„Sir, the following paradigm shifts occurred while you were out“, announces the secretary to its returning boss in a the now classic cartoon related to the Singularity, the moment when even the most intelligent human will be surpassed by a machine in all ways.

If we are to trigger this event, then our own brain is a good starting point. Just like that of humans, the artificial brain will probably have to learn everything about reality and solve the problem of interacting with the environment. In contrast with us, however, it will have to find other ways to get experienced, since it will probably be difficult for it to pass through a genuine childhood.

Arguably, the artificial intelligence could learn everything just like a newborn, perhaps through special training. But will that be of much value? Would it result in something more than an average human? In my opinion, it would be necessary to supplement the processing power and hardware resources that support the artificial brain with sensory experience required to write, say, a poem about a tree. This way, it would be something along the lines of a supercomputer with a character, possessing two types of knowledge: the first one, common knowledge would include information about every subject, and perhaps every existing taxonomy – and the second one would be subjective knowledge, allowing for fuzzy connections and associations which can only be created through personal experience.

But what is a brain, really? Is it just the sum of all its neurons, a collection of automaton cells working together in a predictable way, or is it something more? To attempt to answer this, we could start by looking at how the brain is formed. It may be safe to consider it pretty much empty at birth, with not much in it – other than perhaps some instincts.

In becoming sentient, however, the brain is dependent on its surroundings. It is quite literally their reflection, since it considers true whatever it observes and predicts about the environment. If you look at a tree, for example, you recognize it's a tree but would a put-together simulation of neurons understand that? The simulation didn't go to kindergarten, it doesn't know stories with trees, it hasn't seen movies with trees, it hasn't seen a tree before. In order for it to be effective, the association between the word „tree“ and the experience of perceiving it through all the five senses would have to become possible.

In a story I read (but unfortunately no longer find on the Web to credit the original author), after the extinction of Man, two computers remain, created with the purpose of serving mankind and solving problems such as building cities. A newly created machine, however, is curious about the nature of humans. It studies all the artifacts it finds and gathers all the possible information about them, and after doing so it reaches the conclusion that the way to become human is to transfer itself inside a human body (which it does, then takes over all the other machines).

Ultimately what we would like to possess is an artifact having intelligence and processing power superior to our own, combined with a relentless occupation of this virtual mind on how to solve our problems and make new discoveries, and to elaborate new scientific theories. Of course, we would ask it far more significant questions than „Answer to the Ultimate Question of Life, the Universe, and Everything“, to which everyone knows the answer is 42, as Douglas Adams tells us. In parallel to the Deep Thought supercomputer imagined by the author, our Singularity intelligence will have to allocate resources (energy, time, materials) to find answers to our queries about reality.

POST#0075 2009-MAY-8

Help improve the Fusion Blog - express your opinion about the content on this page:   I like it   Can be improved

Commentary:

No comments so far. Why not add one?