Meta Platforms (META) has released several new artificial intelligence (A.I.) models.
Among the new models is one called the “Self-Taught Evaluator” that’s capable of checking the work of other A.I. for accuracy.
Specifically, the Self-Taught Evaluator can check the accuracy of responses issued by chatbots and other A.I. applications in areas such as science, coding, and mathematics.
Meta says it aims to solve one of the biggest problems with A.I. models such as ChatGPT, which is that they often provide inaccurate or outdated information.
Analysts say the ability to use artificial intelligence to evaluate A.I. is a step towards autonomous A.I. models that operate independent of humans.
Eventually, A.I. models will be able to learn from their own mistakes, self-correct, and improve without any input from human beings, say researchers.
Analysts say this is all part of the move toward “artificial general intelligence,” which is the point at which A.I. matches and then surpasses human intelligence and makes decisions on its own.
Other new A.I. tools released by Meta include an update to the company’s image-identification model and a tool that speeds up A.I. response generation times.
The stock of Meta Platforms is up 67% this year and currently trading at $576.47 U.S. per share.
