Logan Graham

@logangraham

PhD student in Machine Learning. Into comp. intelligence + econ + sustainability + human flourishing. Prev causality ,

SF Bay Area & Oxford, UK
Vrijeme pridruživanja: lipanj 2009.

Tweetovi

Blokirali ste korisnika/cu @logangraham

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

  1. Prikvačeni tweet
    22. stu 2017.

    I curate a weekly list of 10+ great Machine Learning papers submitted in the past week. I'll add each edition to my site from now on: – now featuring fun news.

    Poništi
  2. 1. velj

    More from my brain: unconventional institutions I would like to exist. If anyone has the money, they should build them.

    Poništi
  3. 1. velj

    The day we get there, the human project will have succeeded.

    Prikaži ovu nit
    Poništi
  4. 1. velj

    I spoke with a large handful of applicants around the world. I am reaffirmed in my belief that the world is full of incomparably good people. Humanity's measure of success is how well we raise them up to where they need to be to fully express their hearts and talents.

    Prikaži ovu nit
    Poništi
  5. 26. sij

    I need to print this out and sit in a coffee shop all day and read it. There's got to be a lot of other nuggets in there I've missed.

    Prikaži ovu nit
    Poništi
  6. 26. sij

    For a bunch of people whose jobs are to speculate wildly on unspecified, hard science, they do an amazing job at either (a) having their finger on the future's pulse or (b) confirming my biases.

    Prikaži ovu nit
    Poništi
  7. 26. sij

    Last, I completely missed their other moonshots in machine learning. As ever, I'm impressed by their ability to nail down the machine learning frontiers we're just grappling with: robustness, weak/semi supervision, theory (rather than alchemy), lifelong learning.

    Prikaži ovu nit
    Poništi
  8. 26. sij

    Then there's the moonshot of Automating Science. If we found a general way to do this, it might end up being the most prosperity-producing tool invented in this century. Imagine dark factories (orbiting in space, even!) that report back ever so often with new science.

    Prikaži ovu nit
    Poništi
  9. 26. sij

    Similarly, they want to figure out build a system for learning to model *any* system without bias. This is the meta-problem, named "Big Mechanism". It's like recreating the process of how humans come to understand systems – our multi-hundred year journey of science.

    Prikaži ovu nit
    Poništi
  10. 26. sij

    Modeling the World – their program is literally called "World Modelers". Humanity is a bunch of complex intersecting systems – economic, transportation, information, biological. We should build causal simulations of them all coming together. (right: focus on social systems)

    Prikaži ovu nit
    Poništi
  11. 26. sij

    I've told a few people that two of my personal moonshots are Modelling the World and Automating Science. Good news: DARPA is trying to do both in the next few years --

    Prikaži ovu nit
    Poništi
  12. 26. sij

    Have to retweet with comment because I JUST discovered a bunch of other cool things hidden in this budget! (thread)

    Prikaži ovu nit
    Poništi
  13. 26. sij

    Continuing the theme of institutions for innovation, from : 'what does “McKinsey for the long tail” look like and how do we make it a magnet for the best and brightest?' i.e. how do we bring the many small firms to the productive frontier?

    Poništi
  14. 23. sij

    I literally just discovered what this button does. The future is here.

    Prikaži ovu nit
    Poništi
  15. 23. sij

    Whoops, continuation of the thread here:

    Prikaži ovu nit
    Poništi
  16. 23. sij

    A DARPA budget may not be as engaging as reading a science fiction book, but it's just as fun.

    Prikaži ovu nit
    Poništi
  17. 23. sij

    This foray (from last night) was initially inspired by 's excellent podcast (Idea Machines) interview with

    Prikaži ovu nit
    Poništi
  18. 23. sij

    Overall impressions: 1) The budget increase overall was smaller from 2019->2020 than 2018->2020 2) In a lot of areas the budget is decreasing, possibly from program maturity 3) DARPA is really thinking about what new technology would radically shift the equilibrium

    Prikaži ovu nit
    Poništi
  19. 23. sij

    The hypersonic arms race may be real (cc Russia)

    Prikaži ovu nit
    Poništi
  20. 23. sij

    Some cool things from the AI strategy: explainable AI (DARPA has been several years ahead of academia n spotting this), hypothesis generation, translation for undertranslated languages (cc ), human-machine symbiosis, accelerating adoption

    Prikaži ovu nit
    Poništi
  21. 23. sij

    We're going hard on spotting existential risks / WMD use early in the body:

    Prikaži ovu nit
    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:

    ·