Given how much was announced during Facebook developer conference F8, I thought I would share some thoughts on the major announcements.
Facebook announced it will be soon launching a dating feature. By clicking on a heart on the top right of your ur profile you will now be able to create a profile that’s only visible to non-friends who’ve also opted into the feature. Facebook will match you based on all its data, and messaging will happen in a dedicated inbox rather than Messenger.
- As I pointed out in my article on Wed., this is a tough time for Facebook to be launching something as personal as a dating service given the recent privacy questions raised by the Cambridge Analytica breach.
- Theoretically, the data Facebook has is quite valuable in creating a set of matches based on information shared on users profiles as long as of course people are being honest about what they post. I assume the algorithms used for the matchmaking will also be based on actual information volunteered by the individual.
- Internet dating is not necessarily something you want to share with family, friend and coworkers, which is something Facebook thought of and addressed by not making the profile you create visible to your connections. That said, I do have to wonder if using the data you share coupled with location and events you attend would dramatically increase the chance to have your profile shared with someone you might know in real life.
- Competition in this segment is quite strong with apps that show a higher level of sophistication. Yet, if I look at the sweet spot of the Facebook demographic I do wonder if the appeal will be mostly with older users who might feel comfortable with the platform precisely because they already use it and offer a simpler approach to dating.
The company will soon launch a new privacy feature that will allow users to see and delete the data Facebook has collected from websites and apps that use its ads and analytics tool.
- The exact availability of the feature was not announced, but Zuckerberg said such feature is under development and will launch soon. Clearly, this was done in response to some of the questions raised during the recent congressional hearing.
- While actions should speak louder than words it was really how Zuckerberg talked about the feature that says a lot about his lack of belief this is the right thing to do.
- Zuckerberg pointed out that the feature will make Facebook feel less personal but that this is what consumers were asking for. I thought such a wording was quite condescending and positioned the decision of rolling out the feature to be responsive to a consumer request he did not believe to be right.
- Warning about a less personal Facebook is, by all means, a fair point but the proper focus of the statement should have been on the fact that giving consumer this option was the right thing to do.
Oculus Go and Oculus TV
The wire-free VR headset Oculus Go is now on sale at $199 for the version with 32GB of onboard storage and $249 for the 64GB variety.
- Most of the reviews I read so far can be summed up as decent but not ground-breaking. That said, I think the ease of use that comes from the combination of an easy setup and the lack of cable connections cannot be underestimated.
- A short battery life is somewhat disappointing as this is one of the shortfalls of mobile solutions like Daydream and Galaxy Gear VR that drain the phone quite rapidly and overheats it making prolonged sessions difficult.
- It would be interesting to see how quickly we will see the price point drop as this, for me, would be the best indication of how committed Facebook really is to drive adoption
- Resolution is similar to what can be found with Daydream and Galaxy Gear VR. That coupled with three degrees of freedom rather than the six found in the upcoming Oculus Santa Fe. Basically, you can go side to side, and up and down, but you can’t move closer or further away from an object – and if you try, you’ll start to feel a little nauseous.
- Together with the availability of Oculus Go, Facebook announced Oculus TV. The app puts a TV experience into your virtual environment with specially adapted on-screen controls, which turn the virtual screen in the virtual room into a sort of TV streaming Box experience. The streaming service available later this month will support Facebook Watch, Red Bull TV and Pluto TV directly integrated at launch.
- Despite what Facebook said on stage about being currently working with networks to bring native integrations of Netflix, Hulu and Showtime I am just not sure what the incentives for these brands are as they already have standalone apps in the Oculus store.
- VR in the consumer space remains a chicken and egg problem with content developers waiting for a bigger addressable market and potential buyers waiting for more content. Hopefully, Oculus Go will help broaden that base, although I remain concerned that while the price is more appealing the experience is still not good enough.
There were a few announcements around AI at F8:
On Wednesday, Facebook revealed that it has formed a special team and developed discrete software to ensure that its artificial intelligence systems make decisions as ethically as possible, without biases.
- Facebook is not the first big company talking about the importance of ethics in AI but of course, this does not make it less important. As I will never get tired to say, the industry should prioritize ethics in AI because we cannot afford to train machines to be as biased as we are. Microsoft research organization established a Fairness Accountability Transparency and Ethics group over a year ago and last year Alphabet’s DeepMind Ai group formed an ethics and society team.
- A couple of weeks ago I talked about how AI is not going to be Facebook’s magic wand to solve hate speech on the platform and this announcement is particularly relevant to the points I was making in my article.
- Facebook also develop a piece of software called Fairness Flow, which is now integrated into Facebook’s FBLearner Flow.The software analyzes the data, taking its format into consideration, and then produces a report summarizing it.There are currently no plans to release Fairness Flow to the public but Facebook’s doesn’t plan to release the new Fairness Flow software to the public under an open-source license. However the team might publish academic papers documenting its findings, Facebook said.
Instagram’s hashtagged images help with machine learning
- The idea makes a lot of sense. You take all the pictures tagged as #dog or #cat and you use those to train the machine rather than having to go through a set of pictures of animals to tag dogs and cats before showing those to a machine. Facebook used around 3.5 billion Instagram photos (from public accounts) and 17,000 hashtags to train a computer vision system that they say is the best one that they have created yet.
- When humans tag pictures and then feed them to a machine the process is said to be “fully supervised” As you can imagine this system is pretty error proof but might be difficult to scale or might not be representative enough of the types of cats and dogs in the world.
- The Instagram hashtagged pictures offer a “weakly supervised” process where machines can use users’ generated tags to be trained.
- This “weakly supervised” training is much more “noisy” which creates some challenges on accuracy. For instance, you might have tagged a cat as a dog as a joke or you could have tagged a picture of you and your dog as “dog but you don’t see the dog. Computer vision lead at Facebook Manohar Paluri, said however that the results they achieved are very encouraging. Measured by one benchmark, the system—trained on those billions of Instagram pics—was on average about 85% accurate, he said.