Winners of prominent AI open innovation competition announced
Four innovative AI start-ups from Europe were recently announced as the winners of the Large AI Grand Challenge, a competition launched to promote European innovation and excellence in large-scale AI models. Led by the EU-funded project AI-BOOST, the Large AI Grand Challenge paves the way for future collaborations in which European AI talent and resources are optimally utilised. The competition was launched in November 2023 in collaboration with the European Commission and the EuroHPC Joint Undertaking (JU). It aims to reward innovative start-ups and SMEs for their work in developing large-scale AI models that will boost Europe’s competitive edge. The four winners were selected out of a total of 94 proposals, which testifies to the competitive nature of Europe’s AI landscape. They are: Lingua Custodia (France), a fintech company specialising in AI and natural language processing for the finance sector; Textgain (Belgium), an AI start-up helping companies and governments gain insights from unstructured data through predictive text analytics and focusing on the analysis of hate speech; Unbabel (Portugal), a language technologies company combining AI and human translation for multilingual support in all 24 official EU languages; and Tilde (Latvia), a language technology expert offering machine translation and AI-powered chatbots targeting Balto-Slavic languages. The four start-ups will share a total prize of EUR 1 million and an allocation of eight million graphics processing unit hours on two top-range supercomputers: LUMI and LEONARDO.
A fifth winner
The competition was such a success that it led EuroHPC JU to allocate additional compute time to the supercomputer MareNostrum 5 hosted by Barcelona Supercomputing Center. AI-BOOST therefore awarded funding and 800 000 hours of compute time on the supercomputer to the fifth ranked proposal, submitted by Multiverse Computing (Spain), a quantum computing start-up devoted to enhancing the energy efficiency and speed of large language models (LLMs). The company will use the awarded compute time to build and train an LLM from scratch using quantum and quantum-inspired technology. “Winning this challenge recognizes the specific strength of Multiverse Computing in building faster and more energy-efficient LLMs by using quantum and quantum-inspired technology, today,” comments the start-up’s co-founder and CEO Enrique Lizaso Olmos in a news item posted on ‘The Quantum Insider’. “We are proud of the confidence EU leaders have in our expertise and ability to create a new class of LLMs with faster training, smaller datasets and lower operating costs.” Over the next 12 months, the winners of the Large AI Grand Challenge will have to develop a large-scale AI model with a minimum of 30 billion parameters and train it on one of Europe’s supercomputers. Use of the supercomputers will enable them to reduce training times from years to weeks. Following this, the five teams will have to release their developed models under an open-source licence for non-commercial use or publish their research findings. Launched in September 2023, AI-BOOST (Artificial intelligence for better opportunities and scientific progress towards trustworthy and human-centric digital environment) is striving to create and run highly replicable AI open innovation competitions that attract outstanding talent across Europe. To drive progress in the major AI areas, the project will promote collaboration between key stakeholders in the AI community, leading to trustworthy, human-centric real-world solutions. For more information, please see: AI-BOOST project website
Keywords
AI-BOOST, supercomputer, AI, large language model, LLM, Large AI Grand Challenge, start-up
