Upgrade to Pro — share decks privately, control downloads, hide ads and more …

AI Governance

Ian Mulvany
November 28, 2023

AI Governance

Ian Mulvany

November 28, 2023
Tweet

More Decks by Ian Mulvany

Other Decks in Research

Transcript

  1. Ian Mulvany - CTO BMJ
    Implementing AI Governance
    BMJ Case Study

    View full-size slide

  2. Goal of this talk:
    What actions can each of us take now?
    How might we best share what we are learning?

    View full-size slide

  3. Things are strange

    View full-size slide

  4. How might we make them
    less so?

    View full-size slide

  5. 1. 12 Months since ChatGPT became available
    2. LLMs predictably get more capable with increasing investment, even without targeted innovation.
    3. Many important LLM behaviors emerge unpredictably as a byproduct of increasing investment.
    4. LLMs often appear to learn and use representations of the outside world.
    5. There are no reliable techniques for steering the behavior of LLMs.
    6. Experts are not yet able to interpret the inner workings of LLMs.
    7. Human performance on a task isn’t an upper bound on LLM performance.
    8. LLMs need not express the values of their creators nor the values encoded in web text.
    9. Brief interactions with LLMs are often misleading.
    LLMs - 9 Things to consider!
    https://cims.nyu.edu/~sbowman/eightthings.pdf

    View full-size slide

  6. 1. The pace of iteration is insane
    2. LLMs predictably get more capable with increasing investment, even without targeted innovation.
    3. Many important LLM behaviors emerge unpredictably as a byproduct of increasing investment.
    4. LLMs often appear to learn and use representations of the outside world.
    5. There are no reliable techniques for steering the behavior of LLMs.
    6. Experts are not yet able to interpret the inner workings of LLMs.
    7. Human performance on a task isn’t an upper bound on LLM performance.
    8. LLMs need not express the values of their creators nor the values encoded in web text.
    9. Brief interactions with LLMs are often misleading.
    LLMs - 9 Things to consider!

    View full-size slide

  7. 1. The pace of iteration is insane
    2. They are going to get more powerful
    3. Many important LLM behaviors emerge unpredictably as a byproduct of increasing investment.
    4. LLMs often appear to learn and use representations of the outside world.
    5. There are no reliable techniques for steering the behavior of LLMs.
    6. Experts are not yet able to interpret the inner workings of LLMs.
    7. Human performance on a task isn’t an upper bound on LLM performance.
    8. LLMs need not express the values of their creators nor the values encoded in web text.
    9. Brief interactions with LLMs are often misleading.
    LLMs - 9 Things to consider!

    View full-size slide

  8. 1. The pace of iteration is insane
    2. They are going to get more powerful
    3. We don’t how, they just are
    4. LLMs often appear to learn and use representations of the outside world.
    5. There are no reliable techniques for steering the behavior of LLMs.
    6. Experts are not yet able to interpret the inner workings of LLMs.
    7. Human performance on a task isn’t an upper bound on LLM performance.
    8. LLMs need not express the values of their creators nor the values encoded in web text.
    9. Brief interactions with LLMs are often misleading.
    LLMs - 9 Things to consider!

    View full-size slide

  9. 1. The pace of iteration is insane
    2. They are going to get more powerful
    3. We don’t how, they just are
    4. They look like they are intelligent
    5. There are no reliable techniques for steering the behavior of LLMs.
    6. Experts are not yet able to interpret the inner workings of LLMs.
    7. Human performance on a task isn’t an upper bound on LLM performance.
    8. LLMs need not express the values of their creators nor the values encoded in web text.
    9. Brief interactions with LLMs are often misleading.
    LLMs - 9 Things to consider!

    View full-size slide

  10. 1. The pace of iteration is insane
    2. They are going to get more powerful
    3. We don’t how, they just are
    4. They look like they are intelligent
    5. We don’t know how to control them
    6. Experts are not yet able to interpret the inner workings of LLMs.
    7. Human performance on a task isn’t an upper bound on LLM performance.
    8. LLMs need not express the values of their creators nor the values encoded in web text.
    9. Brief interactions with LLMs are often misleading.
    LLMs - 9 Things to consider!

    View full-size slide

  11. 1. The pace of iteration is insane
    2. They are going to get more powerful
    3. We don’t how, they just are
    4. They look like they are intelligent
    5. We don’t know how to control them
    6. We don’t know how they work
    7. Human performance on a task isn’t an upper bound on LLM performance.
    8. LLMs need not express the values of their creators nor the values encoded in web text.
    9. Brief interactions with LLMs are often misleading.
    LLMs - 9 Things to consider!

    View full-size slide

  12. 1. The pace of iteration is insane
    2. They are going to get more powerful
    3. We don’t how, they just are
    4. They look like they are intelligent
    5. We don’t know how to control them
    6. We don’t know how they work
    7. They will be better than us at things
    8. LLMs need not express the values of their creators nor the values encoded in web text.
    9. Brief interactions with LLMs are often misleading.
    LLMs - 9 Things to consider!

    View full-size slide

  13. 1. The pace of iteration is insane
    2. They are going to get more powerful
    3. We don’t how, they just are
    4. They look like they are intelligent
    5. We don’t know how to control them
    6. We don’t know how they work
    7. They will be better than us at things
    8. They are a tool, and will be used by those who use them
    9. Brief interactions with LLMs are often misleading.
    LLMs - 9 Things to consider!

    View full-size slide

  14. 1. The pace of iteration is insane
    2. They are going to get more powerful
    3. We don’t how, they just are
    4. They look like they are intelligent
    5. We don’t know how to control them
    6. We don’t know how they work
    7. They will be better than us at things
    8. They are a tool, and will be used by those who use them
    9. You need to spend time with them
    LLMs - 9 Things to consider!

    View full-size slide

  15. 1. Stay the same
    2. Get incrementally better
    3. Get dramatically better
    4. Become sentient

    View full-size slide

  16. Customer
    Context
    Certainty
    Disruptiveness
    Language
    Generation
    How we
    interact with
    information
    Workflow /
    Scaling
    Value of
    high quality
    data

    View full-size slide

  17. Let’s Get Boring

    View full-size slide

  18. Kind of task Risk Level
    Autonomous
    decision, high stakes
    domains - self
    driving, clinical
    diagnosis, Peer
    review
    Expert Augmentation,
    coding, legal drafting,
    scholarly text analysis
    Mechanical tasks:
    document analysis,
    copywriting, code
    cleanup
    Danger Danger Will
    Robinson!, stay
    very far away
    Understand,
    explore, exploit
    Just do this
    already!

    View full-size slide

  19. Policy GenAI

    View full-size slide

  20. Governance
    Data Workflow Output
    The risk is our reputation

    View full-size slide

  21. Editorial Users Governance Reputation Legal Tech
    Governance
    Expertise

    View full-size slide

  22. Example use cases inside BMJ

    Software development

    User research

    SAP management

    Report analysis

    Copy editing

    View full-size slide