Mostly AI, a pioneer of structured synthetic data, launched its synthetic text functionality, which gives Fortune 500 companies, including Databricks and Amazon Web Services (AMZN), access to a "vast amount of proprietary text" to train and fine-tune large language models, or LLMs -- without compromising user privacy, it said.

On the Mostly AI platform, users can upload original text data, such as emails and transcripts of customer support calls, and choose an open-source language model from Hugging Face to generate the synthetic data. The original data is used to fine-tune the LLM on the Mostly AI platform, which then generates synthetic text that can be downloaded or stored in a database.

"Today, AI training is hitting a plateau as models exhaust public data sources and yield diminishing returns," Tobias Hann, chief executive of Mostly AI, said in a statement. "To harness high-quality, proprietary data, which offers far greater value and potential than the residual public data currently being used, global enterprises must take the leap and leverage both structured and unstructured synthetic data to safely train and deploy forthcoming generative AI solutions."