top of page
  • Writer's pictureSarah Ruivivar

AWS Unveils Custom Model Import for Generative AI

Image credits: Rubaitul Azad / Unsplash

Amazon's cloud computing arm, AWS, is making a play to be the go-to spot for companies to host and refine their custom generative AI models.


In a recent announcement, AWS unveiled the preview of Custom Model Import, a new feature in Bedrock, its suite of generative AI services aimed at enterprises. This feature allows organisations to import their in-house generative AI models as fully managed APIs, placing them alongside other generative AI models in Bedrock’s library.


Vasi Philomin, VP of generative AI at AWS, explained that this feature allows customers to bring their proprietary models to Bedrock and use them with all of the workflows already on Bedrock.


A recent poll by Cnvrg, Intel’s AI-focused subsidiary, revealed that most enterprises are building their own models and refining them for their applications. However, infrastructure, including cloud compute infrastructure, is seen as the biggest barrier to deployment.


 

Want to learn more about AI's impact on the world in general and property in particular? Join us on our next Webinar! Click here to register

 

AWS aims to fill that gap with Custom Model Import, while keeping up with cloud competitors. AWS' Bedrock offers a wider range of model customisation options than the competition, according to Philomin.


Guardrails, a feature that lets users configure thresholds to filter models’ outputs for things like hate speech, violence, and private information, and Model Evaluation, a tool customers can use to test a model’s performance, are now generally available.


However, it's worth noting that Custom Model Import currently only supports three model architectures: Hugging Face’s Flan-T5, Meta’s Llama, and Mistral’s models.


In related news, Titan Image Generator, AWS’ text-to-image model, is now generally available. It can generate images with more "creativity", according to Philomin. AWS also released the second generation of its Titan Embeddings model, Titan Text Embeddings V2, claiming it is more efficient, cost-effective, and accurate.


As AWS continues to expand its generative AI offerings, we'll be watching to see how these developments shape the AI landscape.



 

Want to learn more about AI's impact on the world in general and property in particular? Join us on our next Webinar! Click here to register

 


Made with TRUST_AI - see the Charter: https://www.modelprop.co.uk/trust-ai

2 views0 comments

Comments


bottom of page