Ekimetrics launches CLAIR to demystify and democratize AI
Back to all articles
Ekimetrics has become the first company in its sector to propose an application based on a multi-agent AI to the general public.
– “Will AI destroy our jobs?” “What’s an algorithmic bias?” “What are the biggest risks of AI for our civilization?” “Does the AI Act protect enough to control the challenges of AI?” These are just some of the questions the general public can ask CLAIR (Comprehensive and Legitimate AI Responses) from today. Available online and on mobile devices, CLAIR is a platform created by Ekimetrics, the European leader in data science and artificial intelligence solutions, to demystify and democratize artificial intelligence.
According to a recent survey by Pew Research Center, 71% of United States adults say they have little to no trust in companies to make responsible decisions about how they use AI in their products.
“It’s true that, in certain aspects, AI can scare people, but it also has multiple advantages. That’s why we want to make CLAIR available to as many people as possible so that everyone can have a more nuanced and objective vision.”, explains Annabelle Blangero, Responsible AI Lead at Ekimetrics.
Multi-agent technology dedicated to responsible AI
CLAIR helps to understand the limitations, risks, and regulations linked to AI from the analysis of official documents and expert reports, such as practical guides on AI by the CNIL (National Commission on Informatics and Liberty), the Oxford Commission on AI & Good Governance or the report by the French interministerial Commission on Artificial Intelligence.
Based on the technology of OpenAI’s GPT-4, the platform relies on a multi-agent AI, a first in an application for the general public. Each of the three agents proposed by CLAIR functions with its own language model, allocated according to the following expertise:
- Ada: Technical expert in AI
- Norma: Expert in the legal regulations of AI
- Sophia: Social sciences expert
Each agent has access to a specific document base and can answer questions either with its internal knowledge or by consulting a database. A fourth model summarizes the input from all three agents to give a complete and qualified reply.
“CLAIR offers a simple and informative interface that’s accessible to everyone. It functions as a sort of enhanced Wikipedia page where users can ask questions and receive precise, detailed answers.”, explains Annabelle Blangero. “We’re very proud to provide users with a multi-agent AI that many experts consider the future of generative AI.”
A tool in line with the promotion of Responsible AI
With CLAIR, Ekimetrics proposes a tool that reflects their engagement to sustainable and responsible AI via three key levers:
- Transparency and ethics: CLAIR guarantees the transparency of data sources by citing all sources. A tab dedicated to responsible AI explains in detail all the practices and technologies used to create the platform.
- Carbon impact: Ekimetrics uses a library developed by the Data for Good association to assess the carbon consumed by queries to the CLAIR model.
- Bias control: Tests are carried out to identify potential biases and prevent attempts to bypass (jailbreak) censorship of the AI model.
“With the launch of CLAIR, we’re very proud to contribute to the promotion of Responsible AI, a cause that Ekimetrics has championed since its creation. The next stage of the project will be to explore smaller models than GPT-4 to achieve a balance between permanence and frugality.”, concludes Jean-Baptiste Bouzige, CEO of Ekimetrics.
.avif)
