Take a theory of consciousness that calculates how aware any information-processing network is – be it a computer or a brain. Trouble is, it takes a supercomputer billions of years to verify its predictions. Add a maverick cosmologist, and what do you get? A way to make the theory useful within our lifetime.
采用一种意识理论,以此来计算人类是如何意识无所不在的信息处理网络的,这是一件非常麻烦的事情。不论是用一台计算机还是一个大脑来计算,这样的计算需要一台超级计算机用几十亿年来验证它的预言。还得要加上一位与众不同的宇宙学家,你能得到什么?解决的方式是找到一种我们在有生之年可以使用的理论。
IIT) is one of our best descriptions of consciousness. Developed by neuroscientist Giulio Tononi of the University of Wisconsin at Madison, it’s based on the observation that each moment of awareness is unified. When you contemplate a bunch of flowers, say, it’s impossible to be conscious of the flower’s colour independently of its fragrance because the brain has integrated the sensory data. Tononi argues that for a system to be conscious, it must integrate information in such a way that the whole contains more information than the sum of its parts.
Integrated information theory (整合信息理论(Integrated information theory ,IIT)是我们描述意识的最佳方法之一。该方法由麦迪逊威斯康辛大学神经科学家朱利奥·托诺尼(Giulio Tononi)发展而成,这一理论基于这样的观察,每一刻的知晓都是统整的。当你凝视一束花的时候,不可能独立于花的芳香而单独意识到花的颜色,因为大脑已经对这些感觉数据进行了集成。托诺尼认为,信息整合必须以这样的方式完成:系统的整体的信息必须大于各部分之和。
The measure of how a system integrates information is called phi. One way of calculating phi involves dividing a system into two and calculating how dependent each part is on the other. One cut would be the “cruellest”, creating two parts that are the least dependent on each other. If the parts of the cruellest cut are completely independent, then phi is zero, and the system is not conscious. The greater their dependency, the greater the value of phi and the greater the degree of consciousness of the system.
衡量系统整合信息方式的方法称为φ(phi)。计算φ(phi)值的一种方式是将一个系统分成两个部分,并且分别计算每个部分依赖另一个部分的程度。一种“最残酷的”切割创造了彼此最少依赖的两个部分。如果这个最残酷切割的两个部分是彻底的独立的,那么φ(phi)为零,并且该系统是非意识的。两个部分之间的依赖性越大,φ(phi)的值就越大,并且这个系统的意识程度就越大。
The cruellest cut 最残酷的切割
Finding the cruellest cut, however, is almost impossible for any large network. For the human brain, with its 100 billion neurons, calculating phi like this would take “longer than the age of our universe”, says Max Tegmark, a cosmologist at the Massachusetts Institute of Technology.
神经元,计算如此之大的φ(phi)值所耗费的时间将会“超过我们宇宙的年龄”。
然而,任何的大网络系统要发现这个最残酷的切割几乎都是不可能的。麻省理工学院的一位宇宙学家特德马克(Tegmark)说,人类大脑有1000亿个Tegmark has come up with a fast way of approximating phi. He treats each neuron in the network as a node and their interconnections as links. He assigns a thickness to each link, proportional to the strength of the interconnection. Now, imagine turning a knob so that the thinnest links fade out. The picture will look somewhat different. Try again, fading out the next-thinnest links. Continue until the single interconnected web breaks into two. Tegmark has shown that this configuration approximates the cruellest cut.
特德马克已经找到了一条计算φ(phi)近似值的捷径。他把网络中的每个神经元看作是一个节点,并把它们之间相互联系看作为一种链接。他为每个链接分配的厚薄程度与连接的强度成正比。现在,想象转动一个旋钮,以便最薄的链接逐渐消失。这样的画面看起来就会有些不一样了。再试一次,下一个最薄的链接消失掉。继续进行,直到单一链接的互联网断裂两个部分为止。特德马克已经向我们呈现了这种接近最残酷切割的格局。
Just a second 刚好一秒
Crucially, this dramatically cuts down the time it takes to find phi. “It goes from being above the age of the universe to something quite manageable,” says Tegmark – he estimates it would take less than a second for a human brain.
最关键的是,这大大减少了发现φ(phi)值的时间。“它从超过宇宙的年龄走到了相当容易管理的地步。”特德马克说。他估计,计算人类大脑φ值用不了一秒的时间。
“It’s cool stuff,” says Christof Koch of the Allen Institute for Brain Science in Seattle. “It’s essential if we are ever to measure phi for real systems and not just toy models with 10 or 20 binary neurons.”