• Sam Altman appears to have responded to concerns that new AI models are hitting a performance wall.

• OpenAI’s CEO said Thursday that there are “no walls” in X.

• A recent report said that OpenAI’s next model shows only a moderate improvement over ChatGPT-4.

Sam Altman said Thursday that “there is no wall”; This is an apparent reference to concerns that OpenAI and other companies are facing a slowdown in developing AI models.

His comment published on X, It follows a report in The Information He said OpenAI’s next model shows only a moderate improvement over ChatGPT-4 and a smaller jump over previous versions.

IT raised concerns In the tech world, the techniques that have historically made AI models smarter (more training data and more computing power) appear to have diminishing returns.

Altman, who had a history of publishing short, cryptic messages on X, suggested that these so-called scaling laws were predetermined. He said on the site in February that these were “determined by God” and that “the constants are determined by the technical staff.”

Ilya Sutskever, co-founder of OpenAI and Safe Superintelligence, recently told Reuters that results from increasing pre-training have remained stable. Others, such as Microsoft’s chief technology officer Kevin Scott, dismissed these concerns. “We’re not in diminishing marginal returns on scaling up,” Scott said on Sequoia Capital’s “Education Data” podcast in July.

AI labs are also exploring ways to overcome slowdowns, such as using synthetic data and improving models after they are trained using a technique called inference.

Large sums of money invested in leading AI companies have created pressure to produce increasingly powerful models.

Last month, OpenAI raises $6.6 billion from investors in a record round of financing.

OpenAI did not immediately respond to Business Insider’s request for comment.

Read the original article Business Content