Hacker News Reader

Comment by isoprophlex | Hacker News Reader

Is this a tactical leak, stemming from a "commoditize your complement" strategy? Open source as a strategic weapon, without having to explain board members/shareholders/whatever that you threw around money on training an open sourced model?

View on HN

I would assume so. Meta’s ML/AI team is very strong, but they probably don’t have a comparable product offering to ChatGPT ready for public use. So instead, they bought themselves some time by letting the open source community run wild with a lesser model and eat into OpenAI’s moat.

They didn't leak it. Someone else did.

I dont think theyre saying Meta AI leaked it, but they anticipated someone else will and still went ahead with it as they wanted the consequences.

It's extremely common for a "leak" to actually be fully intentional, but the organization in question just wants plausible deniability to mitigate legal/political/popular blowback.

In order to preserve plausible deniability, the leak will look genuine in all aspects that are easy to simulate. "Someone else did it" is easy to simulate. A better gauge would be to see if anyone is caught and punished. If so, it was probably a real leak.

I think the key here is that, given the way that Meta distributed the model, a leak was inevitable. So while they may not have directly orchestrated a leak it must have been an intended result.

They have tacitly endorsed the leak. https://github.com/facebookresearch/llama/pull/73#issuecomme...

Only because publicly visible actions are worse for them

People have gotten DMCA takedown requests from them over Llama repositories

If they were interested in limiting distribution, saying essentially "go ahead and seed this torrent more" is worse for them than doing nothing.

I’ve actually beat the streisand effect before by not responding.

The crowd gets bored and my DMCA requests flurry out a month later and all evidence disappears, individuals that might notice dont have the crowd to amplify that they noticed.

You can call that “tacit consent” if you want. But streisand removes all leverage.

Are they going after copies of LLaMA, or after LLaMA derivatives specifically?

https://news.ycombinator.com/item?id=35287733

That's a contributor to the repo, not someone with commit access.

A contributor who is also a Facebook employee and co-author of the LLaMA paper, presumably speaking in official capacity.

Why would you presume that by default? Need a quote to conclude that

It's widely presumed within faang-type-of companies that anything an employee says or does can be interpreted as an official company statement, especially by the press. As a result, many of these companies offer, often mandatory, trainings that underscore the importance of speaking carefully in public, since one's words can end up on the front page of a popular newspaper.

Although I don't know how FB rolls internally, it seems more likely than not to me, that it was ack'd by someone higher up in the organization than line engineers or managers. Someone with a permission of speaking publicly for a given area of a company - doesn't need to be CEO, more like a VP/Director maybe.

Here's a couple more quotes from Yann LeCun, their Chief AI Scientist:

https://twitter.com/ylecun/status/1643945742850031616

https://twitter.com/ylecun/status/1644503237699969026

pretty clear the non-release and extremely half assed response to the leak is just CYA

You would not believe the amount of internal and public facing sabotage done by FB employees.

What would you think is holding back Meta and Google? Why can't they just scale up the compute they throw at the problem?

What are they tinkering on?

I think Meta’s problems are largely similar to Google’s: they have very bad leadership, specifically Zuckerberg, and thus can’t leverage their existing talent to execute/compete. The whole Metaverse fiasco would seem to demonstrate he’s effectively a “mad king” at this point, and probably surrounded by a sycophantic c-suite. Having the best talent in the world (which they obviously do by how fast LLama was spit out) isn’t going to matter that much if its all serving at the behest of someone who has become deluded by their initial success and has no ability to course correct.

Selling 20 million Quest 2 headsets is a pretty good outcome for the so-called Metaverse fiasco.

I think they spent 36 billion on metaverse stuff at last count, so no, not really.

How did literally nothing but Quest come out of that then?

They would've sold exactly as many of those if they didn't jump full-speed into the Metaverse rabbit hole. Even the Quest 2 release predates that direction shift.

Spending $100b on net losses is too much for such an outcome.

Those headsets are subsidized, do you really believe that level of hardware costs less than the unit price of 300 dollars? Meta is trying to gain market share via selling at a loss and perhaps making money through games or simply getting everyone inside the "metaverse" to then show them ads.

It’s not open source. Llama is proprietary, the license hasn’t changed. Just like the source code to windows leaking doesn’t make windows open source.

> Llama is proprietary, the license hasn’t changed.

.. assuming that the weights are copyrightable and that you agreed to license them from Meta (fill out the form). Weights lack at least two requirements to be eligible for copyright protection in the US and many other jurisdictions. For the US, the weights are likely to be considered public domain (unless new legislation is introduced) but we'll have to wait for the courts to know for sure.

(Cont.) If you are of the opinion that weights are copyrightable, I encourage you to show how weights satisfy the requirements of copyright - particularly those concerning originality and human authorship - rather than silently downvoting comments you don't like. At least that way would - hopefully - result in a discussion that is far more informative for all of us.

FYI I think you made a good point and I am not silently downvoting you.

I feel the same regarding people moderating comments to oblivion because they disagree on HN and have commented about this recently to dang.

I should have been more careful with how I wrote my comment. I almost always get silently downvoted when I bring this point up for discussion so I was directing that "you" at those people in particular and not at you. Sorry for coming across as attacking you and thank you for caring about hn.

It’s OK. Have a happy Easter. :)

Reply on news.ycombinator.com