During its first developer conference in November, OpenAI introduced the custom GPTs and store, which were set to be launched later that month.
The company is continuing to “make improvements” to GPTs based on customer feedback, the memo said.
The delay comes against the backdrop of the startup’s surprise ouster of its CEO Sam Altman and his subsequent reinstatement following threats by employees to quit.
The GPTs are early versions of AI assistants that perform real-world tasks such as booking flights on behalf of a user. It is also expected to allow users to share their GPTs and earn money based on the number of users.
Last month, OpenAI announced it intends to work with organisations to produce public and private datasets for training artificial intelligence (AI) models.
Popular chatbot ChatGPT, which can generate poems and prose from simple prompts, is based on large language models that are trained entirely on open-source data available on the Internet.
The company’s latest effort could help it produce more nuanced training data that are more conversational in style.
“We’re particularly looking for data that expresses human intention, across any language, topic and format,” the company said in a blog post.
OpenAI said it is seeking partners to help it create an open-source dataset for training language models. This dataset would be public for anyone to use in AI model training, it said.
The company said it is also preparing private datasets for training proprietary AI models.
© Thomson Reuters 2023