Difference between revisions of "Superintelligence"

From Lesswrongwiki
Jump to: navigation, search
(Blog Posts)
m (added links to Arbital)
 
(6 intermediate revisions by 4 users not shown)
Line 1: Line 1:
 
{{wikilink}}
 
{{wikilink}}
'''Superintelligence''' refers to a hypothesized being that possess intelligence far beyond the abilities of any current human being. These could either be an Super Artificial Intelligence (SAI) or a human that had some form of cognitive enhancement either biological or [[Brain-computer interfaces]]
+
{{arbitallink|https://arbital.com/p/superintelligent/|Superintelligent}}
 +
A '''Superintelligence''' is a being with superhuman intelligence, and a focus of the [[Machine Intelligence Research Institute]]'s research. Specifically, Nick Bostrom (1997) defined it as
  
Superintelligence can be divided into two varieties “weak” and “strong”. A “weak superintelligence” would have a structure with similar cognitive abilities to the human brain, but able to operate at a vastly greater speed. Whilst a “strong superintelligence” would operate superiorly in both speed and cognitive ability.
+
<blockquote> "An intellect that is much smarter than the best human brains in practically every field, including scientific creativity, general wisdom and social skills."</blockquote>
 +
 
 +
The [[Machine Intelligence Research Institute]] is dedicated to ensuring humanity's safety and prosperity by preparing for the development of an [[Artificial General Intelligence]] with superintelligence.  Given its intelligence, it is likely to be [[AI boxing|incapable of being controlled]] by humanity. It is important to prepare early for the development of [[friendly artificial intelligence]], as there may be an [[AI arms race]]. A strong superintelligence is a term describing a superintelligence which is not designed with the same architecture as the human brain.
 +
 
 +
An [[Artificial General Intelligence]] will have a number of advantages aiding it in becoming a superintelligence. It can improve the hardware it runs on and obtain better hardware. It will be capable of directly editing its own code. Depending on how easy its code is to modify, it might carry out software improvements that [[recursive self-improvement|spark further improvements]]. Where a task can be accomplished in a repetitive way, a module preforming the task far more efficiently might be developed. Its motivations and preferences can be edited to be more consistent with each other. It will have an indefinite life span, be capable of reproducing, and transfer knowledge, skills, and code among its copies as well as cooperating and communicating with them better than humans do with each other.
 +
 
 +
The development of superintelligence from humans is another possibility, sometimes termed a weak superintelligence. It may come in the form of [[whole brain emulation]], where a human brain is scanned and simulated on a computer. Many of the advantages a AGI has in developing superintelligence apply here as well. The development of [[Brain-computer interfaces]] may also lead to the creation of superintelligence. Biological enhancements such as genetic engineering and the use of nootropics could lead to superintelligence as well.  
  
 
==Blog Posts==
 
==Blog Posts==
  
*[[http://www.acceleratingfuture.com/articles/superintelligencehowsoon.htm Superintelligence]] by Michael Anissimov
+
*[http://www.acceleratingfuture.com/articles/superintelligencehowsoon.htm Superintelligence] by Michael Anissimov
  
 
==External Links==
 
==External Links==
Line 12: Line 19:
 
*[http://www.nickbostrom.com/superintelligence.html How long before Superintelligence?] by Nick Bostrom
 
*[http://www.nickbostrom.com/superintelligence.html How long before Superintelligence?] by Nick Bostrom
 
*[http://profhugodegaris.files.wordpress.com/2011/04/nocyborgsbghugo.pdf A discussion between Hugo de Garis and Ben Goertzel on superintelligence]
 
*[http://profhugodegaris.files.wordpress.com/2011/04/nocyborgsbghugo.pdf A discussion between Hugo de Garis and Ben Goertzel on superintelligence]
 +
*[http://www.xuenay.net/Papers/DigitalAdvantages.pdf Advantages of Artificial Intelligences, Uploads, And Digital Minds] by Kaj Sotala
  
 
==See Also==
 
==See Also==

Latest revision as of 06:03, 14 October 2016

Smallwikipedialogo.png
Wikipedia has an article about
Arbital has an article about

A Superintelligence is a being with superhuman intelligence, and a focus of the Machine Intelligence Research Institute's research. Specifically, Nick Bostrom (1997) defined it as

"An intellect that is much smarter than the best human brains in practically every field, including scientific creativity, general wisdom and social skills."

The Machine Intelligence Research Institute is dedicated to ensuring humanity's safety and prosperity by preparing for the development of an Artificial General Intelligence with superintelligence. Given its intelligence, it is likely to be incapable of being controlled by humanity. It is important to prepare early for the development of friendly artificial intelligence, as there may be an AI arms race. A strong superintelligence is a term describing a superintelligence which is not designed with the same architecture as the human brain.

An Artificial General Intelligence will have a number of advantages aiding it in becoming a superintelligence. It can improve the hardware it runs on and obtain better hardware. It will be capable of directly editing its own code. Depending on how easy its code is to modify, it might carry out software improvements that spark further improvements. Where a task can be accomplished in a repetitive way, a module preforming the task far more efficiently might be developed. Its motivations and preferences can be edited to be more consistent with each other. It will have an indefinite life span, be capable of reproducing, and transfer knowledge, skills, and code among its copies as well as cooperating and communicating with them better than humans do with each other.

The development of superintelligence from humans is another possibility, sometimes termed a weak superintelligence. It may come in the form of whole brain emulation, where a human brain is scanned and simulated on a computer. Many of the advantages a AGI has in developing superintelligence apply here as well. The development of Brain-computer interfaces may also lead to the creation of superintelligence. Biological enhancements such as genetic engineering and the use of nootropics could lead to superintelligence as well.

Blog Posts

External Links

See Also