Study finds that Google's AI features create "unacceptable risks" for children

Multiple Internet giants have found themselves under the scrutiny of lawmakers, regulators and organizations concerned with the safety of children on the web. Google and its AI Search are the latest to draw criticism.

Interestingly, it may have been the success of the Chromebooks that caused this – they are very popular as school laptops. The Common Sense Media organization explains that the thing that separates Google’s AI from all the other AI services is that it is “ubiquitous on children's personal and school-issued devices.”

Who needs to learn math when AI can just solve the problems for you

The organization found several issues. Homework is one – Google’s AI Mode would happily do all the homework that the student was supposed to do. Worse, it would present different answers from one attempt to another, sometimes right, sometimes wrong.

It gets worse from there. According to the organization, Google’s AI Mode “failed kids in crisis, including missing clear signs of suicidal ideation, reinforcing signs of psychosis and mania, validating disordered eating including purging and celebrating cannabis use.”

AI Mode is the chatbot interface and only appears when the user clicks on the “AI Mode” button. There’s also the AI Overview feature, which appears automatically. An important sticking point is that neither feature can be disabled.

You can read the full report on Common Sense Media’s findings here (PDF). In summary, Google’s AI features performed poorly on seven out of eight “established AI Principles” that the organization tests for.

Source | Via