▲ 641 points
back
600 comments
I think something that doesn't get said enough is Meta did, albeit intentionally kick off the origin of the open source race back in 2023 with the release of llama.
I'm not a big fan of meta in general, but they've done enough good, and it's possible that it was intentional as well. I don't know, I wasn't in the rooms, and I think it's worth giving them some reasonable doubt.
No one is purely good, and no one is purely evil. This is net good regardless.
Disclaimer, I work on Gemma and open models at Deepmind and the opinions here are my own
There were open models from EleutherAI (GPT-Neo), Google Brain (T5X, Bert), and HuggingFace was promoting open models (and others doing open work I haven't listed here) all prior to 2023 and the big Chatgpt moment.
https://github.com/EleutherAI/gpt-neo/releases
https://github.com/google-research/bert
https://github.com/google-research/t5x
If you're learning about AI models it's still worthwhile to review these models and codebases because they continue to be the basis of the technology that's being produced today! It'll give you a good perspective of how things have changed, similar to say learning about propeller planes before moving onto modern jet engines.
That's simply not true. The reason why llama is open source is simply because it got leaked, then llama.cpp was the real game changer which was built from the ground up in depressingly short amount of time. Meta had no choice but to take the L and "support" the open source community. The angry "I-hate-you-and-I-hope-you-die" kind of support.
Open source != open weight.
Big difference and it bugs me that nobody seems to care about using the right words in only this context.
Enough good? You are joking, right?
They also kick off a ton of other nefarious things we are still paying for.
> albeit intentionally
The model was "accidentally" released.
Meta was giving it to approved researchers only until someone leaked a torrent. Whether that was a researcher, an insider, or Meta's plan all along, we don't know.
Meta has withheld its best models, as have a lot of other "open weights" Chinese companies. When an "open weights" company gets ahead in one domain or modality, they tend to start withholding their releases. Tencent, for instance, began withholding their Hunyuan models once they became competitive. Alibaba has done the same.
The "open weights" strategy for the majority of players is this: open source when you're not in first place. Use the ecosystem to poison your rival's margins and play catch up on distribution.
In the West, it tends to take on yet another hook: "shareware weights until you pass $1M ARR, then you must license." See Flux, K2, etc.
The only way for open weights to make sense financially is if you have another income stream and are dumping on the market to destroy competition and/or can get people into using your inference infra / product ecosystem / tooling. Nobody's cracked this yet.
It was leaked which put it in the open, it got widely popular and they rode the wave. I'm not so such if they would have widely released it if it wasn't leak. Nevertheless mucho credits to them for following up with llama2, llama3, llama4 and now muse.
I have mixed feelings along these lines, I know meta have contributed to various open projects, sometimes Mark pays lip service to “the open internet" while his company represents a constilation of walled gardens. I think PHP got some love, and React is an industry goto (I'm more of a PHP... -> Svelte guy) but are these contributions worth what happened in Myanmar? I say no.
Ha-ha they released nothing. It got leaked and without the source for it. Totally wrong to portray them as benevolent benefactors to the ML race
>> No one is purely good, and no one is purely evil.
I will make an exception for Musk and DODGE
My favorite paragraph from Zuckerberg's writeup:
""" [...] it is surprising that the discourse from many developing AI is so filled with doom. I do not understand why anyone who believes that AI will eliminate most jobs and much of humanity's relevance would rush to build that future. The notion that AI is so dangerous that the only safe path is an extreme concentration of power seems inherently problematic. Historically, hoping that an absolute power will benevolently provide for humanity if sufficiently enlightened has not led to safe or positive outcomes. """
Comments here are surprising to me.
I get folks don’t like Zuckerberg and his company and don’t trust his intentions… I don’t either.
But this is an unquestionably good thing right?. The more open source software out there the better. And the more open weights or even over source AI stuff the better too right? More competition the better generally speaking I think.
Unless I’m missing something and am getting this whole situation wrong. Please let me know if I am.
Is this "I'm losing so I think we should change the rules"? Because it seems like that.
Is going to take it one day at a time, on being a less evil bilionaire :-)
"Zuckerberg faces questions over why superyacht reportedly declined to help stranded boat" - https://www.theguardian.com/us-news/2026/aug/09/zuckerberg-s...
If you believe, as an increasing number of folks do, that LLMs are basically commoditized at this point or will be soon then there really isn’t a path forward for non open models.
There simply isn’t any value in the models and only in the compute to run them and potentially some higher-order coordination layers.
That’s very bad for the big model labs that banked their future on the opposite, but there just doesn’t seem to be a viable path forward on that approach anymore.
Meta hasn’t been the best global citizen for most of its life, but there may be some mild redemption if it becomes a source of strong open models.
The relevant paragraph about Meta’s commitment to open source is here, and the statement is significantly less confident than news is reporting
> • Open source is a positive and important force for empowering people and preventing centralization that is detrimental for both safety and the economy. Meta continues to be strongly supportive of open source, including open source AI models. The current open source ecosystem is strong, and we think it would be a mistake to restrict it. Now that Meta Superintelligence Labs are up and running, we will resume releasing some open source models soon.
“We will resume releasing some open source models soon”??? That’s basically the most ambiguous non commitment ever. Even before legal team watering it down I can’t imagine what the intent is here. We are also going to release open models for things maybe or maybe not kinda? Absolutely zero conviction. And moreover, this explicitly frames the open source ecosystem as something Meta is not a part of.
They launched their model a week ago, completely closed, selling an endpoint. Then when nobody bought, they "open" sourced it. Ok.
I have to ask, is this "returns to open models" even correct, considering from a quick look the facebook huggingface collections[0] has been actively releasing different kind of models almost monthly for long time now, possibly even some actual open source ones (looks like Sam 3D had a dataset collection, but didnt look more into it).
Maybe they are not as noteworthy as the recent release, maybe they are too niche for general audience, but its not like they just stopped releasing open weight things and this is their big return, unless i am missing something?
Is this like when they declared E2EE was the future and those not doing it were behind, but as it turned out bad for the bottom line¹² such features are getting rolled out of their properties?
U-turn when meta finds an inconvenience due to open weights in 3… 2… … …
--------
[1] reduced advertising revenues due to not being able to target based on message traffic, and reduced data scraped for training purposes
[2] and possibly in part due to behind-the-scenes pushback from government/police/other interests?
https://archive.is/p1ehR
The essay from Zuckerberg
Mark is not doing this for the money or power.
Mark genuinely cares about your mental health.
He does not want to rewire your brain to be dependent on his models, like he did with FB/IG.
We can see it in his empathic eyes.
The truth is all frontier models are closed. This isn't even an open source vs open weights thing. Even if you accepted that open weights are "open" the capital requirements for running your own Kimi 3 model are significant. It's not like gcc, where the binary just works well enough on random hardware.
> Everyone will have an exceptionally capable personal agent that understands you, your goals, and everything you care about.
"One machine for every man, woman and children of Zion. Sounds exactly like the thinking of a machine to me." Morpheus
It's a joke. OR IS IT? Yes it is, don't worry about it.
This paragraph is wild.
> As a thought experiment, imagine only one person had a superintelligent lawyer. They would have an unfair advantage in court -- even if they were wrong on the merits. That would lead to a worse society. But now imagine everyone has a superintelligent lawyer. In this case, justice would be carried out much more fairly and efficiently than it is today when there is often an imbalance in skills and resources in litigation.
This would only be true in a vacuum. Ignoring several leaps of progress where it would suddenly be allowed for an AI to represent one in court (imagine the size of the context window) and you could also call it from prison. That is not what is imbalanced about the US court system. Whoever has more money can drown someone in court filings and court fees and then drop their case financially ruining someone.
It's also funny how the interest of "AI for everyone" only extends to the national interest but I guess he's got to make sure he stays in good graces with the current administration.
I prefer open models. But in his essay, Zuckerberg's broader picture of the future is not one I want to embrace: everyone with a personal agent that has information on and involvement with every aspect of our lives, including our relationships and hobbies. Like, his daughter loves to bake, so his AI agent chooses recipes and orders ingredients - as if though those aren't enjoyable and meaningful parts of cooking.
Another example Zuckerberg gives is an LLM analyzing his sleep patterns. Do you need an AI to tell you if you're fucking exhausted every day? Is that analysis somehow going to give you more time to sleep?
It's a technocratic fairy tale founded on the idea that modern problems are driven by a lack of personal analysis rather than external pressures that are often systemic and out of our control. So far, LLMs have made a lot of those systemic issues worse: concentrating more wealth in a smaller number of people's hands, increasing consumer electricity and electronic prices, etc.
10 Aug 2026 14:49:05 UTC | Mark Zuckerberg Lays Out New AI Vision in 6,500-Word Essay | https://www.wsj.com/tech/ai/mark-zuckerberg-lays-out-new-ai-... | https://news.ycombinator.com/item?id=49244449
10 Aug 2026 17:15:20 UTC | Zuckerberg Posts 6,500-Word Essay About Giving Everyone AI Superintelligence | https://www.404media.co/mark-zuckerberg-posts-deranged-6-500... | https://news.ycombinator.com/item?id=49246721
10 Aug 2026 18:18:20 UTC | Zuckerberg makes a case for unleashing Pandora's box | https://www.theverge.com/tech/977395/meta-mark-zuckerberg-su... | https://news.ycombinator.com/item?id=49247557
Meta/Zuckerberg realized belatedly that AI models, which are currently in the process of rapidly becoming just a commodity product and that above all else is probably the main reason why Zuckerberg wrote that 6000 word term paper recently because he’s planning on pivoting away from Meta’s current money wasting AI approach.
He’s prepping Wall Street because Meta’s is going to layoff thousands of people in the coming months.
To give him credit, it's amazing that Mark Zuckerberg managed to take an easily defensible position like "we shouldn't allow AI to be owned by a couple of tech monopolies" and still manage to find so many individual points and arguments to be completely clueless on.
My personal favourite:
> Before the industrial revolution, 90% of people were farmers growing food to survive. Advances in technology steadily freed much of humanity to focus less on subsistence and more on the pursuits we choose.
Over a long enough time span, the industrial revolution has certainly improved living standards but the industrial revolution itself was l, for many, an absolutely horrific time to live. Life expectancies in cities plumetted and we saw the birth of workhouses.
It's such an odd example to choose for trying to suggest "technology x will help everyone in a non-complicated way". He might as well have picked the atomic bomb.
i was wondering why meta has such a hard time with llm development
its the organisational goal of that endeavour
they are doing it in a phase of firing people
so the goal of llm at meta is "to make people redundant" and no matter of HR/PR speak can change it
antrophic and openai have the goal "lets create the future"
whomever they hire and no matter how much money they spend on it, the goal alone will create very different outcomes.
What a load of insincere hogwash. If Zuck wanted to dilute the power of large institutions he has several billion ways to do it other than making limp gestures in the direction of open source. The problem is that even if you take him at his word here, he's only willing to do more good and not less harm.
As usual Zuckerberg is talking out of both sides of his mouth. On one hand he's releasing open-weight, commodity grade models while gating their frontier models behind facebook apps. You'll be able to have your cake, but won't be able to eat it.
Not a Meta fan per say but their contribution to open source has always been great. Examples include PyTorch (has had huge impact on velocity of ML research), Llama models, segment anything, many notable research papers from FAIR, etc. I never thought I would say this but I’m with Zuck on this. The call for slowing down AI development for safety seems like an excuse for reducing opex for training better models to stay relevant.
It's ironic Zuckerberg would criticize anyone for closing stuff up. Facebook being the classic walled garden. That said the enemy of my enemy etc.
Discussion on sources:
The Future Is for Everyone – The Path to a Positive AI Future
https://news.ycombinator.com/item?id=49241728
Meta Muse Glimmer – open weights 30B local coding model
From Zuck's own words "we will resume releasing some open source models soon"
Why did you pause releasing them in the first place? Why release only some?
Pushes for export controls on China who have released open weights on much more powerful models than Meta (and I believe that trend's not changing any time soon). I guess open models are bad if they're communist.
The entire essay has not one mention of the Metaverse, that little thing they spent $80b on and rebranded the whole company around.
As usual, the only centralised power Zuck hates is one that isn't in his own hands.
It won't last. Remember (when was it? Last year?) when Musk went after Microsoft for turning Open AI into proprietary closed code? We never know what's going on behind closed doors, but something changed, Musk backed off, and nothing changed.
Have I fallen into a time warp? Didn't Meta have a head of AI who thought open models were the way to go?
All the money in the world and Zuck can’t buy anything worthwhile in AI. The problem at Meta is really Zuck taking interest in AI. Perhaps the adults should steer him back toward VR unless they all left.
We don't trust you Zuck but we'll take it. Thanks.
This[1]:
> This has the potential to begin a new era of personal empowerment where individuals can use this powerful new capability to reach their full potential, pursue their interests, and improve their lives and the world more than ever before.
Plus this[2]:
> Since 2019, thousands of deceptive advertisements on Facebook have used the likeness of the highly prominent businessman to promote scams, racking up thousands of victims, according to the suit.
> The billionaire's legal team argues that Meta's artificial intelligence tools optimised and personalised fraudulent ads before distributing them, thereby making Meta an active participant, rather than a mere intermediary.
Make for a fun juxtaposition today.
[1] https://www.meta.com/thefutureisforeveryone/
[2] https://www.abc.net.au/news/2026-08-11/andrew-forrest-meta-s...
> they've done enough good
well minus driving teenagers to suicide by helping advertisers target girls with poor body image issues, helping launch psyops campaigns to destroy democracy, helping fascist regimes murder their citizens for wrong think, etc.
Attack anything you can't compete against
Great, we can use open or closed models, now that the money is in running the models through their data centers either way.
Ah the we can't win so whine about the playing field
And they can't win because nobody smart wants to work there for long
Why would someone good at running a social media website be good at managing and developing ai frontier models
Meta were fined by Mexico a few days ago so the timing today of a Meta doing good exercise is probably not coincidental
https://www.latimes.com/business/story/2026-08-07/meta-order...
>> promoting his vision of “personal superintelligence”
What personal vision did Zuckerberg ever get right?