• Claude Baudoin with cebe IT

  • Sep 19 2024
  • Duração: 16 minutos
  • Podcast

Claude Baudoin with cebe IT

  • Sumário

  • Claude Baudoin, co-chair of several OMG task forces, including the AI Platform Task Force, discussed the timeliness of AI standards. He highlighted the need for standards to address interoperability and portability issues in AI, citing examples like the portability of neural network models and standardization of image classifiers. Baudoin emphasized OMG's open process for determining standards through RFIs, contrasting it with ISO's more guideline-focused approach. He encouraged active participation in OMG to shape standards, offering a competitive advantage in the market. OMG's efforts aim to provide concrete, technical standards, unlike ISO's general advice. Action Items
    • [ ] Issue a request for information to determine needed AI standards.
    • [ ] Get involved in OMG's standards development process by becoming a member and attending meetings.

    Outline Introduction and Participant Roles
    • Karen Quatromoni introduces herself as the Director of Public Relations for Object Management Group (OMG) and welcomes Bill Hoffman, the OMG CEO and chairman.
    • Claude Baudoin introduces himself as the owner and principal at cebe IT and Knowledge Management, based in San Rafael, California.
    • Claude mentions his extensive background in software engineering, IT management, and his long association with OMG, including his roles in various task forces and working groups.
    • The focus of the podcast is on AI standards, and Claude is co-chairing the AI Platform Task Force for OMG.

    Timing of AI Standards
    • Bill Hoffman discusses the rapid evolution of AI and the timeliness of discussing AI standards.
    • Claude shares an anecdote from 1993 about the premature standardization debate and how it eventually led to the development of the Unified Modeling Language (UML).
    • The fundamental issue is that lack of standards can lead to significant time wasted on interoperability and portability issues, hindering innovation.
    • The goal is to identify when standards are needed to allow developers to focus more on innovation rather than technical challenges.

    Examples of AI Standards Needed
    • Claude explains the current work on portability of neural network models with the help of Zephyr Solutions.
    • The challenge is that neural network models cannot be easily moved from one platform to another due to the lack of a standard representation.
    • Another area of interest is image classifiers, where there is a need for a standard to handle large datasets and descriptions of images.
    • Other potential standards include metadata for data sets and semantic tagging of information, which are widely needed.

    Determining AI Standards
    • Bill asks how OMG determines which AI standards are needed.
    • OMG has an open process that involves issuing requests for information (RFIs) to gather input from the general public, not just OMG members.
    • In 2019, NIST issued an RFI, but it was five years old and did not address the current landscape, including the rise of large language models.
    • OMG is considering issuing a new RFI to gather more up-to-date input on needed standards.

    OMG's Approach to AI Standards
    • Bill inquires about how OMG's efforts differ from other organizations like ISO and IEEE.
    • ISO's standards are more like guidelines, while OMG focuses on more precise, technically detailed standards.
    • OMG collaborates with ISO and IEEE, feeding specifications to ISO and having a liaison with IEEE's project group on AI terminology and data formats.
    • OMG's work is more concrete and provides specific models and formats for users and developers.

    Getting Involved in OMG
    • ...
    Exibir mais Exibir menos
activate_samplebutton_t1

O que os ouvintes dizem sobre Claude Baudoin with cebe IT

Nota média dos ouvintes. Apenas ouvintes que tiverem escutado o título podem escrever avaliações.

Avaliações - Selecione as abas abaixo para mudar a fonte das avaliações.