OpenAI released a 98-page technical report on Tuesday to accompany its unveiling of its latest large language model, GPT-4. Among the hype surrounding the model’s new capabilities, such as its ability to pass the bar exam , is growing criticism from AI researchers who point out that the paper is not transparent or "open" in any meaningful way.
The report, whose sole author is listed as the company rather than specific researchers, explicitly says, “Given both the competitive landscape and the safety implications of large-scale models like GPT-4, this report contains no further details about the architecture (including model size), hardware, training compute, dataset construction, training method."
This means that OpenAI did not disclose what it used to train the model or how it trained the model, including the energy costs and hardware used for it, making GPT-4 the company’s most secretive release thus far. As Motherboard has noted before, this is a complete 180 from OpenAI's founding principles as a nonprofit, open-source entity.