Docusign Acquires Lexion to Enhance AI Capabilities in Agreement Management
Docusign has acquired Lexion, an AI-powered agreement management software provider, for $165 million. This acquisition strengthens Docusign’s position in the Intelligent Agreement Management (IAM) space and expands its AI capabilities. Lexion’s technology uses natural language processing to extract structured data from contracts, providing users with a searchable repository, reporting tools, and integrations. Docusign plans to integrate Lexion’s technology into its IAM platform, offering customers enhanced insights into their agreements, AI-powered contract development assistance, and streamlined document management. This acquisition reflects Docusign’s commitment to innovation in agreement management and its focus on leveraging AI to improve efficiency and productivity for its customers. Lexion’s team, including its founders, will join Docusign in senior product and engineering roles, contributing their expertise to the company’s IAM initiatives.
OpenAI Partners with Stack Overflow to Enhance Coding Capabilities of AI Models
OpenAI has announced a partnership with Stack Overflow, the popular platform for programmers, to improve the coding abilities of its AI models, including ChatGPT. This collaboration will provide OpenAI developers with access to Stack Overflow’s vast repository of vetted technical knowledge and code, enhancing the accuracy and reliability of AI-generated responses. The partnership comes after Stack Overflow previously banned ChatGPT-generated content due to concerns about inaccurate information. However, the platform now recognizes the potential of AI to improve the developer experience and is working with OpenAI to ensure the quality and trustworthiness of AI-generated code. As part of the collaboration, OpenAI will leverage Stack Overflow’s OverflowAPI, a subscription-based service that provides access to the platform’s extensive datasets. This will enable OpenAI to train and fine-tune its models using reliable and up-to-date technical information. ChatGPT users will benefit from this partnership by receiving more accurate and informative coding-related answers. Each response will include links to relevant Stack Overflow posts, allowing users to delve deeper into the underlying technical knowledge. Stack Overflow will also utilize OpenAI’s models to develop its own generative AI capabilities for both its public platform and enterprise offering.
Meta Researchers Develop Multi-Token Prediction Technique to Boost AI Model Speed and Accuracy
Researchers at Meta have developed a new technique called “multi-token prediction” that significantly improves the speed and accuracy of large language models (LLMs). This approach deviates from the traditional “next-token prediction” method, where models predict one token at a time, by enabling them to predict multiple tokens simultaneously. Multi-token prediction addresses some limitations of next-token prediction, such as sensitivity to local patterns and overlooking predictions that require longer-term reasoning. By predicting multiple tokens at once, models can learn more general patterns and achieve higher accuracy with less data. The researchers tested this technique on various tasks and found that it leads to significant improvements, especially with larger models. Models trained with multi-token prediction achieved better performance on coding benchmarks and were up to three times faster at inference time. This research has the potential to enhance the efficiency and accuracy of generative tasks such as code completion, making it valuable for enterprise applications.
Nvidia and Intrinsic Partner to Advance Next-Generation Robotics
Nvidia and Alphabet’s Intrinsic are collaborating to revolutionize robotics with advancements in grasping capabilities and industrial scalability. The partnership leverages Nvidia’s Isaac Manipulator platform and AI expertise to enhance robotic dexterity and automation. Isaac Manipulator, introduced by Nvidia earlier this year, provides foundation models and GPU-accelerated libraries for dynamic manipulation tasks. These models enable robots to perceive and make decisions autonomously, similar to human understanding. Intrinsic has utilized Nvidia’s Isaac Sim on the Omniverse platform to generate synthetic data for vacuum grasping, a challenging skill in robotics. This collaboration demonstrates the potential for a universally applicable grasping skill that can work across various grippers, environments, and objects. The partnership aims to bring state-of-the-art dexterity and modular AI capabilities to robotic arms, accelerating the development of new robotics and automation tasks. Nvidia and Intrinsic will showcase their collaborative efforts at the Automate trade show in Chicago, highlighting the transformative impact of foundation models on the robotics industry.
