December 2022 Summaries

Cohere For AI, launched in June 2022 as a non-profit research lab, aims to revolutionize machine learning research by fostering inclusivity and collaboration. Within its first six months, the lab, led by Sara Hooker, published significant research contributions in areas like metadata archeology and multilingual model compression, collaborating with over 20 institutions, including renowned universities and Google Research. The Cohere For AI Scholars Program commenced to bridge the gap between research opportunities and aspirants lacking formal credentials, while the growing C4AI Community expanded to over 800 members from 90 countries, emphasizing generosity and collaboration. Cohere For AI actively engaged in global ML conferences and hosted events to inspire and educate, with plans for continued innovation and community growth in 2023.

Cohere, a company focused on advancing natural language processing (NLP) for businesses, has announced the appointment of Martin Kon, formerly Google's YouTube CFO, as its new President and COO. With extensive experience in market strategy and commercial data analytics, Martin will drive Cohere's efforts to make NLP more accessible to businesses of all sizes, capitalizing on its potential to transform how companies utilize language data. His leadership is expected to help Cohere establish itself as a commercial leader in language AI by bringing high-demand products to market, optimizing existing offerings, and helping enterprises leverage the power of language models for various applications such as predictive text generation, search, and content moderation. This move is part of Cohere's broader strategy to unlock the potential of language AI across multiple industries, recognizing that a significant portion of human-generated data is in text form, yet remains underutilized by many businesses today.

Cohere has introduced its first multilingual text understanding model, supporting over 100 languages and delivering significant performance improvements over existing open-source models. This model, designed for applications such as semantic search, content aggregation, and cross-lingual content moderation, enables businesses and platforms to effectively serve global markets by understanding and processing text across multiple languages. By mapping text into a semantic vector space, it facilitates improved search results and insights, outperforming traditional keyword searches and other multilingual models, particularly in capturing language nuances and longer text meanings. Trained on a vast dataset of authentic question/answer pairs from diverse languages, the model addresses previous limitations in multilingual natural language processing. Cohere's model showcases improved capabilities in multilingual semantic search, customer feedback aggregation, and zero-shot content moderation, making it a valuable tool for international operations.