Snorkel AI Accelerates Adoption of Foundation Models With Data-Centric AI – insideBIGDATA | Wonder Mind Kids

Snorkel AI, the data-centric AI platform company, today introduced enterprise data-centric foundation model development to unlock complex, performance-critical use cases with GPT-3, RoBERTa, T5, and other foundation models. With this launch, enterprise data science and machine learning teams can overcome customization and deployment challenges by creating large, domain-specific datasets to fine-tune base models and use them to create smaller, specialized models that are deployed within governance and cost constraints can become. New capabilities for data-centric fundamental model development are in preview in Snorkel Flow, the company’s flagship platform.

Foundation models like GPT-3, DALL-E-2, Stable Diffusion and more show promise for generative, creative and exploratory tasks. But companies are still a long way from deploying base models in production for complex, performance-critical NLP and other automation use cases. Organizations need large amounts of domain- and task-specific labeled training data to customize base models for domain-specific use. Creating these high-quality training datasets using traditional, manual data labeling approaches is painfully slow and expensive. Additionally, Foundation models are incredibly expensive to develop and maintain, and pose governance constraints when deployed to production.

These challenges must be addressed before companies can reap the benefits of foundation models. Snorkel Flow’s Data-centric Foundation Management Development is a new paradigm for enterprise AI/ML teams to overcome the customization and deployment challenges that currently prevent them from using Foundation models to accelerate AI development.

Using early versions of Data-centric Foundation Management Development, AI/ML teams built and deployed highly accurate NLP applications in a matter of days:

  • A leading US bank improved accuracy from 25.5% to 91.5% when extracting information from complex multi-hundred-page contracts.
  • A global housewares e-commerce company improved accuracy when classifying products by description by 7-22% and reduced development time from four weeks to one day.
  • Pixability distilled knowledge from base models and built smaller classification models with over 90% accuracy in a matter of days.
  • The AI ​​research team of Snorkel and partners from Stanford University and Brown University achieved the same quality as a finely tuned GPT-3 model with a model over 1000 times smaller on LEDGAR, a complex Class 100 legal benchmark task .

“With over 3 million videos created on YouTube every day, we need to constantly and accurately categorize millions of videos to help brands place their ads appropriately and maximize performance,” said Jackie Swansburg Paulino, chief product officer at Pixability. “With Snorkel Flow, we can apply data-centric workflows to distill knowledge from base models and build high-cardinality classification models with greater than 90% accuracy in a matter of days.”

Enterprise Foundation Model Management Suite features include:

  • Fine-tuning of the base model to create large, domain-specific training datasets to fine-tune and adapt base models for enterprise use cases with production-grade accuracy.
  • Basic model warm start to use base models and state-of-the-art zero and few-shot learning to automatically label training data with the push of a button to train deployable models.
  • Foundation Model Prompt Builder to develop, evaluate, and combine prompts, to tune and correct the output of base models, to precisely label datasets, and to train deployable models.

“Enterprises are struggling to leverage the power of base models like GPT-3 and DALL-E due to fundamental customization and deployment challenges. To work in real enterprise use cases, base models must be customized using task-specific training data and overcome major deployment challenges in terms of cost and governance,” said Alex Ratner, CEO and co-founder of Snorkel AI. “Snorkel Flow’s unique data-centric approach creates the necessary bridge between base models and enterprise AI, solving the customization and deployment challenges so organizations can derive real value from base models.”

Sign up for the free insideBIGDATA newsletter.

Follow us on Twitter:

Visit us on LinkedIn:

Join us on Facebook:

Leave a Comment