What's up with the bad Meta projects?

post by Yitz (yitz) · 2022-08-18T05:34:08.565Z · LW · GW · 29 comments

Contents

29 comments

What's going on with Meta? The other day Mark Zuckerberg posted a screenshot from Horizon Worlds, a supposedly central Metaverse project which is currently being dunked on by the internet like there's no tomorrow. And for good reason too—it looks horrible. Like, really incredibly horrible. I'm not sure I can communicate just how bad it looks in fact; just see for yourself:

AAAAAHHHHHHHHH MY EYES

 

And this is what Zuckerberg posted publicly—from what I've heard, in the actual application nobody has legs (possibly so simulated sex is impossible?), but chairs abound...the whole thing seems to be a surreally bad mess.

Meanwhile, earlier this month, Meta released BlenderBot 3, a chatbot which somehow manages to be a less realistic user experience than chatting with ELIZA. I tested it out, and it was so much less coherent than even GPT-2 that I was shocked the thing was released in the first place (it classified me as a Christian fundamentalist??). Turns out it also has a slight problem with misinformation as well...

So what's going on? My gut feeling is that this has to be on purpose. I mean, it would be pretty hard to make something as ugly as Horizon Worlds if I tried, and BlenderBot is so inferior to the state-of-the-art right now that it almost feels insulting. I'm having trouble imaging a scenario where a multi-billion dollar company would publicize this without realizing what people's reactions are going to be. However, if I try to imagine scenarios where it would make sense to make it look this bad on purpose, I still find myself confused.  What would be gained from doing that? 

All of this is to say that I'm currently befuddled by Meta's behavior, and am publicly noting my confusion here. If anyone has any insight into how this sort of thing happens, feel free to comment!

29 comments

Comments sorted by top scores.

comment by sdr · 2022-08-19T07:11:00.363Z · LW(p) · GW(p)

Cross-posting some thoughts:

Facebook's metaverse strat is focusing heavily on capability / platform, and not content / single-awe-of-moment. To them it's possibly okay if vrchat wins at the expense of horizon worlds, _just as long as majority of peeps access it via quest_ which they do: https://metrics.vrchat.community/?orgId=1&refresh=30s <- quest users now outnumber pc ones 1:2.

Consider the apple & appstore fiasco, whereby apple can basically, in one OS update, kill retargeting by introducing privacy popups into apps at os level, kneecapping the entire ad industry (single major reason for FB's this quarter _decline of revenues for the first time ever_); and unilaterally decide, that everyone who takes payments for digital services, and has an an app on ios (read: entire B2C SAAS market) now has to pay 30% to them. _and make it a reality_ on pain of removal from app store. _and it works_.

Basically, FB wants to position itself into the same capability / platform play -they control the device, they can dictate terms for everyone building on top of it. 

comment by gwern · 2022-08-19T03:12:17.004Z · LW(p) · GW(p)

Please! Explaining a joke kills it. ;_;

comment by Jonathan_Graehl · 2022-08-19T21:00:58.499Z · LW(p) · GW(p)

In Carmack's recent 5+hr interview on Lex Friedman [1], he points out that finding a particular virtual setting that people love and focusing effort on that is usually how we arrive at games/spaces that have historically driven hardware/platform adoption, and that Zucc is very obviously not doing that. The closest successful virtual space to Zucc's approachis Roblox, a kind of social game construction kit (with pretty high market cap), but in his opinion the outcome is usually you build it and they don't come. I believe Carmack also favors the technical results of optimizing a platform along with a particular game, which is part of his strong motivation for making things better in his immediate environment.

[1] 

comment by Miguel S (miguel-s) · 2022-08-18T07:41:02.590Z · LW(p) · GW(p)

Horizon Worlds doesn't look THAT bad. In fact, it looks as good as it can given the circumstances:

  • It has to run on low-spec devices, running on batteries
  • It's a VR app, so they probably prioritize framerate and response rate over graphics quality

The problem is that the screenshot shared by Mark is a very poor example of how Worlds look (just search it in Google Images and you'll see more fair examples.)

Why did Mark share that? Who knows. Maybe the points above are so obvious that he didn't think of what outsiders would perceive when seeing the image?

Replies from: Richard_Kennaway, alexlyzhov, yitz
comment by Richard_Kennaway · 2022-08-18T22:35:16.249Z · LW(p) · GW(p)

Why did Mark share that?

Perhaps he just has no taste. Facebook itself is hardly a joy to use.

comment by alexlyzhov · 2022-08-24T05:13:33.278Z · LW(p) · GW(p)

From my experience of playing VR games on mobile devices (Quest 1 and Quest 2), the majority of in-game characters look much better than this and it doesn't impact the framerate at all. This seems like a 100% stylistic choice.

comment by Yitz (yitz) · 2022-08-19T03:55:12.887Z · LW(p) · GW(p)

Response here [LW(p) · GW(p)]; btw I did look up other in-game screenshots and they weren't substantially different from the quality of Mark's selfie; imo, the same problems I mentioned apply elsewhere.

comment by ChristianKl · 2022-08-18T07:25:43.230Z · LW(p) · GW(p)

The Facebook newsfeed also got a lot of opposition in the beginning. Whether or not a very innovative product is good or bad can not be decided on whether people like to oppose or make fun of it in the beginning. 

BlenderBot 3 is likely running on substantially fewer resources than GPT-2/3 and thus should not be expected to have the same performance.

Replies from: zac-hatfield-dodds
comment by Zac Hatfield-Dodds (zac-hatfield-dodds) · 2022-08-18T08:29:29.945Z · LW(p) · GW(p)

I'm not sure what you mean by "running on fewer resources" - it's a 175 billion parameter model, so it has the same minimum hardware requirements as GPT-3.

Their inference setup likely uses the same hardware per-response, but I'd guess OpenAI has much faster kernels so Meta would need more hardware to serve a given traffic rate. However, that's totally unrelated to the quality of each response.

Blenderbot is based on their largest "OPT" model, presumably fine-tuned somehow, after a training run which was explicitly imitating GPT-3 after they couldn't work out better hyperparameters, though I'd add it's pretty badly under-trained even according to pre-chinchilla scaling.

(all views my own, you know the drill)

Replies from: ChristianKl
comment by ChristianKl · 2022-08-18T09:55:41.108Z · LW(p) · GW(p)

It seems you are right. I thought previously that BlenderBot was supposed to be useable outside of the research enviroment.

I read through the description and a key difference seems to be that Blenderbot actually has long-term memory about the user with whom it interacts.

Replies from: yitz
comment by Yitz (yitz) · 2022-08-18T14:58:26.760Z · LW(p) · GW(p)

Except that it doesn’t really! The interpretations it stores in “memory” (which is just a small notepad) are so out-of-sync with what went on in the conversation previously that I think it may be doing worse with that memory than without it. As an anecdotal observation: I asked what it thought of Eliezer Yudkowsky’s take on AI ethics, and it responded asking me if I knew that [insert first sentence of wiki page about Yudkowsky here], and stored the “memory” that I enjoyed Harry Potter (presumably because of Yud’s fanfic?). It then proceeded to tell me that God is real and I should believe in Him, which assuming that Yudkowsky is not in fact God, had nothing to do with anything previously discussed. It definitely is not usable outside of a research environment, which Meta should have known, but nonetheless their PR pitch has been promoting it as giving general consumers free access to a cutting-edge conversational chatbot (which it really really isn’t).

Replies from: ChristianKl
comment by ChristianKl · 2022-08-19T09:53:26.620Z · LW(p) · GW(p)

That sounds like it performing worse than because it tries to do the thing with memory. That sounds like an okay tradeoff if you want to have long-term success with having memory.

It definitely is not usable outside of a research environment, which Meta should have known, but nonetheless their PR pitch has been promoting it as giving general consumers free access to a cutting-edge conversational chatbot

I expect that you are wrong about that and that there are people who have fun chatting with it. Even if it's a MVP that's not very powerful there's still a good chance that some people will find it useable. 

To improve the product it's important to have people using it and providing more data, so it makes sense to announce it in a way that gets more people to use it. 

comment by Mitchell_Porter · 2022-08-18T14:55:26.553Z · LW(p) · GW(p)

I am reminded of the opening shot of the 2004 movie "Team America"... Maybe this is the dumb free version of Facebook VR, and a better premium version will just be for paying customers. Maybe premium is even as hard as other comments imply, and meanwhile Mark Zuckerberg is putting this stuff out there to reassure the investors that he does have a product. 

I could go further and speculate that the Team-American deep state has encouraged this crippleware business model (is there a "non-ableist" name for that?), in order to keep technology under control, but that's probably just indulging in the fantasy that NSA et al are secretly far ahead in the technology race. 

Replies from: yitz
comment by Yitz (yitz) · 2022-08-19T04:01:04.357Z · LW(p) · GW(p)

a better premium version will just be for paying customers

So you're saying maybe they're deliberately making a bad product so that when they release the "good version" it will feel more incredible by comparison? The closest (unforced) analogue I'm thinking of would be No Man's Sky incredible comeback after an initially botched release. That would seem like an incredibly risky strategy to deliberately pursue for a multi-billion dollar project though...

comment by niknoble · 2022-08-18T17:54:40.091Z · LW(p) · GW(p)

Unless humanity destroys itself first, something like Horizon Worlds will inevitably become a massive success. A digital world is better than the physical world because it lets us override the laws of physics. In a digital world, we can duplicate items at will, cover massive distances instantaneously, make crime literally impossible, and much, much more. A digital world is to the real world as Microsoft Word is to a sheet of paper. The digital version has too many advantages to count.

Zuckerberg realizes this and is making a high-risk bet that Meta will be able to control the digital universe in the same way that Apple and Google control the landscape of mobile phones. For example, imagine Meta automatically taking 1% of every monetary transaction in the universe. Or dictating to corporate rivals what they are allowed to do in the universe, gaining massive leverage over them. Even if Zuckerberg is unlikely to succeed (and it's still very unclear what direction the digital universe will evolve), he knows the potential payoff is staggering and calculates that it's worth it. That's why he's investing so heavily in VR, and Horizon Worlds in particular.

As for the aesthetics of Horizon Worlds being creepy, boring, or ugly, there are 2 factors to keep in mind.

First, VR hardware and software are in their infancy and you simply can't have very crisp graphics at this stage. That is fine according to the philosophy of modern tech companies. Just ship a minimum viable product, start getting users, and react to user feedback as you go. If Horizon Worlds succeeds, it will look far better in 20 years than it does today.

Second, Horizon may get attacked on the internet for being sterile and lifeless, but internet commenters are not the people who are putting direct pressure on Zuckerberg. Rather, he is surrounded by employees and journalists whose primary complaint is that Horizon Worlds is not sterile enough. I'm sure you've seen the articles: Harmful language is going unpunished, women are being made to feel uncomfortable by sexual gestures. Considering that Zuckerberg receives a constant barrage of these criticisms now, can you imagine the kind of heat he would get if he made Horizon more like VRChat, with its subversive culture and erotic content?

Replies from: yitz, gwern, Vanilla_cabs
comment by Yitz (yitz) · 2022-08-19T00:00:52.253Z · LW(p) · GW(p)

VR hardware and software are in their infancy and you simply can't have very crisp graphics at this stage

As an occasional video game developer, I’m going to strongly disagree with you there. To give a counter-example:

Walkabout Mini Golf is a VR game that runs on Oculus Quest, Rift, and Steam VR, made by this fairly small studio (and most of the people listed there didn't even work on the game) [EDIT: I reached out to the studio on Twitter and it turns out the game was mainly developed by a single guy, Lucas Martell].  It looks like this:

I've played this game with a friend of mine (who shows up as a stylized floating head that looks pretty great), and it was crisp, clear, high frame-rate VR perfection. Even in multiplayer, everything works smoothly, and it serves as a really nice virtual social space.

Having limited graphics capabilities does not place a significant limiting bound on aesthetics. As another example,  Anodyne 2: Return to Dust is a jaw-droppingly beautiful game (developed by only two people!) deliberately made with PS1-era  graphics:

Simplicity does not necessitate ugliness.

Rather, he is surrounded by employees and journalists whose primary complaint is that Horizon Worlds is not sterile enough.

This may very well be true (despite my personal distaste for that line of thought), but even if it is, being inoffensive and "bland" doesn't mean you have to look bad! Nintendo's oeuvre, for instance, shows that being friendly for all ages doesn't require sacrificing aesthetic beauty. Meanwhile, in screenshots online and in the "selfie" Zuckerberg posted, model sizes are wildly inconsistent (look at the trees—or is that supposed to be grass?—on the ground), the clothing of avatars are almost surrealistically bad (why is Mark's top button so far off to the left?), the shading is worse than what I could make in half a day with Unity when I was 12, and overall everything manages to look more slapped together than this notorious disaster of an asset flip.

I can't help but feel that on some level this must be intentional, or at least the result of some absolutely horrific mismanagement. 

comment by gwern · 2022-08-18T22:54:46.131Z · LW(p) · GW(p)

Worth noting that even VRChat doesn't want to be VRChat - as they just banned all the mods necessary to access said subversive culture & erotic content, much to the furry of their communities, one might say.

Replies from: MSRayne
comment by MSRayne · 2022-08-21T16:21:16.883Z · LW(p) · GW(p)

That seems like a bad move. Banning people from being human is always counterproductive. Any successful media of the future will be filled with weird porn; nobody wants to be censored.

Replies from: yitz
comment by Yitz (yitz) · 2022-08-21T22:10:04.069Z · LW(p) · GW(p)

I'd certainly like to believe that; the problem comes if a single company (which cares less about consumers than it does shareholders) is able to monopolize the market, as that then takes away consumer's choice to be kinky if they so wish. Alternatively, a moral panic could plausibly cause regulations to be set in place for the metaverse which would make doing anything non-wholesome extremely difficult for the average person. But yeah, from the consumer perspective this is almost certainly a bad move.

comment by Vanilla_cabs · 2022-08-24T10:12:32.038Z · LW(p) · GW(p)

Unless humanity destroys itself first, something like Horizon Worlds will inevitably become a massive success. A digital world is better than the physical world because it lets us override the laws of physics. In a digital world, we can duplicate items at will, cover massive distances instantaneously, make crime literally impossible, and much, much more. A digital world is to the real world as Microsoft Word is to a sheet of paper. The digital version has too many advantages to count.

Either there will be limitations or not. No limitations means that you can never be sure that someone in front of you is paying attention to you; your appearance indicates nothing but your whim of the moment; you can not be useful to others by providing something that they can't get by themselves (art? AIs can make art). My first impression is that it will be very hard to build trust and intimacy in this environment. I expect loneliness and depression to rise as this technology is adopted.

But there will probably be limitations. Except that while in our world the limitations are arbitrary, in the Metaverse they will be decided by a private company and will probably enforce a plutocratic class system.

comment by Ben (ben-lang) · 2022-08-18T17:36:58.496Z · LW(p) · GW(p)

My theory (at least for the image) is that someone at Meta has decided that "big has to be bland". They want their software to be super mainstream, the "default, vanilla" VR world that every other VR place will be compared to. (Maybe one is more Super Mario, another more gritty.).

If you tell the design team to aim for blandness maybe it destroys their artistic tendencies so completely they are unable to make anything that looks remotely good.

Replies from: Jonathan_Graehl
comment by Jonathan_Graehl · 2022-08-19T21:02:29.631Z · LW(p) · GW(p)

IMO this project needs an aesthetic leader. A bunch of technically competent people building tools they think might be useful is very likely to result in a bunch of unappealing stuff no one wants.

comment by pseud · 2022-08-18T10:29:56.787Z · LW(p) · GW(p)

I don't think the screenshot looks that bad. Netizens love to be irrationally extremely negative about Zuck, and it's possible you have been swept up in this.

Replies from: adrian-arellano-davin, conor-sullivan
comment by mukashi (adrian-arellano-davin) · 2022-08-18T11:30:49.513Z · LW(p) · GW(p)

Sir, we seem to have a very different taste I am afraid

comment by Lone Pine (conor-sullivan) · 2022-08-18T12:19:40.388Z · LW(p) · GW(p)

From a 3D graphics perspective, it's very similar to the Wii. The Wii was an underpowered console and so its graphics were simple. I assume Horizon's graphics are simple for a similar reason, because of the limitations of VR (need high framerate, low latency, consumer hardware needs to be reasonably priced, etc). However, the Wii had a lot more charm, IMO.

I think people are also negative about digital minimalism because how pervasive it is the advertising of big tech companies. See "Corporate Memphis". Corporations prefer minimalism since it is cross-culturally inoffensive -- except among those who have learned to associate it with companies they don't like.

Replies from: pseud
comment by pseud · 2022-08-19T01:55:04.556Z · LW(p) · GW(p)

Yes, I think the graphics are quite simple. I think your explanation relating to  current limitations of VR is enough to address a lot of the OP's confusion/questioning. It's not that Meta is purposefully trying to look bad; they're just sharing the state of the art honestly. It's also worth noting that this seen by Zuckerberg as a temporary state, and his goal very much seems to be photorealism. If you listen to his recent interview with Lex Fridman you'll hear him bring up photorealism again and again. 

 

Why is Meta sharing their work now instead of waiting for something closer to photorealism? I am speculating, but I think it is because Meta sees this is the best way to get a head-start on the competition. When photorealism is achieved, Meta will already have years of experience of making useful and enjoyable applications for VR. 

Replies from: yitz
comment by Yitz (yitz) · 2022-08-19T03:41:13.888Z · LW(p) · GW(p)

See my response here [LW(p) · GW(p)]; this is very much not state-of-the-art, and hasn't been for a fairly long time.  As many on Twitter have pointed out, Second Life, which was released in 2003, is so ahead of Meta in terms of both looks and feel that the concept of them getting a "head start" against competition through this seems implausible. Gwern also mentioned VRchat in the comments, which while it is facing difficulty with moderation/censorship (whichever you'd prefer to call it lol), it is also an obviously superior product. If I were an investor, I'd be more worried after seeing what Meta is publicly releasing than what I'd be under the counterfactual where Meta hadn't publicized these products yet.

Replies from: pseud
comment by pseud · 2022-08-19T04:15:35.561Z · LW(p) · GW(p)

I read your post but I thought it was more about aesthetics than technology. 

Horizon Worlds is a program where users can make their own environments and can make aesthetic decisions for themselves.

comment by mukashi (adrian-arellano-davin) · 2022-08-18T11:29:03.765Z · LW(p) · GW(p)

LW is not my usual place to have some laughs, but this made my day.