OpenAI begins training of GPT-4 successor, forms safety board

OpenAI has initiated the training of its next AI model and established a new safety and security board, the company announced in a blog post on Tuesday. This rare glimpse into OpenAI’s future plans hints at the eagerly anticipated successor to GPT-4, the technology behind its popular chatbot, ChatGPT, as tech giants compete to develop increasingly sophisticated AI tools, according to Forbes.

OpenAI revealed that it has begun training its “next frontier model,” which is expected to deliver advanced capabilities in line with the company’s mission to achieve artificial general intelligence (AGI). AGI represents a form of AI that can perform a wide range of tasks at a human-like level.

Also Read: OpenAI Board of Directors fires CEO Sam Altman

This new model is poised to replace OpenAI’s leading foundational model, GPT-4, and will power a variety of generative AI tools including image generators, virtual assistants, search engines, and its flagship chatbot, ChatGPT.

To ensure the safe development and deployment of this new technology, OpenAI has created a committee to evaluate potential risks. This committee will be jointly led by CEO and co-founder Sam Altman, and three board members: chair Bret Taylor, Adam D’Angelo, and Nicole Seligman. It will also include six of OpenAI’s technical and policy experts: Aleksander Madry (Head of Preparedness), Lilian Weng (Head of Safety Systems), John Schulman (Head of Alignment Science), Matt Knight (Head of Security), and Jakub Pachocki (Chief Scientist). Additionally, the company will consult with other safety, security, and technical experts as needed.

The committee’s first task will be to review and enhance the company’s processes and safeguards over the next 90 days, after which it will present its recommendations to the full board.

Although OpenAI has not explicitly named the model it is currently training, the company has historically followed a numerical naming convention for its foundational models, with its latest being GPT-4. Consequently, the next model is widely anticipated to be named GPT-5. This new model is expected to be released in the coming months, amidst increasing competition from rival models such as Meta’s Llama 3, Google’s Gemini, and Anthropic’s Claude 3.

Also Read: Ghana Opens Kotoka International Airport, Charges $150 “Covid-19 Fee”

OpenAI did not specify a release date for the new model, noting that training AI systems can span months or even years. Recently, the company updated GPT-4 to GPT-4o, introducing features like voice mode, enabling it to function as a digital assistant similar to Amazon’s Alexa or Apple’s Siri. This update faced criticism for using a voice resembling actress Scarlett Johansson.

While OpenAI appears transparent with its announcement, there is ambiguity regarding how much of the new committee’s findings will be shared with the public. OpenAI stated it will “publicly share an update on adopted recommendations in a manner that is consistent with safety and security” once the committee reports to the board but did not provide a specific timeframe beyond the initial 90-day period. The company’s commitment to sharing the recommendations it adopts leaves it unclear if the full report or all recommendations will be disclosed, especially if there are disagreements over what to implement.

Given OpenAI’s declared commitment to safety and recent internal disputes and departures among key safety leaders, the company and its new committee are likely to face close scrutiny from the AI safety community.

Subscribe to Notifications