An Update on AI-Powered Translations in Canvas

TamasBalogh
Instructure
Instructure
0
83

Canvas.png

With the end of the year rapidly approaching, we want to share an update on the highly anticipated AI-powered Discussions and Inbox Translations. As part of our commitment to delivering high-quality, reliable, and scalable features, we will continue to keep this feature in beta testing until critical areas of improvements identified through user feedback have been successfully resolved.

Key Findings from Beta Testing

Scalability: While translations perform well in courses with a smaller number of enrollments, it hasn't scaled as we’d expect with increased usage. We recognize how important large-course support is for many of our customers, so ensuring this feature is scalable is of top priority.

Accuracy: Beta testing brought to light inconsistencies in translation responses, so we’re working to refine the model to ensure the translations are as accurate and relevant as possible. 

Performance: Translation speed is crucial for a great user experience, so we’re continuing to optimize the tool to meet our high-performance standards.

Next Steps

The observations described above are primarily linked to the language model that powers Translations in Canvas. To address these concerns, and deliver a robust and reliable translation tool at no additional cost, we plan to migrate to an improved model that better addresses our scalability, accuracy, and performance concerns. Once the adoption of the new language model is released, we will update AI Nutrition facts accordingly.

While this change will push our production release into 2025, it will allow us to offer our customers a more reliable translation tool that better aligns with Instructure's standards of quality and user satisfaction.

What to Expect in 2025

We are working hard to migrate to the new language model and will provide an updated Production Release date as soon as we can. We will continue to share updates via the Product Blog and encourage you to submit questions or feedback in the comments section below. Thank you for your understanding as we work to resolve this issue in the new year.