Beats me why people are designing a computer program to do stuff that a super-recognizer could do standing on their head

Why, why, why do people assume that we need computers to do clever and sophisticated things in face recognition and face perception? Our brains have evolved over millions of years to do this stuff, and some people are even better at this stuff than the amazing feats of visual perception that the average Joe can do with barely a thought or effort.

http://www.newscientist.com/article/dn25776-computer-spots-rare-diseases-in-family-photos.html#.U6mtSvmSx8E

If Abraham Lincoln had Marfan syndrome, identifiable through his facial appearance, then I guess that means that the Australian politician who was identified as a Lincoln double on the TV show Insiders must have it too.

If I were two-faced, would I be wearing this one? Abraham Lincoln

How many other politicians can you think of who could recycle that witty remark?

Jokes aside, its worth taking a look at the larger image of the group of faces in the magazine article, because they change in ways that I find quite fascinating and familiar. Which one do you think looks the most like Alfred E. Neuman? I think the right-edge lower row.

Advertisements
Post a comment or leave a trackback: Trackback URL.

Comments

  • Anonymous  On June 29, 2014 at 1:46 pm

    Because super-recognizers have to be located, tested, hired, and paid, and even then can only process so much footage per hour; computers, on the other hand, work for the cost of electricity, and a large organization with more footage than computers will have very little difficulty buying a couple dozen more.

  • C. Wright  On June 29, 2014 at 8:42 pm

    I don’t work in security, but it is my understanding that at least some CCTV cameras in public places, and I guess private ones too, have people monitoring them 24/7. I guess this is what they get paid to do; just sit and watch people and places for hours on end. How would that be any less labour-intensive than any medical application of facial dysmorphia recognition by a human?

    Medical genetics is a specialty. It is actually surprising how few doctors and clinics there are for particular medical specialties, even in a large city. There is no way that any of these clinics churn through scores or hundreds or thousands of patients in a day or a month. Handfuls of new patients is more likely per day or week per clinic, and the doctors seem to spend half their life OS at conferences. I’m sure a month’s patient-load at a specialist medical clinic wouldn’t overwhelm a human trained to look for patterns of diseases in faces. I guess if this program is to be used by GPs instead of patients of a clinical geneticist there would be more images to process, and if used in the third world and world-wide, that would be a big market, but I’m yet to be convinced that the program would be doing very much that a GP should be able to do on her own. What kind of doctor couldn’t pick progeria or Downs syndrome? They are pretty hard to miss!

    A computer will not have background human knowledge from life experiences. A computer will not be able to say that “In my experience people who look like this have an exceptionally agreeable personality and heaps of friends, but generally low grades in school” or “Check the dental history of this patient for a distinctive abnormality that is often corrected by braces in childhood and won’t always show up in adult photos”. Computers aren’t people and don’t have a lifetime of discovering all kinds of things about other human beings. Oh, and the agreeable personality referred to can be objectively tested for in minutes with well-established psychological tests.

    Locating super-recognizers to hire is no big deal, because for any large organization it has already been done. We are estimated to be as common as 1 in 50, and unless a company or govt department has actively discriminated against hiring supers, it will already have plenty on their payroll. Internal recruiting and testing is all that needs to be done, and testing is so easy with computerised tests. Researchers manage to recruit supers for their studies and pay us nothing or near-nothing for our help. I can’t imagine giving a job to a super would be an impossible task.

  • C. Wright  On June 29, 2014 at 9:15 pm

    Another thought; would a computer program be able to observe and consider the pitch of a patient’s voice, they way they walk and talk, the colour of their complexion, their personality? Humans have all these bonus features in perception in addition to face perception. It might sound as though observing a person as a whole instead of looking at only a colourless image of their face would be straying from objectivity, but it is a fact that all the above things can be symptoms of diseases or syndromes.

  • C. Wright  On June 29, 2014 at 9:28 pm

    I do wonder how the problem of race is handled by these kinds of face image databases. The cross-race effect would have to complicate matters. Are there images of faces from non-Anglophile stock available, representing the many types of medical dysmorphia, for doctors to use to help identify cases of other races? Very dark skin reflects light in a very different way than buff skin, and different races have normally different face shapes.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: