Te-Lin Wu

Title: Te-Lin Wu and AI Language Model Retraining

Introduction

Te-Lin Wu is an inventor based in Menlo Park, California. Her patent work addresses methods for improving artificial intelligence language models by using feedback about problematic responses.

Latest Patents

Wu is associated with 1 patent, titled “Retraining a Large Language Model.” The invention describes receiving an indication that a response generated by a first AI language model has a problem. The first model then generates one or more modified responses, which are provided—along with some or all of the turn-based conversation containing the original response—to a second AI language model. The first model is retrained based in part on the second model’s output.

Career Highlights

Wu is associated with Character Technologies Inc. Her patent focuses on a feedback-based approach to retraining language models, using conversational context and the output of a second model to help improve the first model.

Collaborations

Wu’s listed coworkers include Satyaki Chakraborty and Landon Jonquiere Brittan Smith. The patent’s described process also involves two AI language models, with the second model contributing output used in retraining the first.

Conclusion

Te-Lin Wu’s patent work at Character Technologies Inc. centers on using contextual feedback and a second AI model to support the retraining of a language model.