Following the Asilomar conference last month [https://futureoflife.org/2017/01/17/principled-ai-discussion-asilomar], I started thinking about Human Level AI and potential impacts on society as we know it today. To frame the focus, I chose the topic of addiction and support methods that could be enhanced with this kind of technology.
Conversation is an immensely powerful tool. Influence and social capital are products of conversation, instruction and linguistic expression. Interpretation is subjective, however, which leads to a unique emotional experience for every evaluator. Simple words, spoken or written, can inspire radical happiness, invoke deep sadness or any level of emotional response in between, depending on the recipient’s personal understanding and current state of mind.
Reading up on super AI inspired me to think about the boundaries of this kind of technology; and its potential to utilise the power of these natural emotional responses. Imagine a machine that is able to converse with an individual, deconstruct their responses and then, just as quickly, accurately predict the forecasted release of endorphins or chemical levels in the individuals system with the intent of altering the branch of communication accordingly. This could enable ultimate control over this power of influence.
Humans easily influence others through conversation. Even with a basic understanding of the power of psychology, humans rapidly predict an expected response and their up follow up actions. The unknown variable is the interpreters true understanding of the expressed. Only to be validated by a measurement of the chemical changes physically occurring within the hearer/reader. Assuming an AI that can harness these abilities and instantly assess the data to inform further expressions, there are questionable dilemmas.
Addicts and substance abusers who actively seek help in controlling their addictions often require the help of psychological advice and ongoing moral support. These conversations boost will-power, which in turn helps to suppress cravings through chemical reactions in the brain. Through the art of supporting and encouraging conversations, people can overcome immensely powerful personal challenges.
If human level AI of this magnitude is coupled with the power of instantly monitoring these chemical levels in the brain while soliciting encouragement that specifically boosts the individuals unique will-power, then presumably the affects would be much more instant. To create influence is to manipulate ultimately at a chemical level.
The ethical question that is begged here; to what level is it acceptable to allow AI to influence another living being, if at all. Presumably, to the benefit of the being, but under who’s interpretation of ‘benefit’? The more informed AI or the disadvantaged being?
In the case of above, the human’s reactions and responses are untrustworthy even to himself. Should an overpowering craving cause a relapse, then the human’s intent is compromised. If the AI has the power to subdue this craving, should it act on this power? Or allow this relapse to happen? In this case, help from the AI can only be successfully performed by disobeying the addicts wishes?
Surely professionals who currently offer supportive services with the aim to help someone overcome such daemons inherently tackle this kind of ethical dilemma – to advocate on behalf of those who cannot help themselves. Should Human Level AI hold true this responsibility if it is in its power?
Technology is constantly used and improved to overcome obstacles in medical sciences and this should no different in my opinion.