus-pocus?

“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,” said the research paper by the Princeton team.

The results are useful for authorities looking to use predictive AI models for applications in criminal justice and child protection.  They must consider the results thrown up in the study and the associated concerns.

Related Story:   Artificial Intelligence: Now AI Can Unlock Secrets From 2,500-Year Old Iranian Tablets