February 2024 Summaries

Cohere is committed to preserving customer ownership rights over their creations made using its AI technology, marking a significant milestone in its mission to help enterprises harness AI capabilities. The company aims to enhance the business impact of generative AI and boost productivity for enterprise clients while staying informed about evolving legal and regulatory landscapes. Cohere pledges to collaborate with stakeholders to ensure their approach aligns with industry standards.

In a discussion with UK Secretary of State for Science, Innovation, and Technology Michelle Donelan, Cohere's CEO Aidan Gomez emphasizes the importance of early engagement and experimentation with AI technology for future readiness. Cohere experts Sudip Roy and Neil Shepherd have published a guide to assist enterprises in selecting the right AI model, considering factors such as open-source versus proprietary options, model scalability, and support levels. The guide highlights the significance of choosing between general or tailored models based on specific use cases and performance requirements, and the distinction between transformational and incremental AI adoption strategies. Cohere has introduced new tools, including a semantic search system using Cohere Embed and Pinecone's serverless solution. Additionally, Cohere is involved in AI safety initiatives and has launched Aya, a multilingual open-source model in collaboration with a global network of researchers. The company joins the newly formed U.S. AI Safety Institute Consortium to further AI safety and development.

Avoiding the challenges associated with generative AI projects requires dedicated oversight, which may lead to the emergence of new leadership roles such as Head of Gen AI Transformation. As proof-of-concept projects transition to full-scale production, businesses must address cross-functional and organizational challenges to maximize the benefits of AI assistants. Cohere offers a framework for integrating private datastores with its large language models, boasting over 100 connectors and enabling Retrieval-Augmented Generation (RAG) for enterprises. Neil Thompson from MIT discusses scaling AI within enterprises, emphasizing the importance of efficiency and flexibility. Developers can learn to use Cohere's latest connector mode for secure data access through Cohere Chat. In 2023, Cohere For AI, a non-profit research lab, published 30 cutting-edge machine learning papers in collaboration with over 40 institutions. Cohere's CEO, Aidan Gomez, participated in a discussion at Davos with prominent AI figures about generative models, exploring potential advancements in the field.

In an interview with Graham Neubig, an associate professor at Carnegie Mellon University, he discusses the evaluation and future of large language models (LLMs) and neural network architectures beyond Transformers. Neubig emphasizes the importance of using academic datasets for initial evaluations but stresses the need for iterative examination of real-world data outputs to identify and correct errors. He highlights a growing trend in academia towards developing more complex and industry-representative datasets, and he expresses interest in LLM-backed agents, which can either use tools to solve tasks or act independently to impact the world. Neubig also explores the potential of combining LLMs with reinforcement learning and mentions the emergence of new architectures like Mamba, which challenge the dominance of Transformers. Looking ahead to 2024, he is keen on improving evaluation reliability and developing small, adaptable open-source models that perform well on specific tasks.

The Aya project represents a significant advancement in multilingual AI research, aiming to bridge the gap created by previous language limitations in AI models. By gathering contributions from over 3,000 researchers worldwide, Aya focuses on enhancing support for underserved languages, outperforming existing open-source multilingual models like mT0 and Bloomz in tasks such as natural language understanding, summarization, and translation. It covers more than 50 previously unserved languages and offers the most extensive multilingual dataset to date, with 513 million prompts and completions in 114 languages, including 204,000 rare human-annotated entries. Aya's open-source, fully permissive Apache 2.0 licensed model and data collection provide a valuable resource for developers and researchers, promoting linguistic diversity and offering a foundation for further open science projects. The initiative invites academics, civil institutions, and small companies to participate and contribute to enhancing AI's cultural and linguistic relevance.

Cohere has joined the AI Safety Institute Consortium (AISIC), a new initiative established by the National Institute of Standards and Technology (NIST) to promote the safe development and deployment of AI technology. AISIC brings together over 200 top AI stakeholders, including AI developers, users, academics, government and industry researchers, and civil society organizations, to collaborate on AI safety efforts. The consortium aims to develop safe and trustworthy AI systems that will support future standards and policies. Cohere's participation follows its previous commitments to AI safety initiatives in the U.S., Canada, and the UK, and it continues to work with global policymakers to ensure that AI regulation fosters innovation while prioritizing responsible development.

The text emphasizes the importance of not being hindered by the decision-making process when choosing an AI model but encourages taking the necessary time to understand the associated costs, performance, and risks to ensure successful application in delivering significant value. It highlights contributions from Sudip Roy and Neil Shepherd of Cohere, who hold key positions in the company, and mentions recent developments, such as the opening of a new office in Paris and advancements in AI adoption for national security and secure translation services for global enterprises.

The UK government introduced a regulatory framework aimed at harnessing the benefits of AI while ensuring its safe and responsible development, highlighted by a visit from Secretary of State for Science, Innovation, and Technology Michelle Donelan to Cohere’s London office. During her visit, Donelan engaged with Cohere’s leadership, experienced a demonstration of their retrieval-augmented generation (RAG) technology, and emphasized the UK's status as a major global AI destination. Cohere, established in 2022, has made its London office a key research hub, housing about 10% of its global workforce, and plans to double its UK headcount in 2024. The company values the UK for its tech talent and market opportunities, with CEO Aidan Gomez underscoring the importance of balancing innovation with responsible AI development. Cohere is committed to collaborating with the UK government to advance AI technologies that benefit enterprises.

Cohere is a company that specializes in developing enterprise-grade AI solutions designed to enhance workplace productivity and business insights through a range of products and tools. Their offerings include North, an AI platform tailored for modern work environments, and Compass, a system for intelligent search and discovery. Additionally, they provide high-performance language models under the Command suite and multilingual capabilities with Aya Expanse. The company emphasizes robust AI security and data protection, offering private deployments and customizable solutions across various industries such as technology, financial services, healthcare, manufacturing, energy, and the public sector. Cohere is also active in research through Cohere Labs and supports open science initiatives, while fostering a community with events, resources, and partnerships.