Orca teaches AIs to behave fundamentally differently than traditional deep learning models, including large language models and classic machine learning models. Models built with Orca learn to use both static training data and external, updatable data to create outputs, instead of relying on just the training data and emergent properties.
An Orca-augmented natural language classifier
Because the model learns to use external data sources as a core component in the model, you can update, change and add data. The model immediately produces new outputs - unlocking zero shot learning - and you can observe what data is most impactful within your model. For more technical information about how Orca works, please download our white paper ("Orca: Enterprise-Scale Augmentation and Intervention Methods for Neural Network Based Models").