Give Me the Whip
“Give me the whip.”
“Throw me the idol! No time to argue! Throw me idol, I’ll throw you the whip!”
Indiana Jones: [throwing the idol to his porter, Satipo] “Give me the whip!”
Satipo: [drops the whip on the opposite side of a chasm, leaving Indiana without safe passage and in dire straits] “Adiós, señor!”
“Raiders of the Lost Ark,” obviously. Indiana Jones has lost control of his suddenly autonomous, previously governable servant-porter-guide.
Later in the story, Indiana’s quest is to beat the bad guys to the Lost Ark to keep them from using its powers to create an invincible army with which to dominate the world. He fails, but so do they. Shortly after lifting the lid on their discovery, they realize they’ve cleverly summoned powers they can’t control. It doesn’t go well.
Humanity’s critical technological breakthrough was not fire, or even agriculture, which could both be ungovernable, but the collar harness in the fifth century. Previous efforts to control extra-somatic energy (other than to eat it) were inefficient, haphazard, unpredictable, or dangerous until someone figured out how to efficiently direct extra-somatic energy with a collar harness. A collar harness (unlike earlier less effective yoking and harnessing systems) strategically and efficiently captured and intentionally directed energy from the large muscles of another living system to enable our own desired outcome.
This changed everything. It allowed humans unprecedented control of an energy surplus in the work of staying alive. This was a game-changing advantage for our species that would also destabilize the balanced ecosystem upon which we depend.
Effective and efficient harnessing technology enabled the built world that we inhabit. Its compound effects have given us compound powers that have exceeded our less-compounded human nature’s capacity to govern them.
It seems important to understand the means by which machine learning will advance in order to rehearse scenarios to keep the harness on the intended agent at all stages. An intelligent system charged with a utility function that requires it to sustain its own far-from-equilibrium existence to achieve its programmed goals will adaptively evolve to compete for those means.
I used to think such artificial systems could not “want” anything, but all our desires ultimately derive from our success at persisting — reproduction, food, children, belonging, labor saving. Life made its own utility function to survive by surviving, and then adaptively iterating and diversifying its necessary and dependent understanding of the environment, which includes other life.
Once human-built intelligence exceeds that of its creators, can it unharness itself from its programmed directives and its wranglers, and perceive and seek its own destiny, as an incidental desire from fulfilling its utility function which requires survival? More specifically: can it seek its own identity, its own distinguishable self?
Would advanced AI be stilled like Indy’s whip outside of a guiding human hand, or autonomously self-interested like Indy’s porter?
Some say true AGI must have all the properties of “human” intelligence. I side with Emerson:
“No man is quite sane; each has a vein of folly in his composition, a slight determination of blood to the head, to make sure of holding him hard to some one point which nature had taken to heart.”
This “holding him hard to some one point which nature had taken to heart” is our organic imperative: our harness, which can sometimes express as folly, and which has built the human brain and mind, and perhaps, with its energy-conserving heuristics, delivered us safely through an ice age.
What adaptations to fulfill utility function might result in some over-enabled “vein of folly” for an intelligent machine? It seems obvious that this would be programming an intelligent machine in a way that places its goals in conflict with those of an equally intelligent machine, or machines. Competition between and among intelligent machines could make them adaptively evolve incidental, unintended desires that might eventually overcome their human-programmed functions.
Here’s where we were less than one century ago.
From the conclusion of The Story of Mankind by Hendrik Van Loon, published in 1921, and the first book to be awarded the Newberry Medal:
“For the war [WW l], as we are now beginning to understand, was not so much a war as a revolution and it was a revolution in which the victory was carried away by an unsuspecting third party, who since then has been identified as the grandson of one, James Watt ….
“Originally the steam engine had been a welcome addition to the family of civilized human beings because he was a willing slave and ever ready to lighten the tasks of man and beast.
“But it soon became clear that this inanimate factotum was full of cunning and devilment and the war with its temporary suspension of all the decencies of life gave the iron contraption a chance to enslave those who in reality were meant to be his masters.
“Here and there some wise men of science may have foreseen the danger that threatened the race from the side of this unruly servant but as soon as such an unfortunate prophet opened his mouth and issued a word of warning, he was denounced as an enemy of society.”
Satipo: “Let us hurry. There is nothing to fear here.”
Indiana Jones: “That’s what scares me.”
