The European Union has announced the winners of its “Large AI Grand Challenge,” a competition aimed at accelerating innovation in large-scale AI model making. Four startups will share €1 million in prize money and 8 million GPU hours to train their models on the EU’s high-performance computing (HPC) supercomputers over the next 12 months.
The winners are Lingua Custodia, a French fintech specializing in natural language processing (NLP) for financial document processing; Textgain, a Belgian startup focused on NLP for text processing, with a focus on analyzing unstructured data, such as social media chatter for hate speech; Tilde, a Latvian startup offering machine translation and AI-powered chatbots for Balto-Slavic languages; and Unbabel, a Portuguese startup that blends machine translation with human expertise for customer service and productivity use cases.
The EU received a total of 94 proposals for the challenge, and the winners will be expected to release their developed models under an open source license for non-commercial use or publish their research findings at the end of the training period.
The EU’s supercomputers will support the winners, with Lumi and Leonardo, two pre-exascale HPC supercomputers, providing 8 million GPU hours for model training. A fifth startup, Multiverse Computing, which focuses on improving energy efficiency and speed of large language models using “quantum-inspired tensor networks,” will receive 800,000 computational hours on MareNostrum 5.
The EU has recognized the need to reconfigure its HPC infrastructure for the generative AI age, and has proposed upgrading the supercomputers and building a support layer to improve accessibility for AI startups. The Commission has also aimed to accelerate innovation in AI by providing more access to the bloc’s supercomputing hardware.
