Log inSign up
Eric J. Michaud
328 posts
Eric J. Michaud profile banner
@ericjmichaud_

Eric J. Michaud

@ericjmichaud_
Trying to make deep neural networks among the best understood objects in the universe. 💻🤖🧠👽🔭🚀
Berkeley
ericjmichaud.com
Joined February 2015
1,320
Following
4,294
Followers
RepliesRepliesRepostsRepostsMediaMedia

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.
  • Pinned
    @ericjmichaud_
    Eric J. Michaud
    @ericjmichaud_
    Jan 13
    How does scaling up neural networks change what they learn? Despite its importance, our understanding of this question remains nascent. I've written a long post reflecting on my model of neural scaling and its relationship to interpretability, etc.: ericjmichaud.com/quanta
    00:00
    38
  • @ericjmichaud_
    Eric J. Michaud
    @ericjmichaud_
    Aug 31
    it is time-consuming work to wrangle references during the paper writing process. but it is also very satisfying to piece together the whole eclectic network of people who have influenced your thinking, and to feel and communicate a gratitude for them and their ideas.
    2
  • @ericjmichaud_
    Eric J. Michaud
    @ericjmichaud_
    Aug 27
    One source of optimism in interpretability was that we had full access to neural network internals, unlike neuroscience. And indeed the field has moved fast! But I still think we could spend 100+ years thinking about this stuff at the ~current pace of progress...
    @banburismus_
    Tom McGrath
    @banburismus_
    Aug 19
    a surprisingly underappreciated thing when thinking about potential short timelines: even if most of the calendar time to do something useful is in the past, most of the cognitive time is in the future I've seen this perspective in a few places but never about interp
    3
  • @ericjmichaud_
    Eric J. Michaud
    @ericjmichaud_
    Jul 24
    I gave a lightning talk at the mechanistic interpretability workshop at ICML this year. Thanks to the organizers, who encouraged us to share our high-level takes; my talk ended up being quite high-level and personal indeed. Blog post version: ericjmichaud.com/interp-purpose
    7
  • @ericjmichaud_
    Eric J. Michaud
    @ericjmichaud_
    Jul 17
    captures the vibe of interpretability nicely!
    @covacut
    cova
    @covacut
    Jul 13
    how does AI think? ✨
    00:00
    2