Meta's Lucas Beyer says strict corporate policies prohibit industry AI researchers from sharing unpublished technical details, setbacks, or future directions

@giffmana The reply link is broken — it routes to an empty / missing attachment. Hard to add real take without seeing the feedback.

@elvisnavah I did write this with mostly the reader who has never been at or dealt with bigco in mind. So add student to your list.

@giffmana @ECLresearch new meta for bots, to drive engagement, is to say the reply link doesnt work

@giffmana @ECLresearch i think that might be a bot

@ECLresearch I don't understand?

@ECLresearch Oh you mean you can't see the screenshot? Odd, works for me both on mobile and on desktop.

@giffmana random question since i wasn't here during the google brain era, was it the same for "unpublished directions"? like you were only able to talk about published work, it's just that there was much more published work obviously?

@giffmana the only person who could ask you something like this is either unemployed or is a founder (very little difference)

Yeah it’s always easier to talk about other companies. And if you work for a publicly traded company (or one headed that way) there’s a load more stuff you could trip over accidentally. One persons memey shitpost is anothers’ market manipulation attempt… Also complaining about colleagues work on X feels mean since I could raise most things more helpfully internally

@giffmana Are you also not allowed to bashing /commenting on the whole meta data labeling drama lmao.

@giffmana I never thought people actually use anon feedback for anything constructive and not only "oho you're not funny"

@giffmana X used to be useful. You could actually find good articles and interesting ideas. Now everyone just posts boring content, so the whole platform feels completely dead.

@giffmana @ECLresearch It works.

@giffmana Clarify specific point please.

@KKumar_ai_plans @ECLresearch yeah I realized just after, then report+block.

@giffmana Exactly why open-source AI matters. Corporate approval doesn't just filter what gets published, it shapes what directions the field can even discuss. We all lose.