Fraser

@FraserGreenlee

Pioneer working on the future of programming Previously

Edinburgh, Scotland
Joined August 2016

Tweets

You blocked @FraserGreenlee

Are you sure you want to view these Tweets? Viewing Tweets won't unblock @FraserGreenlee

  1. Pinned Tweet
    29 Sep 2019

    Just made a reinforcement learning agent that teaches itself a small part of Python! It does this by typing into a REPL and trying to create suprising outputs.

    Show this thread
    Undo
  2. Retweeted
    Jan 31

    If you read the old Y Combinator stories about the first few batches and wondered how it felt like, I suggest joining Pioneer. It’s only 100 of us right now, and it will be big in the future (maybe bigger than YC?)

    Show this thread
    Undo
  3. Retweeted
    Jan 30

    Introducing Pioneer 2.0! A standard investment deal, more funding options, and a fresh new website. (I'm especially in love with the animated Harry Potter faces.) Join:

    Undo
  4. Retweeted
    Jan 24
    Show this thread
    Undo
  5. Retweeted
    Jan 22

    Introducing the first 12 Pioneers of 2020. Credit card points for sustainability, Twitter as a CRM, Minecraft-meets-notes, and more:

    Undo
  6. Retweeted
    24 Dec 2019
    Undo
  7. Retweeted
    24 Dec 2019
    Show this thread
    Undo
  8. Retweeted
    15 Dec 2019

    My blog post on how to win just dropped! Make sure to share it with people currently in the tournament ;)

    Show this thread
    Undo
  9. Retweeted
    Undo
  10. Retweeted

    Stuðlagil Canyon in Iceland looks like another planet. Credit: Hörður Kristleifsso

    Undo
  11. Retweeted
    8 Dec 2019

    "Singles Day" achieved the equivalent sales volume of Black Friday in less than 10 minutes. I wrote about Alibaba’s Hong Kong IPO for !

    Undo
  12. Retweeted

    In Silicon Valley, you don't say "a lot" (as in "a lot bigger"), you say "10x" "It's raining 10x today"

    Show this thread
    Undo
  13. Retweeted
    26 Nov 2019

    Introducing the SHA-RNN :) - Read alternative history as a research genre - Learn of the terrifying tokenization attack that leaves language models perplexed - Get near SotA results on enwik8 in hours on a lone GPU No Sesame Street or Transformers allowed.

    The SHA-RNN is composed of an RNN, pointer based attention, and a “Boom” feed-forward with a sprinkling of layer normalization. The persistent state is the RNN’s hidden state h as well as the memory M concatenated from previous memories. Bake at 200◦F for 16 to 20 hours in a desktop sized oven.
    The attention mechanism within the SHA-RNN is highly computationally efficient. The only matrix multiplication acts on the query. The A block represents scaled dot product attention, a vector-vector operation. The operators {qs, ks, vs} are vectorvector multiplications and thus have minimal overhead. We use a sigmoid to produce {qs, ks}. For vs see Section 6.4.
    Bits Per Character (BPC) onenwik8. The single attention SHA-LSTM has an attention head on the second last layer and hadbatch size 16 due to lower memory use. Directly comparing the head count for LSTM models and Transformer models obviously doesn’tmake sense but neither does comparing zero-headed LSTMs against bajillion headed models and then declaring an entire species dead.
    Undo
  14. Retweeted

    There is no qualitative difference between "interviewing" a language model by using it to generate lots of answers then selecting those you like, & the TV "interviews" of Sophia the Robot. In both cases we're looking at a puppet, and projecting human intentions & emotions on it.

    Show this thread
    Undo
  15. 24 Nov 2019

    Not sure on the Cyvertruck hype, how many Ford F-150 fans are going to buy it? Seems like it's only Tesla fans ordering it.

    Undo
  16. Retweeted

    The Bus Ticket Theory of Genius:

    Undo
  17. Retweeted
    19 Nov 2019

    "When there's speech that is completely forbidden and questions that are not allowed to be asked, normally, you should assume that those things are simply true." - Peter Thiel

    Undo
  18. Retweeted
    20 Nov 2019

    And we're live on ! EasyDB is the fastest way to provision a database. Period. Head on over and checkout what and I have been hacking on for the past ~2 weeks.

    Show this thread
    Undo
  19. Retweeted
    19 Nov 2019

    So many thoughts! -It is way past time for U.S. cities to prioritize walkability. -It’s especially cool that Culdesac chose Tempe to start since we know that most growth in cities is to the mid-size market. -The team at Culdesac could not be more stellar. Very bullish ❤️

    Undo
  20. 12 Nov 2019
    Undo
  21. Retweeted
    9 Nov 2019

    Drone Animation in feels like Blade Runner and Ghost In The Shell Via

    Undo

Loading seems to be taking a while.

Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.

    You may also like

    ·