2024-01-18 AIM TF Meeting Notes

Owned by Wenjing Chu
Last updated: Jan 19, 2024
3 min read

Meeting Date & Time

  •  
    • 09:00-10:00 PT / 16:00-17:00 UTC 

Zoom Meeting Links / Recordings

Meeting: https://zoom.us/j/98931559152?pwd=d0ZwM1JHQ3d5cXRqVTh4NlRHeVJvQT09

Recording: https://zoom.us/rec/share/v5eWw321w-jkgX2Cx6YCCS7mgD0JKSOk6cUiP5u3j4ASf_GCmNjvKvJIyWkIT9E.sbiHwDUZXCYkh27V

Transcript: https://zoom.us/rec/share/EVAr000qZRtec10yTBWVXJyDT00K6Wew4KOHo2MId4EIrUfSaEvFnhQ5w44TXzk.vlVcYaVNLZhMEYoh


Attendees:

Wenjing Chu 
@Anita Rao
@Steven Milstein
@Carl Smith
@Kevin Kenyan
@Keerthi Thomas
@Drummond Reed
@Judith Fleenor
@Neil Thompson
@Bo Haralid
@Eric Drury

Main Goal of this Meeting

Agenda Items and Notes (including all relevant links)

TimeAgenda ItemLeadNotes
5 min
  • Start recording
  • Welcome & antitrust notice
  • Introduction of new members
  • Agenda review
Chairs
  • Antitrust Policy Notice: Attendees are reminded to adhere to the meeting agenda and not participate in activities prohibited under antitrust and competition laws.
  • ToIP Policy: Only members of ToIP who have signed the necessary agreements are permitted to participate in this activity beyond an observer role.
  • ToIP TSWG IPR Policy: see TF wiki page. AI & Metaverse Technology Task Force
5 mins
  • Introduction of new members
  • Any general announcements, news, that could be of interest to the TF
  • Last week's action items
All
50 mins2024 plan discussions

ALL 

  1. Authentic AI / Responsible AI problems and solutions using ToIP & related technologies. Recap last meeting's discussion. Continue the discussion. (I'll be calling from a car so I'm providing the necessary links here)
    1. The draft proposal is here: https://docs.google.com/document/d/1Snga-dkUG0XaLaXx1q9FKojaFy42CTZhvyFJbo9KgLU/edit. The proposal is flexible, and a start. 
      1. Describe the problem to help us all to have the same understanding. The group can come together to describe the problem. Trust on the data, processing and actors responsible for the data, processing and governance. Link from Bo - https://www.linkedin.com/posts/bo-harald-4768b51_connecting-ai-to-the-trust-infrastructure-activity-7125393724984991745-wleY?utm_source=share&utm_medium=member_desktop.
      2. Requirements for a solution. 
      3. The solution can have components.
        1. Use protocols/tools of TOIP, C2PA, 
        2. Some aspects of the solutions an be standardized.
        3. How can SSI help AI applications?
        4. Ecosystem trust that needs to be traceable. Entire workflow so people can drill into the details. To trust AI similarly the entire process. The actors. An annual check. Audit. 
        5. From Bo 

          The Trust infrastructure is naturally the foundation also for better AI
          How could SSI support AI applications?

          Self-Sovereign Identity (SSI) can be immensely supportive for AI applications in various ways:

          1. **Data Integrity and Authenticity**: SSI ensures that the data shared with AI systems is authentic and unaltered, which is crucial for AI models that rely on accurate data for training and decision-making.

          2. **Personalized Experiences**: With SSI, AI applications can offer personalized services without compromising users' privacy. Users can selectively share information, allowing AI systems to tailor experiences based on the provided data.

          3. **Secure Data Access**: AI applications, especially in sensitive sectors like healthcare, need secure access to data. SSI can provide a mechanism for users to grant time-limited access to specific data, ensuring AI systems can function without perpetual access to personal data.

        6. . **Reduced Data Bias**: By giving individuals control over their data, SSI can help in collecting diverse datasets that are more representative of the real world. This can aid in reducing biases in AI models.

          5. **Trust and Transparency**: Users might be more willing to engage with AI applications if they know their data is managed through SSI, enhancing trust and transparency.

          6. **Regulatory Compliance**: AI applications often operate within stringent regulatory environments, especially concerning data privacy (e.g., GDPR). SSI can aid in ensuring that data is processed and accessed in compliance with such regulations.

          7. **Decentralized AI Models**: With the emergence of decentralized AI, where models are trained across multiple devices without centralizing data, SSI can provide a framework for secure and authenticated data sharing.

          Pipe dream to see AI services list verified data per strength class it is using - should it be mandatory?
    2. Minor edits highlighted in blue to clarify that we can standardize some aspects of our work as a formal spec should we found that useful. I personally do found several such aspects very much worthy of formal specifications, and urgently, but I'll leave this topic open for others to comment on.
    3. For a technical background of the approach I'm proposing, see this technical paper. or this video presentation.
  2. AI Practice: Roll out a ChatGPT/RAG-like tool for ToIP & other communities.
    1. Savita Farooqiu, Steven Milstein 
    2. Propose to...Maybe it's time to get the ToIP knowledge base (all documentation, recordings, etc.) put into a private LLM such that ToIP members can a) get higher value from the collective expertise and b) be a generative AI sandbox where we can fully understand what effective AI can do.
  3. A blog series to (goals...)
    1. Propose to...
  4. Other ideas welcome...
  5. Write a page on a proposal to provide a Chat GPT like tool 


0 mins
  • Review decisions/action items
  • Planning for next meeting 
  • AOB
Chairs


  • Keep the same time slot (and discuss on Jan 18 if there're needs to adjust)