March 2023 Summaries

Gretel has introduced Gretel Relational, a solution that enables organizations to generate high-quality synthetic databases while preserving cross-table relationships. This tool leverages generative AI models to synthesize large multi-table databases while maintaining referential integrity and statistical accuracy. It offers over 30 available connectors for popular databases and data warehouses like Oracle, MySQL, MariaDB, Microsoft SQL Server, Snowflake, SQLite, PostgreSQL, and more. The Gretel Relational Report helps assess the accuracy and privacy of synthetic databases as a whole and of individual tables.

This blog demonstrates how to utilize Gretel, an AI-based platform, in conjunction with Google Cloud Platform's (GCP) Vertex AI to create high-quality synthetic tabular data for training a classification model. The partnership between Gretel and GCP aims to accelerate MLOps by enabling the use of synthetic data to augment or replace ML training data. By using Gretel, users can increase the number of training samples and balance their datasets, thus eliminating bias in the data. This tutorial uses Gretel Cloud for easy setup and familiarization with its capabilities, but also offers options for customers to deploy their own Data Plane, known as a Gretel Hybrid Deployment, which allows all data processing to happen within their own GCP account by way of Google Kubernetes Engine (GKE).

Gretel introduces Reinforcement Learning from Privacy Feedback (RLPF), a novel approach to reduce the likelihood of language models leaking private information. RLPF combines reinforcement learning with measures of privacy and uses them as rewards for improving language model capabilities in a multi-task fashion. Preliminary results show that RLPF can improve both privacy preservation and summarization quality, outperforming some existing models. This method has potential applications in reducing biased or discriminatory language in AI systems.

Gretel has partnered with Google Cloud to accelerate the adoption of safer generative AI in enterprises by leveraging synthetic data. The collaboration enables developers to create secure versions of sensitive data from existing Google Cloud storage buckets and deliver them directly to Vertex AI, addressing data compliance concerns and supply constraints. Synthetic data also allows for experimentation with foundation models without exposing raw and sensitive data. Gretel's platform provides important tooling for developers utilizing data and building generative AI models and applications on Google Cloud.