#Fun testing boundaries of #DeepFace

Since the announcement of DeepFace and its consecutive reasonance in the media, the facial recognition algorithm from Facebook, it aroused both interest and critique. There are many arguments to consider privacy issues before sending out these types of identity related software's out there ... into the public world. But no matter what the status of the philosophical decisions is, DeepFace is now ready to be fully deployed after a successful pilot.

Every type of technology is embedded in a context and ecology, which makes it an integral part of a holistic society. And as a human instrument, it inevitably leads to many discussions whenever it changes contemporary habits. Nevertheless, each technology also adds to a bit of fun. And I see it as an informal duty of each learning technologist, tech geek... or all-round nerd-joker, to investigate the fun-factor of these types of algorithms. And that was what I was thinking about during last night.

The DeepFaced-facts

• Deep Face recognises more people than I ever would (I have trouble recognizing faces, and not coming anywhere close to the 97,5 % average of most people), and has almost reached human recognition stats (97,25 %).  
• The rotation challenge: DeepFace uses a 3D model for rotating faces virtually so that the person in the photo appears to be looking at the camera. 
• The algorithm draws its power from Deep Learning, a visual as well as audio (language) recognition system set-up by Google. Where deep learning has reignited some of the grand challenges in artificial intelligence, due to its use of computational power, use of big data, and adaptation capacity.

So take DeepFace to the challenge
Provide DeepFace with some additional challenges, while at the same time expand your EdTech tool-use
History is being rewritten, we all know this, and most of the time history is written by the victors (Churchill). It will never be different, nevertheless, it might be fun to try and contaminate some of history's facts with us - the normal people. Which also makes it into a nice 'how would you use this tool'-action for any multimedia class, online or face-to-face. Some options:

• Photoshop yourself into (Facebook) history. It almost feels like old-school this photo-shopping, but it never hurts to rethink old options. By placing yourself into histories key moments, Facebook might pick-up your presence at these key points, and of course Deep Learning might adjust itself to 'this person could not have been here!?!', but then again it might start to calculate you must have been here if you work yourself into these picture from different angles (in doing so, making yourself even more experienced with photoshop). Me with Ghandi, me with the new Greek president Alexis Tsipras (would love this), me with ...
• Exploring the boundaries of morphed images and DeepFace. Another fun activity, that will allow you to see how much tweaking you can do to your own face, before DeepFace stops recognising you. As a test I already morphed me with my son. Quick online morphing option: 3Dthis.com .
• Finally an answer to 'does everyone on earth have a (or more) look-alike/s'. If facial recognition is indeed working, it might reveal that there is another Ignatia out there somewhere... and I would like to meet her, facebook might make this possible (what is the return rate for DeepFace on successfully recognising twins?). But I do hope my look-alike is not mixed up with too much hustling though... how dangerous could that be for my identity? and what if people make masks mimicing my face?... 

As you can see: fun guaranteed. I feel that I should add this concept of the Fun-test to my repertoir on getting and screening new technologies.