schrep

Your work was so impactful over a long period from Firefox to Facebook. Honored to have been a small part of it.

Wow small world indeed!!! The team at CenterRun was amazing

I’ve worked directly for Mark and closely with Sheryl for nearly 14 years and this is very accurate.

Congrats on this huge milestone!

So refreshing to see a leader in this field say “we are not sure which one will work out” rather than just hyping their stuff.

Can I get a test ride soon!

Do you have a reference/citation for any system that does 100x100 translation? That's the claim here - first to do 100x00 language directions from a single model. But if we are wrong I'd happily update with a reference/citation.

“ Neo 2/Neo 2 Eye is available for businesses only. Please fill out this form if you are interested in ordering the Neo 2/Neo 2 Eye and we will get back to you once stock is available.”

Tracking controllers in 6DOF from headset cameras is the hardest part. Nobody has yet replicated this for a stand-alone headset.

What standalone 6DOF headset with 6DOF controllers existed in the market in 2019 besides the Quest? Or now? I certainly can’t think of one?

Our products aren't perfect, and we understand that we have a lot of work to do.

However, the fundamental purpose of our products is to allow people to efficiently communicate with each other. Hard for me to square that with "drain on society." I have many friends who, via Facebook, found a connection that was life changing: from finding a job, a spouse, to a community to deal with the loss of a loved one or support after being diagnosed with a terminal illness.

One of the things that draws AI researchers to come work at Facebook is the opportunity to see their work make a positive impact on billions of people around the world.

The research done by FAIR is helping us do things like deliver billions of translations a day, provide automatic photo captions for people who are visually impaired, and help bring blood donors and people in need together. It also helps us spot when someone is expressing thoughts about self-harm so we can alert first responders.

But we also believe there's even more we can do to help bring the world closer together, to give people a voice, and to open up new opportunities for everyone. AI is a key part of that and we believe pretty deeply in the power of open research to help not just us but the whole industry.

People have been able to download their Facebook info since 2010: https://techcrunch.com/2010/10/06/facebook-now-allows-you-to...

Available here: https://www.facebook.com/help/131112897028467

If these are posts by other people that you were tagged in then those posts should still be up, just without the tags of you, on the original posters timeline.

No - we've had the download your info feature since 2010 (https://techcrunch.com/2010/10/06/facebook-now-allows-you-to...). There is just a lot more public scrutiny now of what's in there - which found this bug.

This was a bug. There was an old feature that used to allow you to record and post directly from the browser. Those videos were streamed to FB as they were being recorded. If you decided not to post those draft videos should have been deleted but were not. They showed up in download your information (DYI) as expected because that tool is designed to show you the data Facebook has about you. Thanks to New York Magazine for the flag. If you see anything in DYI that doesn't look right, let us know and we'll investigate. This was a bug, and we really do appreciate any help in finding them so we can fix them.

I totally get that view and it bums me out. I can say there is reasoned debate about the different licenses as you can see in this thread. We've changed an explicit patent grant into a debated implicit one. We thought the first was clearer to folks but when debate turned into action we decided winning the debate wasn't important we should just do what would allow people to use the code. I know big companies can look opaque and self-interested on the outside so I get your view. Just sharing as someone part of the internal discussions what happened and why.

Actually we just thought it was a shame that people wouldn’t use it because of the license. Not much more complicated than that.

Doesn’t mean I’m right :) - see above as he had a point.

That’s impressive work. Still don’t think we have reached human level for all the categories of things we see in images. But you are correct that my comment about 1k categories is not true for many production systems.

For example marketplace (where you can sell items on Facebook) will suggest a category for the item if you upload a photo.

With what precision and recall?

This is some of the most advanced work out there - but CV is not “solved” most vision systems only can label about 1k categories of objects. So capatchas can still be easiy constructed that would fool these systems. Part of why it is exciting to get this out there others can help us improve it.

This was a result of us throttling traffic to the site - not the original outage.

We actually encounter Thundering Herd problems on a very regular basis. The Starbucks page has nearly 14M fans and posts may get tens of thousands of comments/likes with a high update rate. You have a lot of readers on a frequently changed value which means it is not often current in cache and you can have a pileup on the database.

Since we encounter this on a regular basis we have built a few different systems to gracefully handle them.

Unfortunately, the event today was not just a thundering herd because the value never converged. All clients who fetched the value from a db thought it was invalid and forced a re-fetch.