Joe Faith

@joedotfaith

Ex-Googler now using and to catch money launderers. CPO at . PhD in Philosophy and AI. Tweeting about , , and

Vrijeme pridruživanja: svibanj 2012.

Tweetovi

Blokirali ste korisnika/cu @joedotfaith

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @joedotfaith

  1. 5. velj
    Prikaži ovu nit
    Poništi
  2. proslijedio/la je Tweet
    5. velj

    The biggest difference between statistics and machine learning may be in language! So a few months ago I created this (inspired by ) but haven't made much progress since. Welcoming suggestions for improvements

    Poništi
  3. 4. velj

    Lesson 3/3: ML is opaque. Regulators, including and EBA, mandate effective oversight of models. Banks are accountable for their models, so must be able to justify their decisions. This means all decisions must be auditable, traceable, and explainable.

    Prikaži ovu nit
    Poništi
  4. 4. velj

    Lesson 2/3: ML is brittle. Aggregate performance metrics can hide issues on high risk minority cases (eg premium customers or risky locales). Evaluating risk properly requires schema completion and stress testing to understand operational safe limits.

    Prikaži ovu nit
    Poništi
  5. 4. velj

    Lesson 1/3: Data is a bottleneck. We don't have enough data on verified cases, esp. new threats like , so pure-data approaches miss significant risks. The solution is to encode the knowledge of domain experts -- don't wait until the threats appear in the data.

    Prikaži ovu nit
    Poništi
  6. 4. velj

    $4tn of criminal proceeds flow through banks every year. One third of banks use to detect it, and a third more are running pilots. So what can learn from about building we can trust to protect us? Three lessons...

    Prikaži ovu nit
    Poništi
  7. proslijedio/la je Tweet
    3. velj

    ML papers that report model performance really need to start including uncertainty. An easy way to do this is once you're ready to publish, train the model ~10 times with different random seeds. Evaluate each model and report the mean and variance of the results.

    Poništi
  8. proslijedio/la je Tweet
    31. sij

    "A developed country is not a place where the poor have cars. It's where the rich use public transportation." — (Photo: )

    Poništi
  9. proslijedio/la je Tweet
    4. sij 2018.

    Good science ends when you become more attached to your solution than to the problem.

    Poništi
  10. proslijedio/la je Tweet
    3. velj

    Super excited to do a short Hunting for Monsters in Open Data workshop - building on work with into Monsters, Metaphors and ML to be presented at

    Poništi
  11. proslijedio/la je Tweet
    2. velj

    good tech can’t fix bad policy

    Poništi
  12. proslijedio/la je Tweet
    2. velj

    Become a better communicator overnight: - Replace every “but” you can with an “and” - Never start a sentence with “no” or “disagree” - Only give feedback if there’s a chance it will be implemented - Tell people you will be late before you are late - Bias to overcommunicating

    Poništi
  13. 3. velj

    The point of a PhD is not to gain knowledge, it's learning how to do rigourous, original, reproducible research. The content of most PhDs are forgotten as soon as you submit them. But that's ok. The content is just the proof. The value is in the skills you carry for a lifetime.

    Poništi
  14. 2. velj

    The grizzled veteran cuts them dead. 'They're nothing til I see 'em'

    Prikaži ovu nit
    Poništi
  15. 2. velj

    The experienced ref says 'I call them as I see them' 3/4

    Prikaži ovu nit
    Poništi
  16. 2. velj

    The noob says 'I call them as they are 2/4

    Prikaži ovu nit
    Poništi
  17. 2. velj

    Daniel Dennett used to sum up various epistemological stances by reference to three baseball umpires. 1/4

    Prikaži ovu nit
    Poništi
  18. proslijedio/la je Tweet
    1. velj

    If you want a problem solved make it someone’s project. If you want it managed make it someone’s job.

    Prikaži ovu nit
    Poništi
  19. proslijedio/la je Tweet
    1. velj

    We’re categorically ignorant of anything beyond our consciousness. That explains the epistemic gap between experiences and their neural mechanisms.

    Poništi
  20. 1. velj

    Guide to doing effective research. Key takeaway: focus on goal-driven research, rather than ideas-driven, but look for generalisable solutions. Necessity is the mother of invention. Concrete problems often demand novel solutions.

    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·