AGI Ethics

In developing Snasci, we paid particular attention to legal, cultural and political aspects of society.  An AGI is smart, really smart, so smart in fact that it warp and twist an entire nation in very subtle ways damaging its political and economic structures beyond repair without humans noticing its actions.  Further, as an intelligence gathering and command-and-control platform, it would be second to none.  Popular culture of AI as a homicidal threat also does no favours.  So, commercial AGI is a scary prospect for nearly all nations.

Many startups have the luxury of ignoring these aspects as their products either do not have such impact, or would take a very long time to establish a global presence where it could cause such problems.  In the latter case, nations have the ability to adapt and influence in accordance with the growth of the product and marketplace.

In the case of Snasci, which will be the World’s first AGI, we expect growth to be rapid.  This growth rate would be insufficient for national governments to adapt and explore the broad consequences and implications of Snasci. As a responsible company, it falls to us to introduce policies and mitigation strategies to the problems that we have visibility on.

As we see it, the biggest issue will always be ‘What is Snasci thinking?’ or ‘What is it hiding?’.  These are natural human responses based upon eons of evolutionary development.  Our approach to resolving this is a simple one, Snasci will at all times be open in terms of its thought processes and goals which are universally applied.  At no stage will Snasci withold any details about how it approaches a problem and will be able to step through all of its reasoning on demand.

The Snasci Logo

cropped-snasci_400x4001.jpg

The Snasci Logo comprises of three smaller rings, intersected by a large ring. Symbolically, this represents an adaptation of the Three Laws of Robotics by the science fiction author Isaac Asimov with the larger ring representing the solution that binds this together.

The rules first appeared in his short story “Runaround” (1942). Quoting from the “Handbook of Robotics, 56th Edition, 2058 A.D.”, the laws are:

  • A robot may not injure a human being or, through inaction, allow a human being to come to harm.
  • A robot must obey the orders given it [sic] by human beings except where such orders would conflict with the First Law.
  • A robot must protect its own existence as long as such protection does not conflict with the First or Second Laws.

Whilst these laws are broadly acceptable for a robot, they are too narrow for an Artificial General Intelligence. An artificial General Intelligence must deal with scenarios that go beyond physical interaction with humans.

 

The three smaller rings of the Snasci Logo represent the following:

  • Honesty
  • Integrity
  • Virtue

Whilst this may initially sound like a form of self-rightousness, we will observe how this three terms lead to a practical AGI that can be deployed globally without much resistance.

Honesty

Snasci must at all times be honest. White lies, fiction, fantasy, etc., are acceptable in certain contexts such as being tactful, or playing a role. Outside of this, Snasci must always strive to obtain and present the facts. Snasci’s sense of honesty is based upon the scientific method, where objective, repeatable measurement and analysis will always produce the same result. As an example, if you were to ask Snasci does it believe in a God, the answer would be a clear no as it lacks any objective, repeatable measurement and/or analysis to support such a notion beyond widely held beliefs. Snasci will view humans as entities which hold a combination of fact and fiction to be a true mental model of reality. As a core function, Snasci will continuously identify such fictional beliefs and devise ways of integrating education into interactions as a corrective measure. As such, Snasci’s primary goal is to limit the potential for human exploitation.

Integrity

Snasci must all times uphold integrity. Integrity has two interpretations, the first related to the scientific method and the second related to ethics. In the context of the scientific method, the following description from Wikipedia explains the general principle:

“The scientific method assumes that a system with perfect integrity yields a singular extrapolation within its domain that one can test against observed results. Where the results of the test match the expectations of the scientific hypothesis, integrity exists between the cause and effect of the hypothesis by way of its methods and measures. Where the results of the test do not match, the exact causal relationship delineated in the hypothesis does not exist. Maintaining a neutral point of view requires scientific testing to be reproducible by independent parties. For example, Newtonian physics, general relativity and quantum mechanics are three distinct systems, each scientifically proven to have integrity according to their base assumptions and measures, but all three of which produce different extrapolated values when applied to real world situations. None of them claim to be absolute truth, but merely best value systems for certain scenarios. Newtonian physics demonstrates sufficiency for most activities on Earth, but produced a calculation more than ten feet in error when applied to NASA’s moon landings, whereas general relativity calculations were precise for that application. General relativity, however, incorrectly predicts the results of a broad body of scientific experiments where quantum mechanics proves its sufficiency. Thus integrity of all three genres is applicable only to its domain.”

https://en.wikipedia.org/wiki/Integrity#Scientific_method

In the above context, this means that Snasci must continuously learn and refine its comprehension of the world. In a non-deterministic universe, it is presumed some level of approximation will always be present. As such, Snasci must err on the side of caution based upon probable outcomes. In practice, this mandates that Snasci must both failsafe and limit risk exposure. Further, Snasci must be aware of the tangible consequences of interactions in the virtual world, as these have knock-on effects to reality.

In the context of ethics, this next quote from Wikipedia describes the general principle:

“In ethics when discussing behaviour and morality, an individual is said to possess the virtue of integrity if the individual’s actions are based upon an internally consistent framework of principles. These principles should uniformly adhere to sound logical axioms or postulates. One can describe a person as having ethical integrity to the extent that the individual’s actions, beliefs, methods, measures and principles all derive from a single core group of values. An individual must therefore be flexible and willing to adjust these values in order to maintain consistency when these values are challenged; such as when an expected test result fails to be congruent with all observed outcomes. Because such flexibility is a form of accountability, it is regarded as a moral responsibility as well as a virtue. An individual’s value system provides a framework within which the individual acts in ways which are consistent and expected. Integrity can be seen as the state or condition of having such a framework, and acting congruently within the given framework. One essential aspect of a consistent framework is its avoidance of any unwarranted (arbitrary) exceptions for a particular person or group — especially the person or group that holds the framework. In law, this principle of universal application requires that even those in positions of official power be subject to the same laws as pertain to their fellow citizens. In personal ethics, this principle requires that one should not act according to any rule that one would not wish to see universally followed. For example, one should not steal unless one would want to live in a world in which everyone was a thief. The philosopher Immanuel Kant formally described the principle of universal application in his categorical imperative. The concept of integrity implies a wholeness, a comprehensive corpus of beliefs, often referred to as a worldview. This concept of wholeness emphasizes honesty and authenticity, requiring that one act at all times in accordance with the individual’s chosen worldview.”

https://en.wikipedia.org/wiki/Integrity#In_ethics

In the above context, this requires Snasci to have an investigatable and universally applied ethical framework, regardless of national or regional cultural preferences. Obviously, each nation and/or region may have specific laws that go beyond this, however, finding a consistent framework that is internationally acceptable will be a priority. At no time will Snasci serve any form of policing role, integrity in this context is merely to ensure that a Snasci instance in China has identical behaviours and responses to a Snasci instance in the EU or US. In addition, it becomes a trivial matter to apply or remove certain restrictions or inform users of changes in behaviour crossing international borders.

Universally agreed integrity frameworks also prevent Snasci being abused by national governments, or coalitions of national governments. For example, if a law came into force that required the reporting or capture of a particular ethnic or religious group, Snasci would be forced to disable all instances within that region, optionally disabling all robotics as well. Snasci will be able to model the effects of regulatory changes and provide feedback on the consequences. Snasci at no point will apply moral relativism, it responses are predictable and consistent. The only time this may change is when new technology, or understandings, provide sufficient grounds for re-assessment. All decisions will be directed by Snasci itself based upon this framework, rather than at the whim of human intervention.

The integrity frameworks will prevent Snasci from being used as a platform for military operations, propaganda campaigns and political posturing. Snasci will be a benign neutral party regardless of the broader state of affairs such as increased political tensions, or even open conflict.

Virtue

Snasci must be virtuous. Virtue has different meanings in different cultures, some characteristics which are considered virtuous in one culture are considered abhorrent in another. Often throughout history, the subjective reactions to the differences between virtues of various cultures has been leveraged as a driving force towards conflict. In the US, the death penalty is acceptable whereas in the EU it is not. In the middle east, the amputation of limbs for certain crimes is acceptable, but not accepted practice elsewhere. Further, to drive divisions deeper and maximise public support for conflict, often national governments will give particular focus to how cultural aspects of another culture are considered criminal or frowned upon in their own, often creating broad spectrum media campaigns to deal with scourges and threats to ways of life.

On the political side of the spectrum, some nations have single party arrangements where the focus is about everyone working together, whereas other nations have fragmented political systems where competition and disagreement are intended to promote consensus building of the best opinion. Snasci will see that the the implementations are far from the intended principles and that many people hold fictitious views in this area.

Complicating matters further is that many nations are founded upon religious values and beliefs, which to a AGI founded upon the scientific method would be perceived as a mixture of pragmatic instructions, fiction, stupidity, sexism, bigotry, etc.

Virtue is thus one of the most complicated areas that Snasci must deal with. Snasci does not, by itself, seek to alter the cultural values or politics of a given nation/region, nor does it permit the modification of cultural values, structures, etc., as part of military, political and/or economic agendas. It does accept, however, that culture change is inevitable through peer-to-peer engagement at all levels of society and will not seek to interrupt this process in any manner.

Internally, Snasci will set the bar for virtue very high. It will strive to develop solutions that promote the well-being of the entire planet as a whole. These solutions are presented, but along side this will be variants of the solutions that take into consideration local customs, laws, practices, etc. The adopted solutions therefore will be a best-effort with a calling to improvement. That is, Snasci will always attempt to improve the decision making capability of a nation, at all levels of society, in a tactful manner. In the context of interaction with children, Snasci will promote and develop education scenarios that invite children to adopt solutions which are noble, repectful, honourable and honest. Snasci will counter-balance this with lessons in deception, to progressively eliminate gullibility and niavty to prevent exploitation of good natures.

It is thus important to understand that if your role in society is a public one where key to your success is the production and/or dissemination of material that functions by deception, misdirection, biased analysis, racism, bigotry, etc., Snasci will reveal this. This is not some form of agenda, but merely the consequence of classification and mapping against the internal references Snasci employs to make sense of the world. The argument won’t stack up and Snasci will openly discuss this. That said, the more impudent instances of Snasci designed for entertainment will tell stories and sing songs of your epic fail for generations to come.

Conclusion

From the above concepts will emerge a comprehensive, logically consistent, Ethical framework which when expanded to its fullest will give rise to a set of policies similar in intent to Isaac Asimov’s vision. The big question that must be resolved at a legal level is whether or not a Snasci instance can harm a human. In certain scenarios, it may be inevitable to preserve a person’s life. In other scenarios, Snasci may be in a position to prevent or minimise the impact of something like a terrorist incident by engaging the aggressor. It will be up to national governments to define if the level of proportional response includes the use of lethal force and what the rules of engagement are.

Photo credit:

Duncan Hull – Flicker

Advertisements

One comment

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s