Slashdot Mirror
How AI Can Infer Human Emotions (oreilly.com)
An anonymous reader quotes OReilly.com's interview with the CEO of Affectiva, an emotion-measurement technology company that grew out of MIT's Media Lab.
We can mine Twitter, for example, on text sentiment, but that only gets us so far. About 35-40% is conveyed in tone of voice -- how you say something -- and the remaining 50-60% is read through facial expressions and gestures you make. Technology that reads your emotional state, for example by combining facial and voice expressions, represents the emotion AI space. They are the subconscious, natural way we communicate emotion, which is nonverbal and which complements our language... Facial expressions and speech actually deal more with the subconscious, and are more unbiased and unfiltered expressions of emotion...
Rather than encoding specific rules that depict when a person is making a specific expression, we instead focus our attention on building intelligent algorithms that can be trained to recognize expressions. Through our partnerships across the globe, we have amassed an enormous emotional database from people driving cars, watching media content, etc. A portion of the data is then passed on to our labeling team, who are certified in the Facial Action Coding System...we have gathered 5,313,751 face videos, for a total of 38,944 hours of data, representing nearly two billion facial frames analyzed.
They got their start testing advertisements, and now are already working with a third of all Fortune 500 companies. ("We've seen that pet care and baby ads in the U.S. elicit more enjoyment than cereal ads -- which see the most enjoyment in Canada.") One company even combined their technology with Google Glass to help autistic children learn to recognize emotional cues.
Rather than encoding specific rules that depict when a person is making a specific expression, we instead focus our attention on building intelligent algorithms that can be trained to recognize expressions. Through our partnerships across the globe, we have amassed an enormous emotional database from people driving cars, watching media content, etc. A portion of the data is then passed on to our labeling team, who are certified in the Facial Action Coding System...we have gathered 5,313,751 face videos, for a total of 38,944 hours of data, representing nearly two billion facial frames analyzed.
They got their start testing advertisements, and now are already working with a third of all Fortune 500 companies. ("We've seen that pet care and baby ads in the U.S. elicit more enjoyment than cereal ads -- which see the most enjoyment in Canada.") One company even combined their technology with Google Glass to help autistic children learn to recognize emotional cues.
https://www.youtube.com/watch?...
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
Why do I care what an auto parts store's interview about artificial intelligence?
Thats all i have to say
It is one of the greatest features of computers that they work reliably, all the time, with consistent results, regardless of whether I am hitting my keyboard happily or not.
A computer trying to interpret and/or react differently based on how I feel is exactly the opposite of what I would like to have near me.
They literally don't 'mean' anything about anything. The entire premise is flawed. People don't need thermometers for their emotions. So stupid.
If the answer to the question, 'How old are you?', is under 35,Ii could give a toss about your conjecture or your opinion about technology and pragmatic subjects you aren't yet capable of grasping. Get back to us when some time has passed and your understanding of, well, *everything*, has matured. You are not a scientist *yet*, padwan.
I'm thinking back to that time I slammed my fist on my particle-board desk so hard during a Windows NT installation going sideways in heretofore unmapped N-dimensional space that I snapped the little dowels off at one end of my desk plank. There was a CD involved, and floppy disks to load missing drivers for my primary disk systems, and random errors, and circular dependencies in non-coldstart retry attempts.
I don't even know how many levels deep I was at that point in pointless problems begetting more pointless problems (though I'm pretty sure Towers of Hanoi would have awarded me a Hafnium power-up long before then.)
Fuck 1998 all to hell and back. Shortly thereafter I bought a spare P6 on the cheap and spun up OpenBSD 2.3, my first 'nix box, so that I'd always have at least one system around that behaved like it wasn't authored by closet sociopaths. The year ultimately ended on an uptick.
We've all heard of adhesive duck deficiency?
Well the only reason Redmond wasn't reading the public's lips back in 1998 was a pointy rocket launcher deficiency in backyards everywhere.
I went beyond launch conditions on several occasions.
Technology can read our emotional, the modern technology is so credible.Thanks for sharing the aritcle. photo editor
All this work and data sounds impressive until you realise that FACS ("Facial Action Coding System") is bollocks.
Da Blog
They could train their deep learning networks based on research done on brain of dead salmon reacting to human emotions...
http://prefrontal.org/files/po...
... which is nonverbal and which complements our language ...
Translation: Computers need to 'read' the non-verbal message before they can truly understand us.
As a few posts note, an algorithm to make friends will be used to abuse vulnerable people
... help autistic children learn to recognize emotional cues.
Yes, autistism/asperger people miss the more hidden and subtle cues for "I'm horny"/"I dislike you"/"I'm annoyed" which makes smooth social interaction difficult to achieve (and since most sufferers are male, also getting laid). On the plus side, such suffers can see when the words do not match actions; they are not distracted by the emotional message and 'trust me' gestures.
Now they have to understand that emotion is a state indicator and it is used to communicate to ourselves and others the state of satisfaction of our peculiar accent of motivations in the nominal Human Motivation Array. Think of an N - dimensional motivation array with each axis corresponding to one motivation. The motivation vector points out a vector nulling behavior we have developed in our behavior-space over our lifetime. We convey emotions to each other so we can read another's state and adjust our behavior to their state. The more intensely social a species is, the more intensely emotional it most be according to the complexity of its Motivation Array.
Words and phrases and sentences have motivations satisfaction value. Researchers and programmers need to develop a dictionary of words, phrases, and sentence fragments and append motivation satisfaction values to them. In this way a computer would be able to construct meaningful conversation that is very human-like.
What moves the motivation vector? Air temperature, thought-behaviors, other people, what we see, hear, touch etc etc. Environmental inputs, both internal and external. Connect sensors -- both hardware and software -- to the array.
E Proelio Veritas.
In real life, people don't always scowl in anger or pout in sadness. These are stereotypes, not universal signals to be detected. People may smile in anger, cry in happiness, etc. There's tons of variety in emotion. Software that assumes stereotypes are the norm aren't going to work very well.
Here's a great article on so-called statistical recognition of emotion via software: "Pattern Classification Explained." The author is a well-known emotion researcher.
How long will it be before humans can infer AI emotions? https://www.youtube.com/watch?...