4 ms·
Nice video, but it is a mixed bag of interesting and implausible. In particular, it's unfortunate that the video had to be mostly wasted on the implausible date
by karpathy 14y ago
Nice video, but it is a mixed bag of interesting and implausible. In particular, it's unfortunate that the video had to be mostly wasted on the implausible date scene, with an app reading emotions (such as impatience) and suggesting things to do or say. I'd say that is in fact the perfect example of what NOT to expect in any foreseeable future.
- sp332 14y agoActually the Affective Computing group at MIT has been actively working on this since at least 2007. They had a fully functionaly prototype last year. http://www.newscientist.com/article/mg21128191.600-specs-that-see-right-%20through-you.html?full=true http://www.newscientist.com/article/mg21128191.600-specs-tha... and http://theinstitute.ieee.org/people/profiles/rosalind-picard-develops-specs-that-see-through-you http://theinstitute.ieee.org/people/profiles/rosalind-picard...
- karpathy 14y agoI'm not surprised that this is a research topic but compared to all the other applications, I still think this will take much longer to get running reliably. They are working on centered, pre-registered faces and stationary cameras. Once you add low-quality sensor that moves with the head (introducing huge amounts of motion blur to the image), and rich, real-world scenes full of other distractors, the technology will fall apart. I'm certain we'll get hacked up implementations that work half the time if you keep perfectly still for some of the very basic and easy to tell emotions such as happy and sad, but it will mostly be a disfunctional joke for a long time. Time will tell!
- sp332 14y agoFaces will be pre-registered with Facebook (and the Face.com tech they just bought). Anyway if it's that useful, people will learn to keep their heads still while waiting for the computer's assistance :)