We Launched on Product Hunt and got to Product of the day: Check us out here
APPLICATIONS ARE NOW CLOSED
Large Language Models (LLMs), the AI models behind applications such as ChatGPT, have the ability to enable natural interactions without the need for explicit instructions and manually labelled data, unlike conventional conversational Artificial Intelligence(cAI) systems. This can be a game-changer, particularly for cAI applications, such as customer service automation.
However, this technology is known to generate responses containing false information, known as "hallucinations", which can undermine customer trust and harm a company's reputation, and pose a big bottleneck in the wider adoption of LLMs.
The state-of-the-art methods for addressing hallucinations in LLMs, including identifying, monitoring, and fixing them, are only currently found in academic research and proprietary internal tools developed by major tech companies such as Google, Meta, and OpenAI, with low TRL.
As such, there is currently no commercially available solution that efficiently tackles hallucinations generated by LLMs, hindering the adoption of this powerful technology for multiple use cases, including cAI, particularly for SMEs.
In this project, our consortium will develop a system that will use a company's internal knowledge base to provide real-time reply recommendations to customer agents. Their feedback and responses will be used as training data for Reinforcement learning with Human Feedback, the same technique used by OpenAI to reduce hallucinations on ChatGPT. We will also explore other techniques like constitutional AI. Finally, we will to deploy this model on all of our commercial partners' premises to ensure no data leaves their infrastructure.
• The participating startup will provide feedback that would require (bi)weekly meetings, an operational sandbox, and access to their data.
Algomo is a ChatGPT-powered bot that automates customer support and sales in 100+ languages.
We don't have expertise in AI can we still apply?
Yes. No expertise in AI is necessary.
What is the commitment after getting accepted?
The participating startup will provide feedback that would require (bi)weekly meetings, an operational sandbox, and access to their data.
We are based outside the UK, can we still apply?
Unfortunately, due to UK government grant restrictions, a limited company in the UK is necessary.
When is the deadline?
1st September 2023
