Cohere News | Sep 06, 2024

Brought to you by RivalSense - an AI tool for monitoring any company.

RivalSense tracks the most important product launches, fundraising news, partnerships, hiring activities, pricing changes, tech news, vendors, corporate filings, media mentions, and other developments of companies you're following 💡

---

Company Overview

Cohere is the leading data security-focused enterprise AI company, providing industry-leading large language models (LLMs) and RAG capabilities tailored to meet the needs of enterprise use cases that solve real-world problems. It is a global technology company co-headquartered in Toronto and San Francisco, with key offices in London and New York.

Latest Updates

• Cohere has rebranded its non-profit research lab, Cohere For AI, to a general research lab.
• Pgai integrates with LLMs like Cohere to generate content and perform data reasoning directly in PostgreSQL.
• Cohere's Command R 32b and R+104b models are now available for 4x faster downloading on Hugging Face.
• Cohere's latest Command R series is optimized for business with fast retrieval-augmented generation and multilingual capabilities.
• Cohere's Rerank model can be integrated with Elastic search engines in less than a week, improving accuracy, deployment speed, and cost savings.
• Cohere Embed and Cohere R+ are featured in Azure AI Studio for embedding and generation tasks respectively.
• Cohere's Command R model will be available on Cerebras' new AI cloud service.
• Cohere released the C4AI Command R 08-2024 model with 32 billion parameters, supporting multilingual generation and various use cases like reasoning and summarization.
• Cohere has upgraded its Command R model series, achieving a 50% increase in throughput, a 20% reduction in latency, and halving the hardware footprint.
• Cohere has launched an LLM University with courses on building AI applications using its Chat, Embed, and Rerank endpoints.
• Cohere released the C4AI Command-R model with 35 billion parameters, optimized for multilingual generation, reasoning, summarization, and question answering.
• Cohere is collaborating with Weights & Biases and Weaviate to launch a free advanced RAG in Production course, open for pre-registration as of 29 August 2024.
• Tabnine now allows users to choose between its proprietary models and third-party options like Cohere’s Command R+ for AI-assisted coding.
• Cerebras plans to offer models from Cohere on its wafer-scale engine (WSE) to enhance its AI inference services.
• Weights & Biases, Cohere, and Weaviate have collaborated on a new project as of 2024-08-28.
• AgentOps.ai integrates with Cohere to support AI agent developers.
• Cohere released a 'JSON schema mode' for LLM inference that is over 10 times faster than OpenAI and 45 times faster than Outlines.
• Cohere introduced Command R+, an RAG-optimized LLM for enterprise workloads, now available on Microsoft Azure.
• Cohere's Command models can now connect to external tools like search engines, APIs, and databases to enhance data interaction capabilities.
• On January 1, 2024, Cohere purchased OpaVote, an online voting platform specializing in ranked choice voting.
• Cohere released Aya 23, a family of multilingual generative language models with 8-billion and 35-billion parameters supporting 23 languages.
• Cohere introduced Agentic AI, focusing on building autonomous AI agents for complex, multistep workflows.
• Cohere Command (R, R+) models are now supported through Bedrock.
• Cohere's English and Multilingual embeddings models are now supported by QnABot on AWS.
• Cohere's Command-R model improved AI task performance with longer context windows up to 128,000 tokens and a more affordable pricing strategy.
• Cohere collaborated with Borderless AI to develop Alberni, an AI agent for human resources using Cohere's Command R model and Retrieval-Augmented Generation technology.
• Cohere has introduced Self-Improving Robust Preference Optimization (SRPO) and expanded its research to include multilingual toxicity mitigation and reinforcement learning from human feedback in language models.
• Cohere made Internet-Search, Python interpreter, and Calculator tools available natively via the Chat API.
• trustgraph.AI added support for Cohere and AWS Bedrock, enabling mixed extraction and RAG model deployments.
• Cohere and OpenAI now support Structured Outputs.
• Cohere announced three new features for their Toolkit to enhance building and deploying production-ready Knowledge Assistants.
• David Mataciunas, an independent AI researcher at Cohere for AI, announced the SID project simulating 1000+ autonomous AI agents in a virtual world.
• Sara Hooker, who leads Cohere’s nonprofit research division, argued in a July paper that current compute thresholds are shortsighted and unlikely to mitigate risk.
• Cohere's models can now be registered in Amazon SageMaker Model Registry without needing the inference specification file.
• Cohere will host sessions on Advanced Retrieval-Augmented Generation and Tool Use on September 10 and The Future of AI and Data on September 11.
• Cohere's project 'Maya: Multimodal Aya' led by Nahid Alam, Surya Guthikonda, and Karthik reddy K. was awarded most promising during Cohere For AI’s Expedition Aya.
• Sara Hooker, who leads Cohere’s nonprofit research division, argued in a July paper that current compute thresholds are shortsighted and likely to fail to mitigate AI risk.
• Cohere is sponsoring the Gen AI Hackathon in San Francisco on September 6th.
• Cohere launched Safety Modes in beta, allowing users to toggle between strict and contextual options for model outputs.
• Cohere is hosting a DevEx & DevTools Happy Hour in NYC on 9 September 2024 with Fern.
• Cohere showcased their Enterprise AI solutions, including generative models Command R and R+, at the AWS Summit NY.
• On 28 August 2024, Cohere researchers Nikolas Gritsch, Acyr Locatelli, Sara Hooker, and Ahmet Üstün published a paper on efficient training of Mixture of Experts models using the Nexus framework.
• Johnson Lambert boosted audit efficiency by 50% using Cohere Command on Amazon Bedrock, supported by Provectus.
• Cohere's recent paper found that incorporating code into training data enhances LLMs' performance in language generation tasks and improves their world knowledge.
• CoreWeave’s infrastructure services are used to train some of the largest models from Cohere.
• Aidan Gomez, CEO of Cohere, discusses its business-only focus and the platform’s ability to securely deploy in any environment.
• Cohere is expected to expand its NLP platform in 2024 to support more languages and use cases, including AI-driven content creation and automated translation.
• Cohere AI's study found that pre-training with code data is crucial for generalizing across various tasks, not just coding.
• CohereForAI explored the question 'To Code or Not To Code?' in their recent paper.
• Aidan Gomez, CEO of Cohere, discussed the evolving economic model of generative AI and the importance of the application layer in a recent interview.
• Researchers from Cohere For AI and Cohere conducted large-scale pre-training experiments to study the impact of code data on LLM performance.
• Cohere published a new paper exploring the impact of code data in pre-training.
• Cohere's latest preprint, 'To Code, or Not To Code?', found that including 25% code in pre-training significantly improves natural language reasoning and code performance.
• Cohere published a new research paper on arXiv.
• Aidan Gomez, CEO of Cohere, emphasized the critical importance of data quality for reliable AI, noting that even a single bad example can significantly impact model performance.
• Cohere's recent paper shows that including code in pretraining data improves performance on non-code tasks, with an optimal proportion around 25%.
• Cohere will be heading to Hack the North 2024, hosting an API challenge with a workshop and offering prizes.
• Cohere organized an event showcasing AI Agent orchestrator and AI Agents that transform websites into APIs.
• Cohere's CohereForAI team received recognition for their dedication to advancing AI.
• Cohere's AI Safety team, led by Seraphina Goldfarb-Tarrant, is studying the fairness of LLMs in hiring contexts using real-world data.
• Cohere will be hosting a workshop on their open-source chat interface at the SF AWS Gen AI Loft.
• Cohere will be participating in an event at Siam Paragon SCBX Next Tech on the 4th floor, sponsored by Amazon Web Services.
• Cohere will be participating in the AI/ALL Summit: Future of Equitable and Inclusive AI, a side event of ACL 2024.
• Sama's platform now integrates with Cohere, among other model providers, to evaluate and rank prompts and model responses.
• Hugging Face announced a unified API for tool use across models from Cohere, Mistral AI, Meta AI, and Nous Research.
• Cohere co-founder Nick Frosst emphasized the need for realistic expectations about AI capabilities.
• Nick Frosst, co-founder of Cohere, said on TechCrunch's Found podcast that he doesn't think the AI industry is in a bubble.
• Cohere worked on an open-source vector database with binary and int8 integration, offering in-memory or on-disk vectors.
• Cohere was recognized among the 2024 Forbes Cloud 100 list.
• Cohere will be hosting a hands-on workshop on 'Building Open-Source Local LLMs' with AWS Developers at the AWS GenAI Loft in San Francisco on August 14.
• Cohere's co-founder Nick Frost discussed building a lasting AI's economic potential on TechCrunch's Found podcast.

---

Did you find it useful?

If you liked this report, consider following your own companies of interest. Receive weekly insights directly to your inbox using RivalSense AI.