Companies are reportedly finding it increasingly difficult to forecast their AI spending.

Complete the form to unlock this article and enjoy unlimited free access to all PYMNTS content — no additional logins required.

That’s according to a report Monday (Oct. 5) by The Wall Street Journal (WSJ), which cites one recent study showing that just 11% of almost 400 surveyed businesses could accurately project their artificial intelligence (AI) costs.

The report notes that AI differs from standard software in that it “behaves more like a human worker,” taking actions and making decisions and mistakes, all on the clock.

Although more advanced models typically cost more per token, the WSJ said, they can sometimes carry out tasks more efficiently, which translates to lower costs.

Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!

While more advanced AI models tend to cost more per token, they can sometimes perform tasks more efficiently, leading to lower overall costs. At the same time, asking a cheaper model to do something it isn’t designed to handle could lead to a token run-up, the report said.

This idea was tested earlier this year by researchers from Stanford University, Carnegie Mellon University, and the University of California, Berkeley, as well as Microsoft Research. They had models conduct more than 6,800 tasks in areas such as math, programming and science. In 32% of scenarios, lower-priced models actually cost more than their higher-priced counterparts.

“The practical takeaway is clear,” said Lingjiao Chen , one of the researchers.

Source link