Yield depends on humidity, which varies according to region and season.
It also requires more infrastructure to get yield. In theory all you'd need to have is these etched metal plates, a transparent dome and a source of briny water. (and a cleaning mechanism)
The etched plates creates 100% humidity (probably more as it'll condense out)
The crucial part is that pressure from the capillary action pushes the concentrated brine out onto the non capillary area. unlike fabric the area isn't enclosed so cleaning is easier if the salt starts to accumulate.
Obviously it needs to be cleaned regularly otherwise the salt encroaches into the sensitive bits. However the cleaning method doesn't require dissolving, just scraping.
For me personally I am vaguely indifferent to programmers using LLMs to make more shitty code. My worry is the second and third order effects
For work currently, as an SRE, I'm being asked to maintain and look after slop as if its properly built and instrumented. Our platform has clear rules and conventions, and AI isn't following those.
For the wider world, I fucking hate that image/video generation is evaporating what is "real". For memes sure its great, but for bad actors it gives a brilliant way to say "its AI wasn't me" and then the debate moves away from "did person do bad thing" to "is it wrong to say that things are AI?"
I also worry about the debasement of value of human work. Looking at history, say of the weavers, it didn't work out to well for them when the powered loom came along.
> I also worry about the debasement of value of human work. Looking at history, say of the weavers, it didn't work out to well for them when the powered loom came along.
They eventually moved on to other things, because that was the only option. And the world is better with the power loom. It's scary but we still have to embrace that eventually pretty much all valuable labour will be automated, and by then our society and economy needs to have been restructured for supporting humans providing 0 economic value.
The biggest technical one was when we were making an all day wearable AI assistant thing. It basically had really precise office location (think cm level accurate) a shitty VLM to describe what the wide angle lens was looking at, Speech to text, OCR and a gaze recorder that decribed what you were looking at.
This was all streamed to sqlite. The thing that was really "oh shit" what the thing that made the whole system usable: a 4 paragraph prompt that turned natural language into SQL and reported back to the (non technical user) what they wanted to know.
The most recent one is being caught out by Genai video of a gymnast. I worked in VFX so I am normally able to spot dodgy shit, but this one was close to being real, scarily real.
> It is sad that it takes a Meta developer having some fun to realize they should open up ADB.
Former facebook research lab twat here. It wasn't one dev.
We asked when they shitcanned portal (which was a great product, badly managed) to open it up. Infact one of the kernel devs made a very direct plea to allow the community to adopt the hardware so that we could avoid Ewaste.
It was denied because there are keys on the device that would leak if meta opened it up. (I'm not an android dev so I don't know the ins and outs of that)
However, portal was a casualty of the dash to VR. They scaled up the team briefly, which meant that lots of weird stuff was tried, but the roadmap was diluted. The idea was that they portal would be the "portal" to horizon worlds. this meant that they pushed back the plan for thirdparty app stores that would have meant you had something to actually do on the device.
neglect and stupidity from zuck meant that the portal was killed, even though the next gen device was actually a really great media device (wireless, removable charging stand, excellent speakers, but nothing to run on it.)
> They scaled up the team briefly, which meant that lots of weird stuff was tried, but the roadmap was diluted.
Boz never cared for Portal, it wasn't his product. I was one of the original engineers on Portal. The VP running the research lap responsible for Portal was canned in a political coup, and her entire org moved under Boz, merging it with Oculus into the AR/VR team. There was some ham-fisted justification around why a smart home product should be part of AR/VR, but it never really made sense.
Portal had a bunch of other problems, including:
* Massively over-specced hardware, the SoC was the same SoC as the Quest, even though it had no reason to be. The BOM was something like $500. We were selling these units at a huge loss.
* Cambridge Analytica broke right in the middle of development, which completely tanked any remaining trust in the Facebook brand. Everyone knew the product was completely sunk at that point, but nobody wanted to come out and say it. At the last minute we had to stuff a plastic camera cover into the box as a result.
* Boz was convinced we could build a voice assistant for Portal and Quest that was better than Siri, but the Assistant team at FB was completely out of their depth. We ended up right before launch having to sign a deal with Amazon to ship Alexa on the product.
* So much politics. AR/VR had a virtually unlimited budget so there was a massive land grab to hire as many people as possible, with no consideration around what they'd actually work on. Even though Quest and Portal had the same SoCs, they had completely separate Android OS builds and engineering teams, because everyone was trying to build the biggest engineering teams they could. People were constantly leaking shit: I found out we were delaying the project because an executive leaked it to Bloomberg while the executive meeting was still happening.
> So much politics. AR/VR had a virtually unlimited budget so there was a massive land grab to hire as many people as possible, with no consideration around what they'd actually work on. Even though Quest and Portal had the same SoCs, they had completely separate Android OS builds and engineering teams, because everyone was trying to build the biggest engineering teams they could. People were constantly leaking shit: I found out we were delaying the project because an executive leaked it to Bloomberg while the executive meeting was still happening.
Hnnnnnn
yup, the empire building and land grabs. yup, I had forgotten about the early days before maui was actually universal and people needed different tools to flash different devices.
ar/vr, horizon, and boz ruined a bunch of great software products that people were actually using and enjoyed (i've seen the NPS) to shoehorn horizon worlds and ugly 3d avatars into that no one wanted.
We bought two portals for elderly relatives, predominantly for video calling, and I don't think there has been another product, then or since, that fitted that use case as well, especially with people who maybe aren't as familiar with smartphones.
So somewhat frustrating when it all started to wind down various bits of functionality disappeared a bit at a time, until finally you had something that would receive calls, but not be able to make them - and perhaps not even that any more.
(About the only downside I saw on it was the messenger vs whatsapp tussle caused a bit too much confusion).
But it was a solid bit of household tech for several years, so +1 for that!
I'd not seen that one, and it does look like it has some nice points - although needing its own special app is rather frustrating (vs leveraging apps that people are already using).
Seems not too disimilar from the echo show's video calling capabilities in that respect (i.e. only calls within the Amazon ecosystem).
i had a portal at home for work. great product for VC, i tried using one with my parents and my dad kept it in the trunk of his car outside because that's how negative the facebook brand equity was.
> It was denied because there are keys on the device that would leak if meta opened it up.
Many devices wipe such keys as part of unlocking the bootloader. The better ones restore access upon relocking with a stock OS but that's far from guaranteed.
> It was denied because there are keys on the device that would leak if meta opened it up. (I'm not an android dev so I don't know the ins and outs of that)
Any idea what changed?
> neglect and stupidity from zuck meant that the portal was killed
Is Facebook really set up such that one person's whim is the single point of failure? Is there really no way for teams to progress projects with value somewhat independently?
sadly, or fortunately I am not at facebook anymore, so I don't have the inside track on what changed.
> Is Facebook really set up such that one person's whim is the single point of failure?
Kinda. Zuck sets direction, and he has key interests. The thing that really makes him happy is cutting edge research and new features. The thing that passes him by completely is product experience. Oculus is a great example of that. The user experience was/is trash. the time to fun is/was too high and was for a long time. Carmak spent ages saying "we can't compete on hardware specs, we can compete on ecosystem and experience" he lost that argument.
Outside of zuck there are only a few areas that actually make decisions and communicate them properly, one is monetisation/advertising and the other is Infra planning. _Everything_ else relies on people churning initiatives and seeing what sticks. With loose coordination at the centre based on who know who and who manages to convince others that "this is a Zuck priority, or related to one"
It felt very much like having a Boy king. The Boy king liked playing with toys, and if you made a toy for the king you were in favour. The boring parts were handled by "evil advisors" who are there because they don;t threaten the king's power. Everyone around the boy king is there to gain favour.
> the time to fun is/was too high and was for a long time.
Yup - I got one for the other half as a present... like an hour and a half / two hours into setup/onboarding, they lost interest, it went back in the box and never came out again. :(
> Is Facebook really set up such that one person's whim is the single point of failure?
When I was there (pre-Covid) it was sort of a worst-of-all-worlds situation, compared to other firms.
On the one hand, Zuck maintains an absolute majority of voting shares, so what he say literally goes, with the board having no real authority to rein him in. If your project is something he takes a direct interest in, you are automatically subject to his whims.
On the other, Meta highly values the idea that they are a pretty flat org with no centralised command and control structure. So if your project is not under the baleful gaze of Zuck, there's a good chance that nobody in the executive suite has any fucking idea what is going on in your part of the company.
Contrast this to Bezos-era Amazon, where Bezos would sometimes directly intervene in pet projects like the FirePhone, but the entire company has a strong reporting hierarchy, and executives are expected to maintain direct command-and-control at all times over their reports (i.e. when Bezos sent one of the dreaded question-mark emails, the entire management chain damn well better be able to get their story straight top-to-bottom by the end of day)
To me, with experience of numerous organisations of different sizes outside tech, it is pretty surprising.
(I'm not arguing that this is right but) the typical progression of an organisation as it scales is to move away from the 'scrappy startup with a CEO-dictator' and towards something more mature. Obviously, there are reams of business literature written about growing pains and then stagnation in large companies, but the single-personality-driven model seems hugely flawed - look at Tesla, for example. And I'd certainly expect a public company of the size, resource, and maturity (in years, if not structure) of Meta to have developed beyond this point.
Honestly, that a number of people seem to not grok my questioning this, is possibly quite revealing about the monoculture of the tech world.
It's not that you want teams to be able to go rogue - you want teams to be able to work against a stable mission statement, that doesn't change every 5 minutes as the CEO changes mood
That's on the company owners[1], as represented by its board.
---
[1] Companies like Meta actually has two types of ownership: ownership of the company's current assets (economic equity), which is not the same as ownership of control in the company's decision making (voting power). The owners I reference here are the second category of ownership.
In the example of Meta, a quick search suggests Zuckerberg holds about 61% of the voting power.
With enough encryption, obfuscation, and security-through-obscurity, you can make it extremely difficult to obtain those keys.
Companies like Microsoft, Nintendo, and Sony ship consoles that are the target of a very motivated black market/cheating industry, and it usually takes years before any serious leaks surface.
Well yeah but those are more hardened and also the keys alone is generally not enough because they are asymmetric. So the private part is kept somewhere on some HSM.
I totally believe you that people inside have asked. I think part of the problem is that the people actually making the decisions higher up do not care and aren't really invested in these issues.
To the point that some guy vibecoding for a bit and deciding it was fun had more effect on the company's ability to open up an abandoned product than an entire consumer movement that has been lurking in the zeitgeist for ages.
I was in the Oculus for Business (later AR/VR for Work) group, and we were well into taking Portal and making it compatible with corporate MDM policies, to sell along with Workplace as a general visual conferencing platform. Sadly all went up in smoke when they cancelled the main Portal program.
I always wondered how they messed portal up. It seemed so natural of a piece of hardware to fit into everything else they do. Tying it to horizon says all I need to know.
> Could I use GS to render my living room as a digital playground to train my vision models in?
Yes, its what the autonomous car people are doing.
However you might want to do photogrammetry first (https://github.com/alicevision/meshroom opensource) as that produces a mesh which you can use to detect collisions easier. The downside is that transparent objects render really badly. but it is a lot faster to render
Ehhh if it's just for looking and you don't have anything lidar just go for splats they're way better behaved, mostly because they don't need to understand a concept of "surface" they just understand "splat with spherical harmonics of view-dependant color".
> we didn't know how to pronounce hormuz until bombs started falling
I mean that might be true for you, but its been on the radar over here for a number of years. (it was also a key risk with Iran identified by a number of key intelligence think tanks. especially around the time of the anti-nuclear treaty negotiations)
But to your point, youre right its not an energy issue, its an energy and materials issue. There are lots of pre-cursor chemicals and products that come out of that region that the modern world needs to function.
1) we were always told and legal always pushed back hard on anything face detecting.(ie haar cascade "this is a face" let alone actual this is dave/sally)
2) the FTC would audit us to make sure we weren't doing that kind of stuff
3) all of the research prototypes had inbuilt/inline face removers up until 2024(I left after that so I don't know when/if that changed)
3.1) One of the very first things I worked on was face removal, it was a central core of the entire fucking project. Like if we didn;t have any of those constraints we'd have been 2 years ahead.
4) Stella is the name for v1 rayban stories, so its very odd that they get the update when they've not had any new features since for a long time(unless I am mistaken).
now, I will say that Boz was pushing to have facial recognition when I was were, and some of the early storyboards for XROS were pretty reilant on it (ie in an office having the sims like diamonds above your head that indicated who you were and if you were busy)
I assumed that Zuck said no because he'd had enough time with the lawyer and the FTC sniffing about to not bother.
However the glasses based AI lifelog stuff (which was basically a really effective personal assistant) would be a lot more effective if it could use facial recognition (we weren't allowed to use speaker diarization as that would allow us to record individual audio from users and recognise them like with facial recognition)
I know a few folks who work in FTC compliance roles at Meta. I don't think the wider nerd-sphere appreciates the extent to which the FTC regulates Meta following the consent order after their privacy scandals of the 2010s, or the sheer number of roles within Meta now working only on FTC related compliance. I'm sure Zuck is no doubt desperate to ramp it down.
“We will launch during a dynamic political environment where many civil society groups that we would expect to attack us would have their resources focused on other concerns,” according to the document from Meta’s Reality Labs, which works on hardware including smart glasses.[1]
I recall that Facebook asked me to identify faces of my friends in order to "verify" myself about 15 years ago. Do you know whether Facebook still stores that data?
Wired reported in 2021 Facebook said it would stop using facial recognition technology to identify people in photos and videos and delete accompanying data on more than 1 billion people.[1]
> Do you know whether Facebook still stores that data?
Honestly I couldn’t answer that. I never really touched production userdata (mainly because it was scary and also it was in PHP or some horrid transpiled interface to PHP)
My gut feeling is that facebook doesn't throw data away unless its forced to. So its probably there on your graph somewhere.
Pure speculation, probably they finally figured out the correct legal ToS and privacy policy and everything else that made them feel confident + some regulatory/lawmaker discussions reached a certain point that they finally decided they could do it? Does that add up?
I think the ToS is lenient enough to get away with most things.
I suspect that its a two-fer,
1) zuckerberg has said "it must be done" as part of the AI push
1.1) it might be also that Wang has pushed to get that data, but thats a guess. I doubt he has that kind of sway
2) they've realised that the FTC isn't either capable, or they have bribed the right part of the government to avoid getting nailed.
The thing that gets me is the number of lawyers that are there, and the sheer amount of process that is there to stop this kind of thing happening requires Zuck to explicitly say "I WANT THIS" repeatedly.
> probably they finally figured out the correct legal ToS and privacy policy
Unlikely IMHO - the person who agreed to the TOS is the one person the glasses don’t record.
More likely they’ve decided to launch it and see what happens; they can always withdraw the feature later, and laws can be surprisingly flexible when you’re a large corporation.
It also suggests a certain amount of self-policing by Facebook, which leads to obvious failure cases:
Since Mark Zuckerberg attended Trump's most recent inauguration celebration and NOT non-billionaire average citizens, by and large, I speculate that the FTC threat is no longer a concern to Zuckerberg's Meta corporation. Back scratching all around?
It also requires more infrastructure to get yield. In theory all you'd need to have is these etched metal plates, a transparent dome and a source of briny water. (and a cleaning mechanism)
The etched plates creates 100% humidity (probably more as it'll condense out)
reply