The Brain Google

Written by Carnegie mardi, 08 décembre 2009

Action Potential or Pagerank Algorithm? Learning or Duplicate content management? Is Google, the most popular search engine, that successful because of the similarities that its working reveals, with our own brain? Has this analogy something reliable? Well, it seems that we can find in the program, algorithms reflecting some cerebral functions that optimize our own performances.... What if Google was the brain of the humanity?

Does it sounds strange to you? A few years ago, Joshua Knobe and Jesse Prinz⁽¹⁾ showed that we commonly discern, in entities like God, Microsoft, (and so on Google), intentionality, and some others capacities that we generally attribute to human. We naturally personnalized these entities as if they were relatively close to be human person. So, sentences like "Google plans to create a new service" sounds as plausible to us as "B. Obama plans to create a new administrative section". Among this capacities, we commonly attribute to such entities, willingness, planification, believes, decision-making or knowledge. However, immediate experiences as joy or anger, don't seem to apply to God or business corporations, according to this Scientific American's article⁽²⁾.

Nevertheless, we feel like, face to these entities, that we can attribute them some characteristics that we generally considered as human's ones, or at least, as characteristics of living organisms. With the 5 capacities cited above, we're not as far as we should believe from the consciouness capacity (planification, cognition, motivation...).

Especially concerning Google, the "aim" of this entity looks more similar to the human thought : google's aim is to (automatically) find, like we would want to find (well, with a very long searching time). What the user wants to find, Google must, sometimes guess it, and in the best case, find it. In this way, its aim is : replacing our own brain while searching through information on the web. So it's not as strange as it first looked, that we can see in Google some "brain capacities", simulated and made in order to do what the real brain usually do.

Neuron Networks and Web Pages

Some of the most known facts about the Search Engine Google, is the use of links when creating the index and attributing pagerank. The web crawler analyses pages and follow links on it, like an electric message going through the neurons, reaching others via chemical synapses. Each link is considered as a vote for the page linked. Importance attributed to a Page A is higher when the link comes from a high pageranked page B. So, original pagerank B constitutes a ponderation for vote from B to A. That's kindly similar to neurons : each neuron intersection (synapse) can have more importance than its neighbors. Some will not be sufficiant to get fired the post-neuron, some other will. Post neuron has higher probability of activation if there's a lot of activating pre-neuron (instead of just a few), and higher probability too if the activating pre-neurons are better connected to him (i-e, if there are many synapses connecting them, or if the synapses are more efficient - due to learning by precedent firings). These similarities between the brain machinery and the google's programms are not the only ones !

Informatic - Google	Biology - Brain
Pages network	Neurons network
Links	Synapses
Google exploration bot (web crawler). It goes through pages and follow links they contains.	Nerve influx. Electro-chemical influx go through the neurons and follow synapses they contains.
How googles learn? Google Programs attribute pagerank to Page A, regarding its importance, i-e thenumber of links and their pagerank, which the bot has passed through to join page A.	How network learn? Networks "attribute ponderations" to connexions, regarding the number, activation frequency and power of these connexions.
The more there are links to a Page A, the higher is Pagerank A	The more there are active simultaneous connexions exciting the neuron, the higher is probability of activation
The higher is pagerank with a link to page A, the higher is pagerank A	The better neuron B is connected to neuron A (synaptic efficiency or high number of synaptic connexions between the two neurons), the higher is the probability of activation
Duplicate Content : this parameter increase learning. If two pages have the same content, baclinks don't add, these are two different pages for Google. So pageranking is inefficient, better have only one page that concentrates all backlinks. This duplicate content parameter allows google to "prefer" original pages and separate original content from duplicate, in order to propose the best searching results.	Learning network. When you try for the first time an activity, that's kindly hard, your mind have to be concentrated, and distractors can interfer easily. However, as you repeat this activity, the most efficient neurons ways are "preferred" and some treatments become unconscious, you're becoming an expert. What with your neurons? They're like a wild forest at the beginning of the learning : there's not a prefered way but several to a goal. As you walk through a specific way several times, to reach this goal more rapidly and efficiently, your foots crush grass and begin to form a real road, and other ways are neglected. Same for neurons. At the beginning, there are several neuronal roads, then with learning, the more efficient road is prefered, and the activity becomes progressively automatic or routine.
Semantic. Link anchors are very important for Google to define subject of the page linked. Link as "click here" doesn't mean anything to Google. Same link noted "discover this article from Stephane, dealing with similarities between mechanical Google and the biological brain" allows google to define tags that give a signification, semantic attributes, to the page linked. There are local and global semantic contexts defined by several techniques : key words of the sites gives global contexte (like a specific part of a semantic graph), keywords of the page gives local context (like a node in this graph), links are like semantic arcs between nodes. Note that if you can't really control links from outer sites, you can define and controls your own links, in order to help google in attributing semantic to your pages.	According to some memory theories, connexions define semantics and contextual fields. In that way, thinking of a Fireman will pre-active some concepts as "red", "truck", "hot"... the more the context is define and important, the better the learning is. Power of the context is high.For example, you may not recognize a student if you see him on hollidays, cause you are used to see him only in the university. Some experimentations have shown that informations are better remembered if you have the same conditions (like color of the ink, noise, emotionnal state) that were when you learn. Some researchers believe these properties are effective at the neuron's level

A summary of some Google's "cerebral capacities"

So, conclusion?

Google is created to be a simulation, the more efficient as possible, of the search and decision capacities of the user. So, if you want to create a correct site, you can try to considered search engine as a replica of human brain. When you want to train your brain, you work and learn. In that way, you increase its activity by receiving informations from outside and reasoning on information from inside. Google do the same. It's not sufficient that backlinks points to your site, that's reasonable to improve your site's structure in order to give meanings to your pages and links, in order to make your pages accessibles (meaningful links between your pages of your site). Don't use the "click here" or "this page" links, prefer meaningful ones like "alzheimer disease" or "Neurological characteristics of the cortex".

Keep in mind that semantic information of a page (i-e the meaning that google associates to your page) is closely related to 4 parameters :

Url : http://www.exemple.com/?page=3&categorie=45 is less meaningful than http://www.psyblogs.net/fr/internet-psychology/the-brain-google.html
Metas :principally keywords, define your content and help google to associates search words to your content. Description of your page too, and in addition, it appears on the search results, so that often the first thing users see when searching.
The content ! it is analysed by Google, so better be your content in adequation with your keywords and description, as well as when you search through your memory a recollection of you when you were 8 years old, you start to search through your child recollections, not through your recent recollections. Your content should contains your keywords, so they will be seen by google as a resume, indexation of your page.
Links and anchors. In order to "judge" what a site "think" about yours, google can only base its statement on the url, the "alt and title" attributes, the anchors (words displayed), and the thematics of the page. So, in order to help google, others sites and yours, define good and meaningfull links.

Pour en savoir plus : Knobe J., Prinz J. (2006). Intuitions about consciouness. Experimental studies.

Sciam : Can a robot, an insect or God be aware? (2008)

Bookmark

Email This

Comments (0) Add Comment

Show/Hide comments

Write comment

Show/Hide comment form