OpenAI CEO Sam Altman recently stated that size will no longer be the defining factor of developing artificial intelligence models, upending the way AI models have been designed over the past few years. According to Altman, developers of AI must now explore “other ways” to make their models better, instead of relying on massive models with plentiful data. The current version of OpenAI’s popular ChatGPT is based on GPT-4, one of the most advanced large language models (LLMs) available.
Although OpenAI’s ChatGPT has been extremely successful, Altman believes that parameter count should no longer be the primary focus of AI models. He likened this change in focus to the race for higher gigahertz in processor chips in the late 90s and early 2000s. This means that advancements in AI will still be made, albeit with a more modest parameter count than what is currently present. For example, GPT-2 had 1.5 billion parameters, while GPT-3 improved this number to 175 billion. However, OpenAI has not revealed GPT-4’s parameter count, showing that the scale of AI models is no longer of primary importance.
It has been speculated that the training of GPT-4 cost approximately $100 million, but Altman recently disclosed that it cost more than that. He did not provide any comments on this, however. The success of ChatGPT has provoked reactions from renowned personalities in the technology industry, including Elon Musk, Steve Wozniak, and Tristan Harris, who signed an open letter urging AI companies to halt their progress for at least six months due to safety concerns. When asked about it at the MIT event, Altman noted that safety governance was critical in the industry, but some of the claims made in the letter were inaccurate, citing that OpenAI is yet to start working on GPT-5.
OpenAI is a San Francisco-based artificial intelligence (AI) research laboratory, co-founded by Elon Musk and Sam Altman. It is a non-profit organization with a mission to ensure that artificial general intelligence (AGI), which is the technological capability of a machine to comprehend or learn any task that a human can, is developed safely and responsibly. The company is well-known for its advances in natural language processing, with the development of their large language models, GPT-2 and GPT-3, as well as their ChatGPT bot, which can be trained to perform human-like tasks such as writing poetry and debugging code. Sam Altman is an American programmer, entrepreneur, investor and executive, famous for co-founding Y Combinator in 2005. He is also the current CEO of OpenAI and previously served as President of Y Combinator.