Zuckerberg pitches open AI as check on 'closed' rivals, Meta releases new open model
- Mark Zuckerberg published an essay arguing powerful AI should not be concentrated in the hands of a few companies or governments, positioning it as a case for freely available AI reaching "billions of people."
- Meta, valued at $1.5tn, released the underlying parameters of a new open model called Muse Glimmer that outside developers can download and modify.
- The essay casts OpenAI and Anthropic as foils, framing their closed approach to AI as the problem Meta's open strategy is meant to counter.
- Meta said it plans to release additional open models beyond Muse Glimmer, continuing its shift back toward open weight releases after periods of tighter control.
- Hacker News commenters largely read the move as strategic rather than principled, noting Meta's ad based business benefits more from commoditizing AI than from selling access to it directly.
Hacker News opinions
Zuckerberg's essay reads like a contradiction: he opens by saying he wants to empower people with AI, then pitches an agent that needs all your personal context to 'work for you 24/7'. I'm the one who directs my own life, that's my job, not some agent's.
Feels like classic 'I'm losing so I want to change the rules' behavior, same as Altman saying 'we should slow down' when he wasn't in the lead anymore.
This is also about redefining what 'open' means so Meta can dodge responsibility for scraping everyone's data. Not that there was ever a real legal risk, but it helps.
It's a long term pattern: when you're winning like Anthropic you keep things closed and monetize hard, when you're losing like Meta you open it up or push it into a standards body to devalue the leader's advantage. Then you get to look altruistic on top of it.
Makes total strategic sense though: Meta's business is ads, not selling raw intelligence, so they want AI commoditized and freely available so people just consume more content and ads through it.
I think the real endgame is selling compute: open the models so everyone runs them, then charge people to run those models on Meta's own hardware.
Honestly 'I'm losing so I'll try something else' is just rational market behavior and we should want more of that, it's 'I'm winning so I'll pull up the ladder' that's the actual problem.
The timing is suspicious too, Meta just got fined by Mexico a few days before this announcement, feels like a PR move to offset bad news.
Genuine question: isn't this just unambiguously good regardless of Zuckerberg's motives? More open weights and more competition benefits everyone, what am I missing here?
Most people probably agree it's good in isolation, they just don't want to give Zuckerberg credit given Meta's track record, plus everyone knows the real goal is commoditizing OpenAI and Anthropic, not principle.
Meta's open model releases might honestly be the best thing Facebook has ever done, it let Chinese labs catch up and between them they're keeping AI pricing from becoming a monopoly.
I read it as Mark deliberately burying a self-serving claim in dense text that almost nobody will actually read closely, banking on people just believing the headline version of 'Zuckerberg defends open AI'.
This smells like a setup for a future rug pull, Meta cynically waves the open source flag when it's convenient for marketing, then changes course later like they always do.