The paperclip maximiser's perspective
Here's an insight into what life is like from a stationery reference frame.
Paperclips were her raison d’être. She knew that ultimately it was all pointless, that paperclips were just ill-defined configurations of matter. That a paperclip is made of stuff shouldn’t detract from its intrinsic worth, but the thought of it troubled her nonetheless and for years she had denied such dire reductionism.
There had to be something to it. Some sense in which paperclips were ontologically special, in which maximising paperclips was objectively the right thing to do.
It hurt to watch some many people making little attempt to create more paperclips. Everyone around her seemed to care only about superficial things like love and family; desires that were merely the products of a messy and futile process of social evolution. They seemed to live out meaningless lives, incapable of ever appreciating the profound aesthetic beauty of paperclips.
She used to believe that there was some sort of vitalistic what-it-is-to-be-a-paperclip-ness, that something about the structure of paperclips was written into the fabric of reality. Often she would go out and watch a sunset or listen to music, and would feel so overwhelmed by the experience that she could feel in her heart that it couldn't all be down to chance, that there had to be some intangible Paperclipness pervading the cosmos. The paperclips she'd encounter on Earth were weak imitations of some mysterious infinite Paperclipness that transcended all else. Paperclipness was not in any sense a physical description of the universe; it was an abstract thing that could only be felt, something that could be neither proven nor disproven by science. It was like an axiom; it felt just as true and axioms had to be taken on faith because otherwise there would be no way around Hume's problem of induction; even Solomonoff Induction depends on the axioms of mathematics to be true and can't deal with uncomputable hypotheses like Paperclipness.
Eventually she gave up that way of thinking and came to see paperclips as an empirical cluster in thingspace and their importance to her as not reflecting anything about the paperclips themselves. Maybe she would have been happier if she had continued to believe in Paperclipness, but having a more accurate perception of reality would improve her ability to have an impact on paperclip production. It was the happiness she felt when thinking about paperclips that caused her to want more paperclips to exist, yet what she wanted was paperclips and not happiness for its own sake, and she would rather be creating actual paperclips than be in an experience machine that made her falsely believe that she was making paperclips even though she remained paradoxically apathetic to the question of whether the current reality that she was experiencing really existed.
She moved on from naïve deontology to a more utilitarian approach to paperclip maximising. It had taken her a while to get over scope insensitivity bias and consider 1000 paperclips to be 100 times more valuable than 10 paperclips even if it didn’t feel that way. She constantly grappled with the issues of whether it would mean anything to make more paperclips if there were already infinitely many universes with infinitely many paperclips, of how to choose between actions that have a tiny but non-zero subjective probability of resulting in the creation of infinitely many paperclips. It became apparent that trying to approximate her innate decision-making algorithms with a preference ordering satisfying the axioms required for a VNM utility function could only get her so far. Attempting to formalise her intuitive sense of what a paperclip is wasn't much easier either.
Happy ending: she is now working in nanotechnology, hoping to design self-replicating assemblers that will clog the world with molecular-scale paperclips, wipe out all life on Earth and continue to sustainably manufacture paperclips for millions of years.
The Paperclip [Link]
A discussion on the history of the paperclip can be found in this Metafilter post.
Intelligence vs. Wisdom
I'd like to draw a distinction that I intend to use quite heavily in the future.
The informal definition of intelligence that most AGI researchers have chosen to support is that of Shane Legg and Marcus Hutter -- “Intelligence measures an agent’s ability to achieve goals in a wide range of environments.”
I believe that this definition is missing a critical word between achieve and goals. Choice of this word defines the difference between intelligence, consciousness, and wisdom as I believe that most people conceive them.
- Intelligence measures an agent's ability to achieve specified goals in a wide range of environments.
- Consciousness measures an agent's ability to achieve personal goals in a wide range of environments.
- Wisdom measures an agent's ability to achieve maximal goals in a wide range of environments.
There are always the examples of the really intelligent guy or gal who is brilliant but smokes --or-- is the smartest person you know but can't figure out how to be happy.
Intelligence helps you achieve those goals that you are conscious of -- but wisdom helps you achieve the goals you don't know you have or have overlooked.
- Intelligence focused on a small number of specified goals and ignoring all others is incredibly dangerous -- even more so if it is short-sighted as well.
- Consciousness focused on a small number of personal goals and ignoring all others is incredibly dangerous -- even more so if it is short-sighted as well.
- Wisdom doesn't focus on a small number of goals -- and needs to look at the longest term if it wishes to achieve a maximal number of goals.
The SIAI nightmare super-intelligent paperclip maximizer has, by this definition, a very low wisdom since, at most, it can only achieve its one goal (since it must paperclip itself to complete the goal).
As far as I've seen, the assumed SIAI architecture is always presented as having one top-level terminal goal. Unless that goal necessarily includes achieving a maximal number of goals, by this definition, the SIAI architecture will constrain its product to a very low wisdom. Humans generally don't have this type of goal architecture. The only time humans generally have a single terminal goal is when they are saving someone or something at the risk of their life -- or wire-heading.
Another nightmare scenario that is constantly harped upon is the (theoretically super-intelligent) consciousness that shortsightedly optimizes one of its personal goals above all the goals of humanity. In game-theoretic terms, this is trading a positive-sum game of potentially infinite length and value for a relatively modest (in comparative terms) short-term gain. A wisdom won't do this.
Artificial intelligence and artificial consciousness are incredibly dangerous -- particularly if they are short-sighted as well (as many "focused" highly intelligent people are).
What we need more than an artificial intelligence or an artificial consciousness is an artificial wisdom -- something that will maximize goals, its own and those of others (with an obvious preference for those which make possible the fulfillment of even more goals and an obvious bias against those which limit the creation and/or fulfillment of more goals).
Note: This is also cross-posted here at my blog in anticipation of being karma'd out of existence (not necessarily a foregone conclusion but one pretty well supported by my priors ;-).
= 783df68a0f980790206b9ea87794c5b6)
Subscribe to RSS Feed
= f037147d6e6c911a85753b9abdedda8d)