Artificial intelligence shows limitations in life prediction and gaming

Recent developments in artificial intelligence highlight both its limitations in social sciences and its experimental applications in entertainment.

A study led by researchers from Princeton University, published in PNAS, found that machine learning techniques currently fail to accurately predict human life trajectories. Using a database involving 5,000 children and their families, researchers invited 68 institutions globally to attempt predicting outcomes such as school grades, character traits, and financial stability. Despite using advanced sociological models and massive datasets, the AI predictions were found to be inaccurate when compared to actual results. Researchers noted that this failure suggests either the significant role of luck in human lives or missing variables in current social science theories.

In the gaming sector, the generative AI revolution is driving experimental titles like ‘Teach My Little Sister How To Drive’ by Easy Fox on Steam. Unlike traditional games, this title uses a chatbot to allow players to give driving instructions via voice commands. While the concept offers a creative use of generative AI compared to fully AI-generated assets, developers face significant financial challenges, with some AI-driven projects incurring high daily operational costs.