[HN Gopher] Sharing new research, models, and datasets from Meta...
       ___________________________________________________________________
        
       Sharing new research, models, and datasets from Meta FAIR
        
       Author : TheAceOfHearts
       Score  : 173 points
       Date   : 2024-06-18 17:01 UTC (5 hours ago)
        
 (HTM) web link (ai.meta.com)
 (TXT) w3m dump (ai.meta.com)
        
       | pbronez wrote:
       | Interest that they're releasing a deepfake detector. I expect
       | that to get integrated into image generation training pipelines.
        
         | ronsor wrote:
         | Throw "deepfake loss" onto the pile of loss functions.
        
       | sghiassy wrote:
       | Is it just me or does Meta seem to be doing better than most
       | companies right now open sourcing their AI research?
        
         | IncreasePosts wrote:
         | Yes, I'd also like to know why Meta has chosen this path,
         | whereas many of the other big players haven't. Usually they all
         | settle upon the same viewpoint.
        
           | hlfshell wrote:
           | It started by accident, with the original llama weights being
           | leaked by two separate employees. They've since embraced
           | opening the weights, which I'm all for.
           | 
           | As for why? I have a theory: Meta is not in a position to
           | capitalize upon the model itself. Yes, they can use it
           | internally, and maybe their competitors can copy it to - but
           | there are no real competitors to Facebook or Instagram that
           | can benefit from it enough to make it a differentiating
           | facet.
           | 
           | Thus, releasing stuff for open source does two things:
           | 
           | 1) Make them more attractive to research talent (Apple
           | famously recently started to publish research because their
           | traditional secrecy was causing issues with hiring top
           | talent) and...
           | 
           | 2) Continues to undermine the ability to make $$$ off of
           | model alone, driving it towards being a commodity rather than
           | the long term profit engine for other companies.
        
             | sdenton4 wrote:
             | In short: "Commoditize the complement."
        
             | onurcel wrote:
             | > It started by accident, with the original llama weights
             | being leaked by two separate employees.
             | 
             | This is not true. Meta Fair has been built on openness from
             | day 1. We published many papers and open-source d many
             | repositories to reproduce the work
        
             | ipsum2 wrote:
             | Wrong. FAIR has been open sourcing ML models and source
             | code for the last 10+ years. It did not start with llama.
             | Also, llama was not leaked by employees, but people in the
             | broader community, who the weights were shared with.
             | 
             | For example:
             | 
             | Faster R-CNN - state of the art image segmentation,
             | released in 2017.
             | 
             | FastText - text embedding models, 2016.
             | 
             | FAISS - vector DB, 2018.
             | 
             | https://github.com/orgs/facebookresearch/repositories has
             | over 1,000 repos.
        
               | hlfshell wrote:
               | I was speaking about LLM weights specifically and llama,
               | not all models and work at FAIR.
               | 
               | Per your links, it's clear that FAIR does have a good
               | history of open source work.
        
             | freehorse wrote:
             | > weights being leaked
             | 
             | You can hardly call that "leak" when they basically were
             | sending the weights to thousands of people who applied for
             | access. It is not that they kept them secret.
        
           | righthand wrote:
           | Same reason they open sourced reactjs or encourage jestjs
           | usage. If they give it away they can benefit by being
           | embedded in the stack. Then all the scale issues they can
           | beat implementers on because they have the money to run it.
           | To keep ahead of people implementing their tech they have
           | their own data trove to train on. You only get a small piece
           | of it. It's all to position themselves as a sensible
           | solution.
        
           | downWidOutaFite wrote:
           | My guess is that when Llama leaked on 4chan and it blew up in
           | the community it somewhat forced their hand to go with an
           | open strategy. But they also have a history with pytorch of
           | reaping the benefits of an open strategy. The benefits are
           | well explained in Google's leaked "We Have No Moat" document.
        
           | bg24 wrote:
           | 1/ Build a community (think Linux for OS, Android for mobile,
           | React for frontend) and figure out monetization later. What
           | is clear to most people is that something fundamental is
           | changing in how we build and consume applications.
           | 
           | 2/ Prevent OpenAI from cornering the future $$$ market.
           | Unfortunately, Google search is hit as well, but it is more
           | due to the generational shift.
           | 
           | 3/ Attract the best AI researchers. A product is a good as
           | its core set of people (often just a few).
        
           | timy2shoes wrote:
           | They've chosen the path of commoditizing their complement. To
           | ensure that ML capabilities are not a differentiating factor
           | in the market, make ML capabilities a commodity available to
           | everyone at the marginal cost.
        
           | nicce wrote:
           | There isn't conflict in business model. And if people can
           | help to improve their models, they can extract better value
           | by themselves.
           | 
           | Also, maybe they need to improve their brand. Hoarding data
           | for over a decade, maybe bringing something back now.
        
           | gillesjacobs wrote:
           | Facebook/Meta has been doing open research in ML and NLP for
           | far longer than the current LLM era: Convolutional Neural
           | Networks for Sentence Classification (2014), FastText (2016),
           | PyTorch (2016), fairseq (2017), LASER (2018), RoBERTa (2019),
           | XLM-R (2019), and BART (2019),
           | 
           | They were always present at ACL with decent open research as
           | far as I have been studying/working in NLP (2014).
           | 
           | It's part of a strategy to attract top talent in the field.
           | If you want top researchers you have to let them publish,
           | which in turn hones a reputation of solid research,
           | attracting more talent.
        
             | behnamoh wrote:
             | > It's part of a strategy to attract top talent in the
             | field. If you want top researchers you have to let them
             | publish, which in turn hones a reputation of solid
             | research, attracting more talent.
             | 
             | This. Although, it turns out, if you pay them well enough
             | (like OpenAI), they'll forego publishing. If you can make
             | enough to retire by 40, why work at places that pay less?
             | 
             | ofc, not all researchers think that way, and there are
             | those who are in it for the science, not just money.
        
           | lllaaaammmaaaaa wrote:
           | All their users are writing "content" for free and
           | communicate with each other. AI generated "content" does not
           | really fit into this.
           | 
           | They do not want their users to go to ClosedAI or similar and
           | communicate with an Artificial Stupidity instead talking to
           | each other on Facebook.
           | 
           | So it is in their interest to undermine the market for
           | Artificial Stupidities by releasing the models for free.
        
         | dvngnt_ wrote:
         | they've been on this wave for years.
         | 
         | thanks to meta i have been creating instrumentals of some of my
         | favorite songs with vocals
         | https://github.com/facebookresearch/demucs
        
         | hiddencost wrote:
         | They are unable to compete with the top players on quality, so
         | they instead compete by winning the largest user base with open
         | models.
        
           | 12345hn6789 wrote:
           | Note that due to the licenses no company will touch any of
           | these with a 10 ft pole. Great for individuals looking to
           | experiment though
        
         | occamrazor wrote:
         | How so? Maybe in the past, but nothing announced today is open.
        
       | Filligree wrote:
       | Pity they aren't including image generation. Multimodal
       | generation with reference inputs is my #1 ask for things like
       | novel illustrations.
        
         | advael wrote:
         | It seems like their image-tokenization model might be useful
         | for this, but also have you looked into stuff like ControlNet?
        
       | mi_lk wrote:
       | Is Llama3 400B still going to be released?
        
         | ai_what wrote:
         | Yes, Yann LeCun posted about it on twitter ~2 weeks ago.
         | 
         | Edit, it was a month ago:
         | https://twitter.com/ylecun/status/1793181068943639014
        
         | ein0p wrote:
         | Under a "do not even look at it" license I'm sure. /s
        
       | RobotToaster wrote:
       | > non-commercial/research-only license.
        
       | zxcb1 wrote:
       | Last time they did this, I was not granted access
        
         | ai_what wrote:
         | Try a different email address, that's what fixed it for me.
        
           | zxcb1 wrote:
           | Sure, that may work; but what does it mean?
        
             | ai_what wrote:
             | I can't speculate that for you. Try reaching out to them if
             | you must know.
        
               | zxcb1 wrote:
               | When we say open X, we expect something like the Via
               | Negativa principle; instead of inviting members into a
               | closed community, exclude those that violate the code of
               | the open community.
        
               | ffsm8 wrote:
               | This is false. Any "open" event (or anything, really) can
               | be closed for select individuals.
               | 
               | The reason for this exclusion just can't be a banned
               | criteria such as race.
               | 
               | (I.e. kicking someone out after they've stirred up
               | controversy)
               | 
               | I'm not saying that this happened to you, I'm just
               | addressing the point you made, that anything "open" can't
               | or shouldn't be exclusionary.
        
       | zelphirkalt wrote:
       | Considering, that this is all based on data and personal
       | information of people, the people in general should have the
       | right to dictate Facebook what they can and cannot do with that
       | data and what happens with the products made using the data. It
       | is laughable, that they act here as if they are doing something
       | friendly or fair. Same is true for other companies like Microsoft
       | and "Open"AI.
        
       | tech_ken wrote:
       | What's the argument that the early-fusion setup scales/maintains
       | easier than late-fusion? Does this come at a tradeoff or is it
       | just uniformly better?
        
       ___________________________________________________________________
       (page generated 2024-06-18 23:00 UTC)