OpenAI is reportedly focused on developing new strategies to address the recent slowdown in AI development.

According to a report published in The Information, the artificial intelligence, the company’s next big model, codenamed “Orion”, is just as fast as previous versions. may not offer a big leap It is stated.

The Orion model could not make the expected jump

Employees who tested the Orion model say that while the performance of this new artificial intelligence model exceeds OpenAI’s existing models, it does not reach the level of the quantum leap made in the transition from GPT-3 to GPT-4 . In other words, a decrease in progress rate has been observed compared to previous models. So much so that in some areas, especially specific skills like coding, It’s also worth noting that Orion doesn’t offer a significant advantage over older models. It is stated.

In light of this perceived slowdown, OpenAI is developing new strategies to continue the development of artificial intelligence. A dedicated team set up by the company is exploring ways to improve model performance despite dwindling data sources.

OpenAI is moving to synthetic data

According to the report, these strategies include using synthetic data produced by artificial intelligence in Orion’s training process, as well as improving the model in post-training processes. This use of synthetic datacan be an important step for the development of new models at a time when available natural data is decreasing.

Sources reached by OpenAI were unable to get a response from the company on the matter, but the company had previously stated that they had no plans to launch the model called Orion this year.

So what do you think about this issue? You can share your thoughts with us in the comments section below.