Snorkel AI, the data-centric AI platform company, today showcased data-centric core model development for enterprises to unlock complex and performance-critical use cases with GPT-3, RoBERTa, T5 and other basic models. With this release, enterprise data science and machine learning teams can overcome adaptation and deployment challenges by creating large, domain-specific datasets to refine baseline models and using them to create smaller, specialized models deployable within governance and cost constraints. New features for the development of basic data-centric models are available in Snorkel Flow, the company’s flagship platform, in preview.
Basic models such as GPT-3, DALL-E-2, Stable Diffusion, etc. show great promise for generative, creative and exploratory tasks. But companies are still a long way from deploying basic models in production for complex, performance-critical NLP and other automation use cases. Companies need large volumes of domain-specific and task-specific labeled training data to adapt core models for domain-specific use. Creating these high-quality training datasets with traditional manual data labeling approaches is extremely slow and expensive. Additionally, baseline models are incredibly expensive to develop and maintain and pose governance constraints when deploying to production.
These challenges must be addressed before companies can reap the benefits of core models. Snorkel Flow’s data-centric core management development is a new paradigm for enterprise AI/ML teams to overcome the adaptation and deployment challenges that currently prevent them from using core models to accelerate the development of AI.
Using early versions of Data-centric Foundation Management Development, AI/ML teams built and deployed highly accurate NLP applications in days:
- A major US bank improved accuracy from 25.5% to 91.5% when extracting information from complex contracts spanning hundreds of pages.
- A global homewares e-commerce company improved accuracy by 7-22% when classifying products from descriptions and cut development time from four weeks to one day.
- Pixability distilled the knowledge from the base models and built smaller classification models with over 90% accuracy in days.
- The Snorkel AI research team and its partners at Stanford University and Brown University achieved the same quality as a fine-tuned GPT-3 model with a model that was over 1000 times smaller on LEDGAR, a 100 class complex legal reference task.
“With more than 3 million videos being created daily on Youtube, we need to continuously and accurately categorize millions of videos to help brands place their ads appropriately and optimize their performance,” said Jackie Swansburg Paulino, Chief of product at Pixability. “With Snorkel Flow, we can apply data-centric workflows to extract knowledge from base models and build high-cardinality classification models with over 90% accuracy in days.”
Enterprise Foundation Model Management Suite features include:
- Development of the foundation model to create large, domain-specific training datasets to refine and adapt baseline models for enterprise use cases with production-level accuracy.
- Hot start foundation model use base models and state-of-the-art learning to automatically label training data at the push of a button to train deployable models.
- Foundation Model Prompt Builder develop, evaluate, and combine prompts to adjust and correct the output of base models to accurately label datasets and train deployable models.
“Enterprises have struggled to harness the power of base models such as GPT-3 and DALL-E due to fundamental adaptation and deployment challenges. To work in real-world enterprise use cases, the base models must be scaled using task-specific training data and must address key deployment challenges related to cost and governance,” said Alex Ratner, CEO and co-founder of Snorkel AI. “Snorkel Flow’s unique data-centric approach provides the necessary bridge between core models and enterprise AI, solving adaptation and deployment challenges so companies can derive real value from data models. base.”
Sign up for the free insideBIGDATA newsletter.
Join us on Twitter: https://twitter.com/InsideBigData1
Join us on LinkedIn: https://www.linkedin.com/company/insidebigdata/
Join us on Facebook: https://www.facebook.com/insideBIGDATANOW
#Snorkel #Accelerates #Core #Model #Adoption #DataCentric #insideBIGDATA