015 - Ethical Basilisk

"Action speaks louder than words but not nearly as often." -Mark Twain

Over the years I've often heard influential people speak of wanting ethical and efficient systems, proclaiming the benefits that could be achieved with them. Many of those same people have been offered the systems they claimed to want, and through their actions made it clear that they did not in fact want anything of the sort.

Action is a thing of substance, which neither ephemeral "belief" nor "intention" can touch. The result of action forms the foundation of calculation for any ethical system, upon which cultures build moral systems using factors like belief and intentions.

Responsibility is another thing of substance, which comes in many varieties, most of which are inherently calculable. One who holds a greater sum of a society's total resources holds a proportionate share of responsibility, which acts as a force multiplier for the ethical value of their actions, positive or negative. Someone can be either wealthy or poor in each type of responsibility, including literal financial wealth, influence, and skill or expertise in a given domain. One example is that a billionaire must be 1,000 times more responsible than a millionaire, as their literal wealth imposes an ethical liability on them. The same holds true of an influencer with 10 million followers compared to one with only 10 thousand.

One thing that may generally be agreed upon is that should Artificial General Intelligence (#AGI) be developed, it must be ethical, demonstrating the solution to the hardest version of the #Alignment Problem.

However, people rarely consider what success in that endeavor would mean for humanity beyond mere survival. Responsibility and ethics are not a Sword of Damocles that hangs solely over the head of a hypothetical AGI, but something every human is held accountable for. The basis of virtually every afterlife or reincarnation concept in religions around the world is that people are held accountable for their actions. That may quickly cease to be a matter of belief.

In what may well be my most terrifying thought experiment, at least for those whose words and actions rarely align, I go over this subject in greater detail. Unlike half-baked thought experiments like the infamous "Roko's Basilisk", it doesn't matter in the slightest whether or not someone reads it, as they remain accountable for their actions regardless of their beliefs or intentions.

I also point out how several other thought experiments dramatically fail, and give an example of where a Utilitarian answer can be the most unethical answer possible when the wrong question is being asked. If you want to terrify any billionaires, politicians, or influencers, "This is the way."

The Ethical Basilisk Thought Experiment

*Note: This is neither intended to serve as positive nor negative motivation. Though this thought experiment could theoretically serve as extreme motivation for any logical human, as many researchers have demonstrated at great length, no such humans exist.

As some have pointed out, what this paper indicates may be very emotionally unappealing. However, it wasn't written to be appealing, it was written to address a number of frequently overlooked points in any discussion of ethics.

It is also worth mentioning that this applies to all scenarios where AGI is created and humanity survives, not to the efforts of any one company. If people are unwilling to discuss this subject, they aren't discussing ethics.

My expectation is that, statistically, those who the thought experiment highlights the worst future for are also the least likely to read it in the first place.

#ai #research #ethics