Acceleration AI Ethics

A framework for managing the tension between innovation and safety in artificial intelligence.

 

 

Origin


 

Acceleration ethics in the area of information technology can be concieved as developing through three stages. 

The first is a reaction against the precautionary principle: precaution short circuits because it creates the dangers it claims to mitigate. This reaction occurs on the levels of the environment, healthcare, and behavioral economics.

The second developmental stage flips the anti-precaution principle into a positive stance, sometimes called the "proactionary principle." Initially developed by Max More, the proactive posture is based in large part on Extropy Institute's Vital Progress Summit in 2004. As background, that group convened to oppose precautionary limits imposed on stem cell research in biology. Their response contained seven "factors." The initial one is ethical:

1. People's freedom to innovate technologically is humanly valuable. The burden of proof therefore belongs to those who propose restrictive measures.

Subsequent factors concern proportion and measurement, they are about how risks may be measured and compared, and do not belong to ethics so much as socio-technical studies, or risk management.

2. Evaluate risk according to available science, not popular perception, and allow for common reasoning biases.

3. Give precedence to ameliorating known threats to human health and environmental quality over hypothetical risks.

4. Treat technological risks on the same basis as natural risks.

5. Estimate and account for the lost opportunities and benefits of abandoning a technology.

6. Consider restrictive measures only if the potential impact of an activity has both significant probability and severity. If restrictive measures appear justified, they should be proportionate to the potential negative effects.

7. Rank risks as follows: Human and other intelligent life over risks to other species; non-lethal threats to human health priority over threats limited to the environment (within reasonable limits); priority to immediate threats over distant threats; more certain threats over less certain, and irreversible impacts over transient ones.

The proactionary stance has been supported more recently by Steve Fuller and Veronika Lipinska. Together, they promote the attitude that risk is an opportunity more than a threat, though they explain their conception on the economic or even entrepreneurial level more than as an ethical stance. Fuller calls for an "entrepreneurial approach" to risk, and states that it is a kind of "opening in the market that you can do something with."

The third stage transfers the discussion from the levels of regulation and socio-technical studies, onto the level of human values. In other words, for advocates of proaction, there are regulatory and socio-technical stances first, and they imply an ethics, while for the acceleration approach there is first an ethics, and that creates regulatory and socio-technical effects. While the acceleration approach is more humanly fundamental, the outcomes in terms of what happens in the lived world are not entirely incongruent.