OpenAI is training a GPT-4 successor (GPT-5?) and establish a safety committee

OpenAI announced that it has begun training a new AI model made to surpass its current GPT-4 technology. This new model is expected to bring big advancements in AI capabilities and propel the company further on its path to achieving artificial general intelligence (AGI). Could it be GPT-5, which is said to be “materially better” and “better at everything“?

On a similar note, when asked about AGI by Elon Musk, he said it’ll be out next year.

OpenAI is also forming a new Safety and Security Committee to address the potential risks associated with future AI developments. The committee will evaluate and refine OpenAI’s safety protocols over the next three months and make recommendations to the company’s board.

OpenAI’s announcement comes amid ongoing debates regarding the safety and ethics of AI development. The company highlighted its commitment to responsible AI advancement while acknowledging the importance of open discussion on these critical issues.

The training process for AI models can take months or even years. Following the training phase, additional time is required for testing and refinement before a model is publicly released. This is a clear example of the fact that OpenAI’s new model might not be available for at least another year.

The formation of the Safety and Security Committee comes after a period of internal turmoil at OpenAI. Ilya Sutskever, a co-founder and safety leader, recently left the company. John Schulman, another co-founder, will now head the safety research efforts.

More here.