Also don't underestimate the value they're getting from making more overtures to the developer community. It could be a coincidence, but it's only since they started releasing these models that I started noticing people on HN calling them "Meta", and attitudes towards them have been far more positive of late than usual.
Good will isn't worth as much as cheap moderation automation and fancy features, but it's worth something.
> Also don't underestimate the value they're getting from making more overtures to the developer community.
I wonder if it's significant. As developers, we're biased to think it matters, but in the grand scheme of things, 99.99% of people don't have a clue about open source or things that matter to hackers. As far as recruitment go, developers look primarily at how much they make, possibly the tech and how it looks on resume. There's always been a stigma around social networks and generally big tech companies, but not to the point it's going to hurt them.
I agree there's a lot of decisions that have to be made on the basis of "the tech and how it looks on resume" (as I'm finding out by not having React Native on mine).
> There's always been a stigma around social networks and generally big tech companies, but not to the point it's going to hurt them.
I'm not sure it was "always":
The one Facebook developer event I've been to made me feel dirty just to associate with them, but before that I had no negative feelings. It started off as "the new LiveJournal".
Deleted my account for a few years, only came back to it when I started planning to move country and wanted to keep in contact with those who stayed put.
AI researchers get huge checks working anywhere. So how does your company distinguish its appeal? Play to a researcher's desire for recognition in their field. Lean on open source. (I'm just paraphrasing Zuckerberg.)
Similar how places like SpaceX or game development are known for low pay and bad working conditions, yet have no shortage of applicants.
No matter whether you want to hire the best of the best or just average people at a lower than average price, being a place where people want to work helps immensely
It's funny how quickly Zuck managed to turn his image around from "data-stealing actual lizard person" to "kind of normal guy" with a few years and a haircut. It's also not lost on me that he's the only "hacker" major tech CEO remaining:
- Sundar is a glorified bean counter and his company is rotting from the inside, only kept afloat by the money printer that is ads.
- Satya and Microsoft are in a similar boat, with the only major achievement being essentially buying OpenAI while every other product gets worse
- Tim Cook is doing good things with Apple, but he still runs the company more like a fashion company than a tech company
- Amazon was always more about logistics than cool hack value, and that hasn't changed since Bezos left
- Elon is Elon
Meanwhile Zuck is spending shareholder money pushing forward consumer VR because he thinks it's cool, demoing true AR glasses, releasing open-source models, and building giant Roman-style statues of his wife.
I feel like owning Oculus must be the equivalent of the curse of the monkey's paw. It drove Palmer Luckey veritably[0] insane[1] when he was the head of it, and between the statue nonsense, the weaponizing of Llama and his angling for a position as a Trump advisor, it sure looks like Zuckerburg wants to head down the same path.
Facebook Engineering has always been well regarded — starting with React on the front end, but also projects like Open Compute.
Their product management on the other hand— well, I mean, Facebook and Instagram are arguably as popular as McDonald’s. So they’ve got that going for them.
It's also got an infinitely better reputation since it's arguably not social media, but just a solid instant messenger.
I was personally very concerned when they acquired it, but went from a begrudging user (it was the de facto standard in the country I was living in at the time) to an excited one (Facebook/Meta added end-to-end encryption, multi-device capability, device migration etc.)
* More UI space devoted to brands (the Updates tab
* Communities UX is, uh, suboptimal
* Brands / business accounts can message you with no opt out, you can block a chat after the fact but won’t stop them from messaging you from another account.
These made the news in WhatsApp’s favourite testbed market[1], but if you think it’s not coming to other markets…
Anyhow, tl;dr — WhatsApp’s no longer in the “put users first” phase, it’s in the “monetize” phase of Facebook Product Management.
As a person who paid for WhatsApp pre-Facebook, that does make me sad.
Spot-on. Hacker News has a hate-boner for Facebook and the Facebook-branded products, but from an engineering standpoint they're up there with Netflix solving problems at scale. FAIR had a better reputation than OpenAI since the start, and they've developed the Oculus acquisition to it's logical extremes. You can do a lot worse, looking at the way their peers handle similar situations.
All the big tech companies have a Facebook-esque product they wish they could get rid of forever. Meta has Facebook, and instead of imploding like everyone said they would (for decades) they demonstrated competency in engineering and culture. The next 4 years will be a gauntlet with a literal "Mr. X" advising social media policy, but I frankly don't think Facebook has ever been down for the count in a pragmatic sense.
I think the only concept I’ve heard as to the decline of Facebook is, Facebook is for “older people” (read: millennials+) but I have no idea how true this is.
I used to live in the city and now I live out in nowhere land. Everything around here is done thru facebook groups and facebook events. If you want to keep in touch with the community, it is really only on facebook.
Some of my older facebook friends seem to have left the platform and I seen more than once of my old city friends posting final goodbyes declaring they are leaving the "dead" platform for discord or some stuff where their family will never find them.
It's funny. The only time I've ever seen Hacker News unanimously applaud a Facebook product was when Pytorch announced they merged Apple Silicon support. Seems like Mr. Zuckerburg knows how to play a winning hand.
I'm also glad that those who chose to give my 19 year-old self a break ignore all the other scandalous shit my company did in line with those comments in the decades after the fact.
Profiling people for advertising, building shadow profiles, buying 3rd party user data, tracking people across the web, devices and geographically, psychological experiments, optimizing for engagement, limiting post reach without pay... generally prioritizing profit over people.
Ok, I'm going to assume that you're somewhat technical given where we are.
First off, if I estimate a series of numbers on you and those result in you being served a set of ads, is that wrong? If so, can you help me understand what's wrong with that?
Shadow profiles are mostly bullshit, yes data was collected for non users due to how the SDK and pixel worked. This data was all assigned to one user ID and was filtered out by basically everyone using that data.
I'm a little confused as to why buying third party data is wrong, the problem with this is that it's legal to collect and sell the data.
Speaking as a psychologist can you clarify what's wrong with psychological experiments?
I think your point about optimising for engagement was definitely a mistake, given the downstream consequences. However, they needed to find some way of ranking feed after Zynga almost killed them (a chronological feed would have been all Farmville all the time for a number of years) and they picked likes.
They also optimized for time spent but people complained about that so they started optimising for comments and shares which made everything worse, sadly.
Limiting post reach for pages was a legitimate business decision, particularly given the ranking constraints.
Like if you look into what that company actually did all the data stuff was a smokescreen for their speciality of getting your opponents caught in compromising positions.
Fundamentally, neither the Big 5 traits nor friend data is particularly useful for ad targeting (internally neither approach was successful).
Can you please be specific about the manipulation of elections?
I presume we're talking about Myanmar and the genocide. Personally I generally place responsibility for bad actions on the people engaging in genocide rather than the communication mechanisms involved. Should we have banned radio after the Rwandan genocide?
Hitler used radio very effectively, should we have banned that?
It would be strange if they didn't also use these models to generate much more sophisticated models of their user's interests and hyper-targeted advertising that always looks and feels like trusted friend recommendations for the exact product that's been missing from your life.
> It would be strange if they didn't also use these models to generate much more sophisticated models of their user's interests and generate the hyper targeted advertising
You have the sequence reversed as Meta already created ad targeting models. Meta was forced to scale its AI competence for ad targeting when Apple sent out a privacy update that destroyed tracking-based ad-serving and tanked Meta's share price by deleting billions in revenue for Meta over many quarters. Now that Meta has this skill as a core-competence, they are creating new models for public release. Why they are doing so is debatable[2], but I imagine the cost is marginal since they already had the GPU clusters, talent and know-how for survival purposes.
2. I suspect Zuckerberg is not enthused by the idea of a future AI Apple-analog unilaterally shutting him out of the market. Having your net worth cut in half by a press-release has got to hurt.
Thanks for the article but I don't think I have the sequence reversed.
Without access to the tracking signal, it's been more important to build out a system that can recreate the value from that lost signal by analyzing what users are actually sharing and saying on their platform. Hence the importance of chat (VR, text, video...) and AI that can be used to process and extract value from a chat signal.
I believe Meta's primary revenue source is still advertising (98%), so that is probably 98% of the why.
> Thanks for the article but I don't think I have the sequence reversed
I suppose I read your first sentence as being in future tense when it might not be. The thrust of my argument is that Meta already successfully built those ad targeting models (Advantage+), and they preceded the Llama releases, so they don't need to use Llama-derived models for ad targeting, as I understood your comment to be suggesting. The sequence was not/will not be "Llama -> ad targeting", but was "ad targeting -> Llama"
Meta didn't have to release the weights of the models. Ad revenue doesn't explain why they did so.
Lots of applications for the technology are likely and already being used.
Llama to improve Customer to Product Matching through content analysis.
Llama for Ad Generation based on customer and product profiles.
Llama to generate additional chat signal for humans not chatting with other humans.
Llama as a site content moderator.
Llama as a code generation assistant.
Llama to clean up and improve internal wikis.
Llama as an operational analysis tool to detect inefficiencies in internal communication...
They will use it every in every situation where they find a successful use case. Open sourcing the weights buys goodwill but it won't be long before everyone open sources their weights since the only real long term moats are the successful products built around the technology and not the technology itself. Facebook is its own moat. Altman suggested this outcome a few days ago in his NYT DealBook interview when he compared the transformer model to the transistor.
Good will isn't worth as much as cheap moderation automation and fancy features, but it's worth something.