TL;DR:
- Cohere introduces a connector development tool for Command LLM.
- This tool enables Command to access data from external applications, expanding its capabilities.
- Cohere’s valuation reached $2.2 billion after a successful $270 million Series C funding round.
- Command LLM is designed to optimize performance across common enterprise use cases.
- Prebuilt connectors on GitHub facilitate data retrieval from popular cloud services.
- Cohere’s RAG feature allows for the incorporation of external data into AI models.
- Companies can now enhance their answers with proprietary information.
- Competitors like OpenAI also offer customization of AI models with external data.
- OpenAI’s recent program allows extensive customization of AI models.
Main AI News:
In a strategic move, Cohere Inc., the well-funded artificial intelligence startup, has rolled out a groundbreaking connector development tool tailored for its flagship Command large language model. This innovative tool empowers customers to construct integrations that facilitate Command’s access to data from external applications. By leveraging this external data, the Command model can now provide responses to user inquiries that were previously beyond its reach.
Aidan Gomez, the Co-founder and Chief Executive Officer of Cohere, spearheads the Toronto-based company. Gomez, recognized as one of the eight pioneering researchers behind the Transformer neural network architecture in 2017, has led Cohere to impressive heights. The company recently achieved a remarkable valuation of $2.2 billion in July, following a successful $270 million Series C funding round that saw participation from tech giants such as Nvidia Corp., Oracle Corp., and other industry leaders.
Cohere’s offerings include a suite of Large Language Models (LLMs) that enterprises can seamlessly deploy on major public cloud platforms or on their on-premises hardware infrastructure. Command, Cohere’s flagship LLM, boasts a wide array of capabilities, including text generation, document repository searches, and efficient business record categorization. Notably, Command was meticulously trained on a dataset explicitly designed to optimize its performance in addressing common enterprise use cases.
By default, Command relies on its training dataset to respond to user queries. However, with the introduction of the connector tool unveiled today, Command can now tap into information from virtually any external system equipped with an application programming interface (API) that supports data retrieval. Furthermore, this tool integrates seamlessly with cybersecurity controls in external systems, ensuring secure and controlled access to sensitive records.
To expedite the development process for its customers, Cohere has made approximately 100 prebuilt connectors available on GitHub. These connectors enable Command to retrieve data from popular cloud services like Asana, Slack, and GitHub. Additionally, prepackaged integrations are provided for various other types of applications, including widely used open-source databases.
The newly introduced connectors build upon Cohere’s existing retrieval-augmented generation (RAG) feature within the Command model. RAG is a cutting-edge machine learning technique that empowers AI models to access external application data without requiring extensive code modifications or retraining. This technique involves pairing an LLM with a secondary neural network, which translates data from external systems into mathematical embeddings, making it comprehensible for the LLM.
Cohere has incorporated citations into Command’s RAG feature, ensuring transparency and enabling users to quickly identify the source documents used to generate responses. In a recent blog post, Cohere engineers Roy Eldar and Beatrix De Wilde emphasized how companies can now ground their responses in proprietary information, regardless of the third-party applications used for corporate collaboration.
It’s worth noting that several of Cohere’s competitors also offer similar capabilities to expand their AI models’ knowledge base with external data. For instance, OpenAI is currently testing an API that allows customers to fine-tune its flagship GPT-4 model using custom datasets, providing an identical feature for five of its earlier language models.
At a recent DevDay developer event, OpenAI introduced a program that enables companies to extensively customize their models, offering complete control over every step of the model training process. These customized LLMs will be exclusively available to the organizations that commission them, marking a significant step in the evolution of AI customization and deployment.
Conclusion:
Cohere’s innovative connector tool for Command LLM marks a significant advancement in enterprise AI, allowing for seamless integration with external data sources. This development positions Cohere as a formidable player in the market, offering enhanced customization and competitiveness. It also reflects a broader trend in the AI industry towards empowering organizations with greater control over their AI models, as seen with OpenAI’s recent program. The market can expect increased adoption of such customizable AI solutions, driving innovation and efficiency in various industries.
