How much AI is too much AI?

While it’s pretty amazing we’ve got to the point where artificial intelligence can think critically and come up with solutions out of thin air, it can also be a pain in the behind.  

What happened: The National Eating Disorder Association (NEDA), a US-based non-profit, shut down a chatbot named Tessa which answered questions about eating disorders after the company that programmed it added an unwanted AI component. 

  • Tessa was designed and tested as a closed system, only able to provide pre-approved info from authoritative sources and unable to veer from its script.

Cass, the company running Tessa, added an AI aspect a year after it launched, which led to the bot going rogue and giving dieting tips—something that NEDA did not want it doing.  

Why it matters: Stories like this could become increasingly common. With generative AI seemingly the only tech attracting any interest these days, companies are feeling the pressure to find a way to use it—even if it’s unnecessary, or in this case, harmful. 

  • It’s no secret that AI still makes lots of mistakes—from fabricating court cases when doing legal research to a persistent inability to determine what a gorilla looks like. 

Zoom out: Generally speaking, people don’t like having things foisted upon them—and that’s true of AI, too. A study from the IÉSEG School of Management found that workers who had positive views of AI lost enthusiasm for it when they were forced to use AI technology.—QH