Technology
OpenAI is reportedly developing new strategies to deal with the slowdown in AI improvement
According to new report in Information.
Employees who tested the new model, codenamed Orion, reportedly found that while its performance outperformed existing OpenAI models, the improvement was lower than they noticed after switching from GPT-3 to GPT-4.
In other words, the rate of improvement appears to be slowing. In fact, Orion could be reliably higher than previous models in some areas, reminiscent of encoding.
In response, OpenAI has formed a founding team to determine how the company can proceed to improve its models in the face of a dwindling supply of new training data. These new strategies reportedly include training Orion on synthetic data produced by AI models, in addition to improving the models in the post-training process.
OpenAI didn’t immediately respond to a request for comment. In response to previous reports about plans for its flagship model, the company said: “We have no plans to release a model codenamed Orion this year.”