Think different

I meet very few people who really surprise me. People who will say something I really do not expect. But also, people who look as I have never seen someone like them. This world is a constant repetition of itself. There are at maximum one hundred, maybe two hundred really different type of people. Identified them, you have seen them all.

Face recognition

There is scientific evidence, for this observation. Most face recognition softwares (the same used by Facebook or iPhoto to suggest the friends to tag, but also by airports security to monitor single passengers in transit), most of these codes take advantage of principal component analysis (PCA) and single value decomposition (SVD) methods. A nice way to say it, is that each face is decomposed on an ordered basis of orthogonal eigenvectors: experimental evidence show that eigenfaces with larger expansion coefficients are common to all 7 billions humans. In a simpler way, we may just say that all the process is based on the idea that mixing something like 25 basic faces, in different proportions, allows to reconstruct any possible human look.

You read my mind

I am not sure if someone has ever tried a similar method to describe what people usually say: conversations and dialogues. I have my own PCA software installed somewhere in my epithalamus. I decompose what a person tells me in the first minute of our conversation. Then I predict what s/he is going to say next. If I fail, maybe this interlocutor has something to tell me. Unfortunately, most of the times, I guess correctly. Someone may say “you read in my mind”, if I have a good reason to be polite, I will make them believe so.

Truth is, most people have nothing new to say.

Surprisingly unique

Surprisingly unique. American model Devon Aoki has possibly one of the most peculiar looks among human being. She is beautiful, in her way. Nothing I have ever seen before. For once.

11 responses to “Think different

  1. I strongly agree with the author. If you are a born psychologist then you won’t have problems to “read” the thoughts of others. I can only add that you can “understand” some person so well that you predict not only his thoughts but also his actions, so to say his “game”. As a result you get a feeling that somebody does not exist in reality but is created by your own imagination.

    • The “game” idea is interesting, and I will maybe develop it in some future post. I am not sure about the fact that you may actually believe that reality is just imagination: any meaningful reasoning should originate from the postulate that an external reality exists. Interesting or not, still real. 🙂

  2. I’m a fan of your posts in general, they are always original and often aesthetic. This however, is a nothing more than a mistake. Mathematically partially incorrect, and a little arrogant. Of course, everything in nature can be described by a linear combination of a set of (optimally) orthogonal functions. The beauty is not contained by the functions themselves, but by their (sometimes) unique combination.

    • Thank you! I have never denied my dose or arrogance. And I enjoy when a post sparks debate. However, I am a bit curios about the math: what is incorrect? This could be interesting to discuss! 🙂

      • Incorrect might be overstated. I have my dose of dramaqueen ;-). I can’t see the extrapolation from decomposition of the faces into eigenmodes, towards predictive capabilities for conversations. It is not possible to predict faces using PCA, so I don’t see how you can predict conversations using the same method.

        • Saying first of all that it was a provocative statement (the prediction of conversations), I may defend it from a mathematical point of view. In principle, once I have an ordered orthogonal basis, and I have identified a finite set of most relevant eigenstates, I may build a finite database of possible linear combinations, assuming a reasonable interval for the expansion coefficient. I will have then a set of predictable conversations, that is not a complete set, but a reasonable one given the border conditions. Now, if a person will tell me something that is matching at 95% (just to set a level of confidence) one of the vectors in my database, it means that I expected it. On the contrary, if a person is able to build — with the same basis — an unexpected, interesting linear combination of words, then I will be more than happy to recognize that that person has something original to say.

          In other words: there are seven notes, do, re, mi, fa, sol, la, si. (In this case, they are a finite set, but the idea is the same). Still not everyone is able to produce interesting and original musics. Often, people just sing the same song they heard on the radio. Few individuals may produce new symphonies: these are the only interesting interlocutors for my human dialogues.

      • With this version I can live. Although I’m not sure that the expansion coefficients are discrete, and therefore their combination can yield an infinite number of combinations. However, since one would only be interested in approximately predicting any conversation it can suffice to have a sufficiently accurate value for the expansion coefficients. It is an interesting, but scary idea. On the bright side, the expected dominant eigenmodes of any facebook post do not include dominant eigenmode as content, which makes you unpredictable, and arguably interesting… 😉

  3. Pingback: I love portraits « The Incredible Tide·

  4. Pingback: Intelligence is sexy | The Incredible Tide·

Leave a Reply to SV Cancel reply

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

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

Google photo

You are commenting using your Google 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 )

Connecting to %s