Back

Innovate UK Supports Our Vision for Full-Stack Generative AI Infrastructure

Backed by Innovate UK, Instill AI and InfuseAI are leading the way in developing a versatile, full-stack generative AI infrastructure that’s set to revolutionize AI's capabilities across languages and industries.

Ping-Lin's github avatar

Published by

Ping-Lin

on 4/25/2024

The theme of this tutorial

In a major development set to transform the landscape of artificial intelligence, we at Instill AI, along with Taiwan's InfuseAI, are excited to announce a significant new collaboration. Under the banner "Optimising Industry Dynamics: A Performance-Persistent AI System Adaptable to Rapid Data and Model Changes", this project is jointly funded by Innovate UK and the A+ Industrial Innovative R&D Program in Taiwan, with a shared funding pool of up to £2 million.

#Tackling Generative AI's Greatest Challenges

Our partnership aims to address some of the most pressing challenges faced by large language models (LLMs), particularly those that have hampered broader application and reliability across diverse industries and languages. We are focusing on three specific challenges: enhancing non-English LLM applications, breaking through the context window limit, and addressing the persistent issue of domain or concept drift.

#Ambitious Goals, Practical Solutions

For non-English applications, a notable gap in the current LLM landscape, InfuseAI is poised to develop a cutting-edge solution for performance evaluation and validation. This initiative will be integrated with our products at Instill AI, ensuring a seamless and robust framework that supports a wider range of languages and cultural contexts.

We, at Instill AI, will take the lead in overcoming the context window limit, which currently restricts the breadth of data that LLMs can process at any given time. By developing a multimodal retrieval-augmented generation (RAG) framework, we aim to extend the knowledge base accessible to AI systems, thus significantly enhancing their utility and effectiveness.

Furthermore, our collaboration will tackle the tricky problem of domain/concept drift—where changes in data input or context over time can degrade the performance of AI models. Our development of a drift detection algorithm promises to alert users automatically when the integrity of the knowledge base or the LLM’s performance is compromised due to such drifts.

#Our Collaboration: A Symbol of Innovation and Unity

This joint initiative not only highlights our innovative capabilities but also demonstrates a significant international collaboration between the UK and Taiwan. By pooling resources and expertise, we are poised to make substantial advancements in AI technology, ensuring that LLMs can be more dynamic, reliable, and inclusive of diverse linguistic and cultural nuances.

The outcomes of this collaboration could potentially set new standards for AI applications in various sectors, promoting greater adaptability and precision in how AI systems understand and interact with human language. As this project progresses, the AI community and industry watchers will be keenly observing its impact on the future of technology and global AI capabilities.

Ready to dive into building Generative AI with our solution? Schedule a call and let’s get started!

Last updated: 4/25/2024, 7:46:40 AM