Tag: OpenAI

  • OpenAI｜Update on Safety & Security Practices September 16, 2024

    —

    by

    air-beans
    in

    OpenAI has established an independent governance committee chaired by Zico Kolter to oversee safety and security processes for AI model development and deployment. Key efforts include enhancing cybersecurity measures, fostering transparency, collaborating with external organizations, and building unified frameworks for model safety and monitoring. The company is working towards setting industry-wide safety standards and advancing…

  • OpenAI｜Release of the OpenAI o1 System Card

    —

    by

    air-beans
    in

    OpenAI has published the system card for o1-preview and o1-mini, detailing the safety assessments and external red-teaming efforts conducted before their release. The report evaluates risks based on the Preparedness Framework, assigning an overall “medium” risk level to o1. Its advanced reasoning capabilities improve safety by reducing the likelihood of generating harmful content, demonstrating state-of-the-art…

  • OpenAI｜Introducing OpenAI o1: Enhancing Reasoning with Large Language Models

    —

    by

    air-beans
    in

    OpenAI has introduced the “OpenAI o1,” a new large language model trained using reinforcement learning to handle complex reasoning tasks. This model excels in areas such as coding and mathematics, achieving impressive results in competitions like Codeforces and the American Invitational Mathematics Examination (AIME). The model also outperforms human experts on scientific benchmarks, demonstrating its…

  • OpenAI｜Announcing OpenAI o1-mini: A Cost-Effective AI Model for STEM Reasoning

    —

    by

    air-beans
    in

    OpenAI has introduced a new cost-efficient reasoning model, “OpenAI o1-mini,” specifically optimized for STEM fields, particularly excelling in math and coding. The o1-mini performs almost as well as the larger OpenAI o1 model on evaluation benchmarks, while being 80% cheaper. It is designed to offer a faster and more affordable option for developers who need…

  • OpenAI｜Announcing the OpenAI o1 Series: A New AI Model for Tackling Complex Problems

    —

    by

    air-beans
    in

    OpenAI has announced the release of a new AI model series, “OpenAI o1,” aimed at solving complex problems. The initial version excels in fields such as science, coding, and math, offering improved reasoning capabilities compared to previous models. Additionally, a new safety training approach enhances the model’s ability to follow safety guidelines, resisting attempts to…

  • Ada｜Using GPT-4 to Deliver a New Customer Service Standard

    —

    by

    air-beans
    in

    Ada, a customer service automation platform, has utilized OpenAI’s GPT-4 to shift from the traditional “containment rate” to a new metric called “resolution rate,” which measures how well customer queries are resolved. With this approach, Ada has significantly improved service quality, achieving resolution rates above 80% for some clients. Additionally, Ada has leveraged OpenAI’s fine-tuning…

  • Lowe’s｜Using OpenAI to Improve Ecommerce Data Quality

    —

    by

    air-beans
    in

    Lowe’s, a leader in home improvement with over 16 million weekly transactions, aims to improve data accuracy on its e-commerce site. The Data and AI team leveraged OpenAI’s GPT-3.5 to enhance product information precision. By employing prompt engineering, Lowe’s reduced tagging errors and boosted search result accuracy by 20%. The initiative also eased the burden…

  • OpenAI｜Fine-Tuning GPT-4o Webinar Hosted

    —

    by

    air-beans
    in

    On August 26, 2024, OpenAI hosted a webinar on fine-tuning GPT-4o. The event covered methods for customizing GPT-4o to meet specific industry needs and adjusting the model for optimized performance. Attendees learned detailed steps for model fine-tuning and practical applications for businesses and teams. The session also highlighted the use of the latest technologies, such…

  • OpenAI｜OpenAI Partners with Condé Nast

    —

    by

    air-beans
    in

    OpenAI has announced a partnership with Condé Nast to display content from major brands like Vogue, The New Yorker, Condé Nast Traveler, GQ, Architectural Digest, Vanity Fair, Wired, Bon Appétit, and more, within its products, including ChatGPT and the SearchGPT prototype. The new SearchGPT prototype is being tested to make finding information and reliable content…

  • OpenAI｜Fine-tuning Now Available for GPT-4o

    —

    by

    air-beans
    in

    OpenAI has launched the fine-tuning feature for GPT-4o, responding to one of the most requested features by developers. This feature allows developers to enhance GPT-4o’s performance by using custom datasets to tailor responses for specific use cases. Additionally, OpenAI is offering 1 million free training tokens per day to every organization until September 23, 2024.…