Facebook Research

Holographic Optical Elements certainly have potential for applications like this, and the 3D Hologram Printer could be a great tool for simple prototyping and quick iterative testing. Recently, Maimone et al at Oculus/Facebook Reality Labs used our hologram film to make a design for a thinner VR headset using pancake lenses incorporating holograms: https://research.fb.com/wp-content/uploads/2020/06/Holographic-Optics-…

> the best data point FB Do you have any evidence to back this up? I can't find any published literature where Facebook has hinted this, including this paper [0] saying: > As expected the most important thing is to have the right features: those capturing historical information about the user or ad dominate other types of features. 0. https://research.fb.com/wp-content/uploads/2016/11/practical-lessons-f…

Facebook Research estimated 3.5 degrees of separation for users on its platform back in 2016. I'd be very interested to see any follow-up studies. https://research.fb.com/blog/2016/02/three-and-a-half-degrees-of-separ… Personally, I got a score of 2.91 degrees of separation at the time, and at that point I had 3715 friends. Now I'm at 4037. I only add people who I met in real life, even if I only met them once at the CouchSurfing meetup or a similar international group. I try to get back in touch with people when I travel, so I can be welcomed in every city around the world (and I try hard to welcome others when they come to wherever I'm based).

Here is a paper where they had users training a bot to follow request (given in natural language) of increasing complexity to accomplish task in a simulated environement : https://research.fb.com/wp-content/uploads/2018/04/mastering-the-dunge…

Amazing!. The actual research paper: https://research.fb.com/wp-content/uploads/2019/10/Live-Face-De-Identi… As super interesting this is I'm not sure yet how this could be deployed for people who are concerned about their privacy in the short term. If you didn't put pictures on to social networks and only sent stuff end to end encrypted your family and friends can see the images / videos.

I'm not following. I understood from your first comment that you think the amount of data is low ("underwhelming") and from your last comment that it's a lot ("that much data"). In any case, the data is "whatever needs to be logged". And it's not "server logs", which is what I'm interpreting from your comment. Scribe transports most data at Facebook to be processed by real-time systems (e.g. Puma, Scuba) and also "batch systems" (data warehouse). So, it's quite a lot, being "the ingestion pipe" for Facebook. Does this answer your question? :-? Puma: https://research.fb.com/publications/realtime-data-processing-at-faceb… Scuba: https://research.fb.com/publications/scuba-diving-into-data-at-faceboo…

They have a very strong research arm covering a significant number of research areas. https://research.fb.com/

The Facebook paper on this topic ("Legacy Contact: Designing and implementing stewardship at Facebook") is one of my favorites from their publications repo. Not highly-technical, but relatively detailed as it explores the human side of the design tradeoffs needed to accommodate an unavoidably difficult and varied situation: https://research.fb.com/publications/legacy-contact-designing-and-impl…

Bitten by queues in every team I was at AWS: > As a rule of thumb, target utilization below 75% This is one good reason to rely on serverless. Simply outsource the problem to a system that knows how to handle this better. > Steer slower workloads to paths with lower utilization This is fraught with all sorts of perils [0], so take caution going down this valid but tricky route. > Limit variance as much as possible when utilization is high This is key, and one of the most elegant solutions to this problem I know of comes from a Facebook talk on CoDel + Adaptive LIFO. [1] > Implement backpressure in systems where it is not built-in Making downstream dependencies behave is never an option. Selectively isolating noisy neighbours, if possible, from other well behaved clients, tends to work well for multi-tenant systems [2], in addition to monitoring long work (by dropping work that takes forever) [3], or better yet, doing constant amount of work [4] (aka eliminating modes) [5]. > Use throttling and load shedding to reduce pressure on downstream queues. One way is to impl admission control (ala Token Bucket) to minimise the impact of thundering herds. Though, clients do not take kindly to being throttled. [6] A great article; many have been written at this point. Very many still have been burned by queues. Though, in my experience, without an exception, some component somewhere was always building up that backlog! [7][8][9] [0] Interns with Toasters: How I taught…

Strangely, https://research.fb.com/ is UP.

Facebook has a similar system, unsurprisingly. An agent runs on every host to sample 1 in N packets, accumulating counts by source/dest host/cluster/service/container/port and so on before sending the aggregate data to Scuba [0] for ad-hoc analysis. This tool was really useful — in a matter of seconds you could see traffic types and volumes broken down by almost any dimension. Did service X see a huge jump in traffic last week? From where? Which container or service? How much bandwidth did your compression changes save? And so on. It also had some really neat stuff to identify whether or not a flow was TLSed or not, which was crucial for working out what still needed to be encrypted in light of the Snowden revelations. TCP retransmits were also sampled in detail. Being able to see at a glance if e.g. every host in a rack is the source of (or target for) retransmits made troubleshooting much faster. These systems were really awesome and a good example of what could be built with relatively little effort (the teams involved were small) when you already have great components for reliable log transmission and flexible data analysis. [0] https://research.fb.com/publications/scuba-diving-into-data-at-faceboo…

Closest I can think of is Purely Functional Data Structure (Chris Okasaki) or Category Theory for Programmers (Bartosz Milewski), but they're not exactly what you're after. I would submit that this paper from FB is more in the vein of what you're looking for: https://research.fb.com/wp-content/uploads/2020/08/Eliminating-Bugs-wi… . This interview might also be insightful: https://www.microsoft.com/en-us/research/podcast/functional-programmin… The truth is that Haskell is still very academically focused, so you're likely to see papers of substance much more often than books or blog posts. When you hire people to write Haskell there's two groups that always show up: young enthusiasts who are frustrated with imperative programming in their (usually first or second) day job, and academics who got transplanted into industry to work on hard problems. I've interviewed people doing formal verification of CPU circuits at Intel, people who work on compilers, people who work on verifying termination of programs (for missile guidance), and people who work on financial institution backends. What I haven't seen is too many experienced, pragmatic engineers (rather than computer scientists) who have spent their career writing Haskell.

Hmm and global shutter cameras are much more expensive. Though AFAIK a modern mirrorless camera with mechanical shutter can be used for this, yes? My GH5 can shoot at 12 frames per second with the mechanical shutter, so that may be of some use. I did see in this paper that it seems like rolling shutter compensation is possible: https://research.fb.com/wp-content/uploads/2019/09/An-Integrated-6DoF-… I wonder if a library could be produced for this? As far as my street capture I am currently using 830 frames from my 4 minute video (2 fps), images are 4992 x 3744 shot with a 12mm lens. The reconstruction is still texturing so I don't have the final result. What kind of numbers are you interested in? Please contact me yes! My email is in my profile. See also my four wheel drive robot with four 4k cameras, and feel free to actually initiate discussion on that site where others can see it too (or emails is good). Thanks! https://reboot.love/t/new-cameras-on-rover/

Hey, Facebook VP of Integrity here (I work on this stuff). This WSJ story cites old research and falsely suggests we aren’t invested in fighting polarization. The reality is we didn’t adopt some of the product suggestions cited because we pursued alternatives we believed are more effective. What’s undeniable is we’ve made significant changes to the way FB works to improve the integrity of our products, such as fundamentally changing News Feed ranking to favor content from friends and family over public content (even if this meant people would use our products less). We reduce distribution of posts that use divisive and polarizing tactics like clickbait, engagement bait, and we’ve become more restrictive when it comes to the types of Groups we recommend to people. We come to these decisions through rigorous debate where we look at all angles of how our decisions will affect people in different parts of the world - from those with millions of followers to regular people who might not otherwise have a place to be heard. There’s a baseline expectation of the amount of rigor and diligence we apply to new products and it should be expected that we’d regularly evaluate to ensure that our products are as effective as they can be. We get criticism from all sides of any decision and it motivates us to look at research, our own and external, analyze and pressure test our principles about where we do and don't draw lines on speech. We continue to do and fund research on…

Hi there :) Apologies for the super long response, but you had a lot of points. > Am I really missing something here or this thing is a complete nonsense with no actual use cases what's so ever in practice? Hopefully you're missing something, or we've been wasting a great deal of our time ;) > There are a number of off-the-shelf models that would give you image/sentence embedding easily. Anyone with sufficient understanding of embedding/word2vec would have no trouble train an embedding that is catered to the specific application, with much better quality. For images and text, it's definitely true that you can train your own embeddings with an off-the-shelf model. But I think it's more likely that we end up in a place where a small number of people train a bunch of really good models and everyone else uses them. I think this for three reasons: 1. It's what we've seen with word2vec. The vast majority of people that use word2vec aren't training it themselves, they're downloading pretrained weights. 2. Most people don't have enough data to train a good embedding themselves. There are good public datasets for images and text, but we're planning to produce embeddings for more niche verticals too. Keep in mind that modern deep neural nets are very data hungry, and the problem gets worse every year. In a few years I think we're going to be in a spot where getting state of the art performance requires a lot of compute, and more data than most people have access to. 3.…

Facebook Research has indeed made a prototype for a sunglasses-sized VR headset! https://research.fb.com/wp-content/uploads/2020/06/Holographic-Optics-…

Similar post from FB from 2018 - https://engineering.fb.com/2018/11/21/developer-tools/predictive-test-… and a more recent one (though just the raw technical paper) from 2020 - https://research.fb.com/wp-content/uploads/2020/12/Predictive-Test-Sel…

Here is a link to all the tools we provide: https://research.fb.com/data/

I don’t know, the paper[1] describes how they setup their systems so I think it is relevant to what happened [1] https://research.fb.com/publications/running-bgp-in-data-centers-at-sc…

Facebook's News Feed model was trained daily in 2018. (Source: https://research.fb.com/wp-content/uploads/2017/12/hpca-2018-facebook.… If you know of update please let us know.) "requiring constant retraining" is pretty accurate in this case. Facebook trained image understanding CNN ("Lumos" in the paper) only once in many months, but this is not usual.

Direct source is available at: https://github.com/facebook/mysql-5.6 The social graph is indeed stored in MySQL via TAO. A recent white paper about MyRocks is: https://research.fb.com/publications/myrocks-lsm-tree-database-storage…

This is also one of the tricks used to get real-time 3D reconstruction on the Oculus Quest. https://research.fb.com/wp-content/uploads/2020/07/Passthrough-Real-ti…

What you saw likely wasn't a UPS, but an RPP > A data center typically spans four rooms, called suites, where racks of servers are arranged in rows. Up to four MSBs provide power to each suite. In turn, each MSB supplies up to four 1.25 MW Switch Boards (SBs). From each SB, power is fed to the 190 KW Reactive Power Panels (RPPs) stationed at the end of each row of racks. https://research.fb.com/wp-content/uploads/2016/11/dynamo_facebook_s_d… The UPS role is taken more by BBUs (Battery Backup Units). https://www.youtube.com/watch?v=KNsposM0sJE has lots of info about them. (I work for FB, on entirely unrelated things)

Literally any analyst or data science job related to growth or user engagement on social media. https://research.fb.com/category/data-science/

Pretty sure this paper describes what they're doing now: https://research.fb.com/publications/flighttracker-consistency-across-…

Similar to what Facebook does automatically [0] which was easily visible with an outage they had in 2019 [1]. [0] - https://research.fb.com/videos/automatic-alt-text-for-photos-on-facebo… [1] - https://www.theverge.com/2019/7/3/20681231/facebook-outage-image-tags-…

Facebook might disagree with your assessment... hand tracking research for VR has come a long way: https://research.fb.com/publications/constraining-dense-hand-surface-t…

* 50% of all US currency is printed between now and the end of 2021 * 1 Bitcoin will cost over $100k on Dec 31, 2021 * An animation tool rivaling Adobe Flash for the web will emerge * FB releases a thin virtual reality headset https://research.fb.com/blog/2020/06/holographic-optics-for-thin-and-l… * Austin Tx or Seattle Wa succeeds Silicon Valley as the next major tech hub. San Francisco and its century old Victorians become the Detroit of the tech world * Section 230 doesn't get repealed * WFH and quarantine continues until the summer * A React competitor that compiles to WASM with promises of perf and space gains will emerge * A Reddit competitor emerges and wins majority marketshare * 2021 will be the year of punk and rock n' roll

There’s a host of further info in the paper itself: https://research.fb.com/wp-content/uploads/2020/06/Holographic-Optics-… “Finally, holographic lenses can be constructed so that the lens pro- files can be independently controlled for each of the three color primaries, giving more degrees of freedom than refractive designs. When used with the requisite laser illumination, the displays also have a very large color gamut.” They have full colour working, and apparently well, in the benchtop prototype. There’s discussion towards the end as to options for implementing full colour in the HMD prototype.

This is the hardest source I can find, but it only measures what happens on Facebook. The numbers do seem higher than what I'd expect for IRL conversations, though: https://research.fb.com/blog/2015/05/exposure-to-diverse-information-o… > Online I am connected to those whom I socialize with or am otherwise professionally connected to. In the “real world” this constraint is largely absent. This seems entirely backwards to me? Maybe you talk more with strangers IRL than online, but I doubt it. I only have n=1 (me), but we are talking right now. Who knows where we live in relation to each other? So much of politics is split between urban and rural environments. Those groups are defined by where they live, so I expect very few conversations in person between the two, especially about politics.