OpenAI launches new GPT-01 model…but with limitations

American artificial intelligence (AI) research organization OpenAI has launched its much-anticipated GPT-01 model which has scored impressively on benchmark tests.

The company released a preview of its next-generation large language models on Friday. The company claims that these models outperform previous models, but there are some limitations.

When OpenAI launched its o1-preview model, it boasted about how well it performed on a variety of human-designed tasks: in the Codeforces programming competition, the model achieved an 89th percentile score, and on an International Mathematical Olympiad qualifying test, it answered 83 percent of questions correctly, compared with 14 percent for GPT-4o.

READ ALSO: Microsoft and Open AI face lawsuits over alleged copyright infringement

”The o1-preview and o1-mini models represent “the beginning of a new paradigm: AI that can do complex general-purpose reasoning,” according to OpenAI CEO Sam Altman. He also noted that “o1 is still limited, has flaws, and looks more impressive when first used than when used more often.”

Additionally, according to OpenAI, intentional hallucinations typically occur when o1-preview is asked to provide references to comparable papers, websites, books, or other sources that are difficult to independently check without access to the internet.

“Intentional hallucinations primarily occur when o1-preview is asked to provide references to articles, websites, books, or similar sources that cannot be easily verified without access to an internet search, causing o1-preview to instead create plausible examples,” the company wrote in its model system card.

The post OpenAI reveals new GPT-01 model…but with limitations first appeared on Latest Nigeria News | Top Stories from Ripples Nigeria.