1. X
  2. Wilka Carvalho
Log inSign up
Wilka Carvalho
3,505 posts
Wilka Carvalho profile banner
user avatar

Wilka Carvalho

@cogscikid
@KempnerInst research fellow @Harvard. I want to build AI that empowers the human reinforcement learning algorithm. prev: deepmind, umich, msr.
Cambridge, MA
cogscikid.com
Joined November 2015
942
Following
1,770
Followers
RepliesRepliesMediaMedia
  • Pinned
    user avatar
    Wilka Carvalho
    @cogscikid
    Jun 2
    Task diversity is supposedly key to generalization in RL. But what does it do to continual RL, where agents face one new task distribution after another? We find that past a point, more diversity actually inhibits continual reinforcement learning 🧵
    Image
    00:00
  • user avatar
    Wilka Carvalho
    @cogscikid
    Aug 12
    I can’t help but contextualize these results in the normative values of most people I meet at Anthropic where humans are mere components of AI systems that solve tasks as opposed to AI systems being fundamentally to empower human cognition and behavior I wonder how much this
    user avatar
    ClaudeDevs
    Anthropic
    @ClaudeDevs
    Aug 7
    Replying to @ClaudeDevs
    One reason we trust it more than manual approval: in a study with 1,053 paid testers, we swapped a permission prompt for a clearly dangerous command (text only, nothing actually ran). Testers caught it 13.6% of the time, and closer to 5% after 50 prompts. Auto mode blocked the
    Image
  • user avatar
    Wilka Carvalho
    @cogscikid
    Aug 7
    haha he’s definitely right
    user avatar
    Gabriele Berton
    @gabriberton
    Aug 6
    This is becoming ridiculous Trying to rewrite history after bashing on LLMs for years I gave Francois's post to Gemini 3.1 Pro, GPT 5.6 Sol, Fable, and they ALL say he's wrong (screenshots in next post) [1/3]
    Image
  • user avatar
    Wilka Carvalho
    @cogscikid
    Aug 5
    we are but humble servents
    Image
    Image
    Image
    Image
    user avatar
    Amanda Long
    @_amanda_long
    Aug 4
    Avital Balwit is Chief of Staff to Anthropic’s CEO. She reflects on what it’s like to be one of a select few people shaping frontier AI systems and the moral gravity of what they half-jokingly call “building God.”
  • user avatar
    Wilka Carvalho
    @cogscikid
    Aug 4
    "half-jokingly"
    user avatar
    Amanda Long
    @_amanda_long
    Aug 4
    Avital Balwit is Chief of Staff to Anthropic’s CEO. She reflects on what it’s like to be one of a select few people shaping frontier AI systems and the moral gravity of what they half-jokingly call “building God.”
    Image
    Image
    Image

Log in or sign up for X

See what’s happening and join the conversation

Continue with phone
or
Log in with username or email
Terms·Privacy·Cookies·Accessibility·Ads Info·© 2026 X Corp.
Advertisement
Advertisement