What is the Definition of an 'AI Weapon'
July 30, 2015
Link: https://www.linkedin.com/pulse/definition-ai-weapon-timothy-holborn/
I've been considering the recent global news regarding the dangers of 'AI Weapons', I was looking for some feedback on what would be defined as an 'AI Weapon'...
Basic Research says, common terms talk of 'arms race'[2], 'weapons'[3], yet perhaps distinctions reasonably relate to any form of 'warfare'[4], whether that be 'economic warfare'[5], perhaps 'information warfare'[6] or perhaps more explicitly, conventional warfare or other acts of war [7] generally....
Yet, from a laypersons point of view, a weapon seems more like something that can be used to harm or control another by way of act or threats of acting as to utilise the 'weapon'...
Factors of self-defence are also important to consider. In-turn, at least from a philosophical point of view, far more complex issues in-turn come to mind, such as the statistics on police shootings in different regions
So, at the risk of reading far too deeply into the work that has been promoted via various news networks, World-Wide; I've posted my first thing on linked-in, seeking the input of others who consider the merits of the aforementioned open-letter in the broader sense, and therein seek your views.
Within the field of work for which i am involved, some of the Questions I ponder, relates to both storage and in-turn how systems enable the use of agents[8], pending their distinction (what sort of 'agent' they are).
- How non-human agents are made capable?
- How should they be shut down if necessary?
and perhaps implicitly, pursuant to some of the web-science [9] concepts within the field of linked-data,
- How might these sorts of concepts apply to new industrious concepts such as autonomous organisations [10].
- How do these concepts apply to 'choice of law' with relation to online services and data-storage?
Choices are made between methodological manifests that result in some form of product, whether they centralise control methods, corruptible or believed to be otherwise, or whether they decentralise and/or depend upon human intervention.
A lay concept may look something like defining mean that enable distinctions to be made between 'trust' as it is applied by; either,
- Accountable people / groups of people (whilst maintaining a decentralised method in the groups / people / interdependencies, similar to global systems of democracy, et.al.); OR,
- 'Trust', as it is applied by a computational / Hardware + Software Communications platform that supports the manifest outcome of trust computationally, subject to the variabilities of the underlying assumptions made; OR,
- A mixture.
When considering computational trust, I think of bitcoin and it's 51% vulnerability, which in this thread [11] is described in relation to $ rather than accountability methodologies. Bitcoin could technically be used as a means to support e-contracts and other formats of trade. When considering the IOU nature of $ and transactional demands of systems, should they effectively be capable of high-frequency transactions as required by markets extend the considerations, in-turn may consider that it is the h/w perhaps that is most important ?
So, the chain could issue an academic credential based on a bitcoin block / chain / notation; yet, i'm not sure how that would change if it was subject to change in future (ie: someone cheated)...
So, whilst i've prompted the question with a few considerations; the underlying question is about the views of those on the list, about what is an 'AI Weapon'...
Given so few in the world understand linked-data [12], i figured it was a reasonable question to ask the various groups.
What would protect and/or uphold the concept, of preventing the creation of 'AI Weapons'. What is the definition of an 'AI Weapon'.
If others were going to add-value to the sentiment made about AI Weapons [1] what would you say...
Underlying it all, is the reality of living in a world with such advanced technologies, yet comprehending the difficulties that so many humans experience regardless of what is technologically available, as would protect persons from the same issue, should they be of different circumstance, the implications of improving public data is one very simple example of how things might improve for the better, once better data is available for research, and consideration by our systems of law.
If we are to find a means in which to listen to those who have signed this document, what is it that we'll need to consider?
First thing is perhaps, defining what an AI Weapon actually is...
So what are your views?
LINKS
[1][http://futureoflife.org/AI/open_letter_autonomous_weapons](http://futureoflife.org/AI/open_letter_autonomous_weapons)
[2][https://en.wikipedia.org/wiki/Arms_race](https://en.wikipedia.org/wiki/Arms_race)
[3][https://en.wikipedia.org/wiki/Weapon](https://en.wikipedia.org/wiki/Weapon)
[4][http://dictionary.reference.com/browse/warfare](http://dictionary.reference.com/browse/warfare)
[5][https://en.wikipedia.org/wiki/Economic_warfare](https://en.wikipedia.org/wiki/Economic_warfare)
[6][https://en.wikipedia.org/wiki/Information_warfare](https://en.wikipedia.org/wiki/Information_warfare)
[7][https://en.wikipedia.org/wiki/War](https://en.wikipedia.org/wiki/War)
[8][http://xmlns.com/foaf/spec/#term_Agent](http://xmlns.com/foaf/spec/#term_Agent)
[9][http://www.w3.org/2007/09/map/main.jpg](http://www.w3.org/2007/09/map/main.jpg)
[10][https://en.wikipedia.org/wiki/Decentralized_autonomous_organization](https://en.wikipedia.org/wiki/Decentralized_autonomous_organization)
[11][http://bitcoin.stackexchange.com/questions/658/what-can-an-attacker-with-51-of-hash-power-do](http://bitcoin.stackexchange.com/questions/658/what-can-an-attacker-with-51-of-hash-power-do)
[12][https://en.wikipedia.org/wiki/Linked_data](https://en.wikipedia.org/wiki/Linked_data)