Facebook's AI lab
facebook.com
facebook.com
Here are some things where smart AI could help: 1. Better filtering of spam bots. 2. Better filtering of ads that would significantly improve ad quality. 3. Smart assistant that learns your habits and makes smart suggestions such as: which events to attend, your writing style, and filters immediate notifications based on personalized importance score.
Or Facebook could exploit machine learning/AI techniques to simply drive up click through ratio on their ads.
Recognize your romantic partner using FB network[0]
People are already used to people saying and doing dumb things online, so something that is nearly "smart human like" is a big advance (on average).
I'd imagine projects that group similar topics (i.e. these people are all in front of the Great Pyramids or these people are all taking pictures of their babies) which probably will help group/filter content
AI that has personal information of 500M+ people, using it to manipulate people...(first to click on advertisements, and then much more). In the hands of government, I shudder to think what's possible. With NSA already snooping around, perhaps it's not all that distant.
http://web.mit.edu/people/amliu/Papers/PentlandLiu_NeuralCom...
Just saying.
EDIT: Just came across an article: "The Mother of All Demos" is 45 today; Douglas Engelbart did exactly what I'm hoping someone today will be able to pull of.
I don't think the Neural Net library was quiet what it was in Lush, but it was getting there.
How many people are they teaching in person these days?
I wouldn't be surprised if, at this point, his work split 90% FB, 9% NYU Center for Data Science, 1% NYU CS.
LeCun is now running a high profile lab for a $100B company. He's got the personal attention of the founder. He's only going to be at NYU as a talent conduit. It's not the same as Hinton's arrangement at all.
Pulling someone to industry doesn't stop one from contributing back to the field. This is what they do, what people actually care about. In fact, if you are a big player in a field, you spend most of your time conducting your research and running a lab, not teaching classes. This is the norm and half of the famous professors I had barely ever come to lecture, only the TA would show up (some schools are strict on this though, like MIT). And I often tell people if you want real good solid education, especially for an undergraduate, don't attend big player's lecture. Your TA or your average not-so-well-known professor can be more helpful.
Those who know, do; those who understand, teach.
-Aristotle.
Obviously it's an iterating/evolving meme.
[1] http://answers.yahoo.com/question/index?qid=20100823190727AA...
I own Facebook stock, so I might be biased.
https://plus.google.com/103199870243230279182/posts/JaqgU1sX...
"Fucking disgusting"
>> Regardless of whether they are good or evil, what do you think the potential implications are?
Less privacy, more data in the hands of governments, more police state.
I think Moxie Marlinespike's talk at Defcon 18 (2010) was a really good one: http://www.youtube.com/watch?v=eG0KrT6pBPk
He talks about how we do not really have a free choice when it comes to things like having a Facebook account or having a mobile phone. Namely due to the network effect, you don't really have a choice if you want to be a normal member of society.
As repeated by resonator several posts up.
As I see it, Facebook could at least attempt to act like they have the user's best interest at heart. This means actively educating them about privacy, risks, and not setting defaults to public. Also not everything on private as this would confuse some people, but at least use non-public and reasonable settings. And/or have a public discussion about it.
Facebook is becoming ubiquitous enough that it's almost something like "internet" or "calling" in general. For such services we have made laws, e.g. in the Netherlands there is a net neutrality law, and sending spam is illegal. But for individual website we have no laws, nor do I think we can make any in the near future without a huge lot of fuzz about it. I'm not saying governments should make laws for Facebook to follow, I'm just saying it won't any time soon, if at all.
So then how do we make sure such a big service acts in our best interest? I think Facebook should start acting less like a for-profit company and more like a government would. Not with elections per se, but at least act in the public's interest instead of working against the users.
And that is how I'd answer the question of "how is this Facebook's fault?". Being a popular for-profit company is not a 'fault', but with so many users they have a responsibility, and I think that it's fair if we, hackers that understand the technology, try to make them act responsibly.
If it's my goal to create a product that forces people via social pressure to do something, it's definitely my fault when it works.
Your response is the same as people saying "don't fly" when they complain about the TSA. 'Someone had an unpleasant experience with the TSA? How is that the TSA's fault? That person didn't have to fly.'
Facebook has designed a killer product that makes it really easy to use for social interaction. But Facebook is not forcing you to use it, Facebook is not forcing your friends to use it, Facebook is not forcing social peer pressure on you, and if in fact someone's friends are wholesale excluding that person because they're not on Facebook, Facebook is not responsible for the fact that someone has shitty friends that are too lazy to include that person.
This is one thing that people often miss-understand, just because there is no explicit threat of retaliation does not mean systematic coercion does not exist.
"Facebook CEO Mark Zuckerberg, CTO Michael Schroepfer and I are at the Neural Information Processing Systems Conference in Lake Tahoe today. Mark will announce the news during his presentation at the NIPS Workshop on Deep Learning later today."
It seems Zuckerberg will have an apres-ski Q&A, and be in a panel discussion at day's end (https://sites.google.com/site/deeplearningworkshopnips2013/s...). Hope the room is big enough.
I bet the guys over at Fort Meade are beside themselves with happiness over this :-)
But Facebook? They could create a million virtual DanBCs and then A/B test something against them, and then present me with the irresistible ad, perfectly pitched to draw me in?
Microsoft Research, for example, doesn't restrict itself to the domain of features for Windows, Office, or XBox. Instead, you get giant touchscreens (which you could say inspired Windows 8's unified tablet-PC approach, but didn't really) and PhotoDNA.
Likewise, I'd expect the things coming out of a Facebook AI Lab to be more of the "new ideas that could replace our entire business model when this one runs dry" than "better ways to advertise." Personal, private social agents that live in the cloud and continuously execute on your goals without requiring direct instruction, say (I'm sure there's a speculative-fiction name for these, but all that comes to mind is "NetNavis.")
Not to mention the huge amounts of data they collect through searches and Gmail that they already have been using to tailor ads to each visitor.
I see Google announcing AI research, and I think "self driving car" or "understanding search terms", but I have no doubt that it's also "serve better ads" or "slurp more data".
You're right, I should have put that in.
It's a story about an AI who has the job "to satisfy everybody's values through friendship and ponies" in an MMORPG but breaks out and starts optimizing the real world.
I took Yann's ML class during the first semester of my program at NYU two years ago. It was a terrible experience, given that I hadn't worked with linear algebra or mathematical statistics since about 2008. I could barely write Python at the time, so my Lush code was some ungodly non-functional crap. That said, I'm very happy I took the class in 2011 despite hyperventilating my way through the final--it was only offered once more, and I suspect it will never be offered again given today's news.
Yann really knows his stuff. His convolutional nets (and their application to MNIST and image segment classification) represented a significant improvement in computer vision, and he demoed some incredible low-latency image segmentation stuff for us. He runs one of the best neural net labs in the world (up there with Hinton's and Bengio's), and he has some incredible students at NYU. I can see this shaping up as a delayed acquihire of sorts...he will not struggle to find excellent candidates.
There's a lot of discussion around the privacy implications, but I think everyone's rehashing old points. Facebook already hires excellent researchers and data scientists -- John Myles White and Sean Taylor just moved out west -- but they don't focus on images just yet, from what I know. This hiring represents an investment in image analysis on Facebook's behalf that matches what they put into unstructured textual data and graphical inference. If you've already stayed with Facebook through the "graph search" announcement, this shouldn't surprise you either.
As someone interested in this type of work, it's an exciting time to live in New York City. Finance and adtech have been here forever, but things have expanded. Many startups (Foursquare, Tumblr, Knewton, Etsy) have invested in applied statistics and machine learning, hiring excellent researchers and engineers. Columbia and NYU have announced data science initiatives in the last 6 months. Very smart people (and others, like me ;) ) are very active in the community here.
There are some obvious applications of Yann's work to Facebook's advertising goals:
> Identifying strong friendships through co-occurrence in photos
> Digit or character recognition applied to marketing in photos (shop signs, brands, etc)
> Image segment classification (e.g. beach, park, road) for use in predicting a photo's location (for those uploaded after the fact)
> All the Instagram photos. I mean seriously. They're committed to making ads seamless--why not use image segmentation + likes to identify photographic structures people are attracted to?
Can anyone think of others?