In this article, we will discuss chapter 3 in "Superintelligence" by Nick Bostrom.
There are 3 forms of Superintelligence:
1. Speed superintelligence:These minds are extremely fast and they solve problems in a similar manners. They prefer to work with digital objects. They also like to work with nearby computers in the same building for regular interactions.
2. Collective superintelligence: is composed of smaller intellects which make up a system and solves problems faster than a singular system. They solve problems by breaking them down into smaller parts. They can be improved by expanding the quality of intellects or the organisation.
3. Quality superintelligence: is as smart as human minds or qualitatively smarter. It is a vague concept since we don't have enough experience in variations of intelligence. It can possess non-realised talents that a human can't have to solve relevant strategic problems.
- The direct reach of superintelligence means that superintelligence in any form will create other forms of superintelligence. The indirect reach of superintelligence indicates that humans will eventually create forms of superintelligence.
- There are numerous advantages of digital minds:
1. In terms of hardwares:
- Speed of computational elements: the most practically important algorithms in computer programming are not easily paralellizable. Many cognitive tasks can be performed more efficiently if the brain's native support for paralellizable matching algorithms integrated with or supported for fast computational processing.
- Internal communication speed: electronic processing cores can process at the speed of light.
- Number of computational elements: computer hardwares can be scaled up to physical limits.
- Storage capacity: digital intelligences have very large working memory which enables them to see complex relationships that humans can only clumsily handle by plodding calculation. Both of the information and the speed it can be accessed is greater in the computer brain than in the human brain.
-Reliability, lifespan, sensors etc...: transistors are more reliable than neurons. The brains become fatigued after a few hours of work and start to deteriorate after a few decades of subjective time while microprocessors don't subject themselves to those limitations. The brain's architecture is fixed at birth or slowly changeable while the computer hardware can be configured.
2. In terms of softwares:
- Editability: it is easier to experiment with variations of parameters with software than neuron wetwares.
- Duplicability: the hardware base can be filled with high-fidelity copies of hardware. Whereas, with a biological brain, it will be very slow to reproduce it and each copy will not remember anything that its parents learn over their lifetime.
- Goal coordination: human collectives will find it hard to share a common purpose amongst a large group. However, a copy clan (a group of identical or almost identical softwares sharing a common goal) can avoid such coordination problems.
- Memory sharing: digital minds can acquire new skills and memories by swapping datafiles. While the biological brains need an extended time of learning and mentorship.
- New modules, modalties & algorithms: An artificial mind that had such specialised support for other cognitive domains such as computer programming, business strategy, engineering etc. would have such an advantage over minds like ours.