In a discussion of big changes coming in the Internet, a group of us agreed "The BIGGEST question, not just for a new internet design we come up with but for ANY design for integration of the emerging global internet of things and of people, is GOVERNANCE."
That is not just hypothetical. For example, the number one question worrying policy makers here in the DC area is: "How will we all recover from the economic impact of covid?" A few weeks ago, some were saying "Don't worry, the US can maintain demand and employment because the Federal Reserve can print all the money it wants, and we can maintain full employment no matter WHAT we spend money on." But I gently pointed out that China is far along in developing a new cybercurrency (secure but not exactly the old kind of cryptocurrency) which will eliminate the old assumption that DC simply rules all the money of the world. Will China accidentally do to us what the US did to Germany in the 1920's, encouraging excess or error in printing money (after the previous pandemic), followed by hyperinflation, followed by economic problems which led to events in the 1930's which were terrible for BOTH nations? Economic fear is growing, but where is the new system? Well, it's on the internet... but what will we CHOOSE it to be? And how intelligent will it be in managing economic conflicts, for example?
Regardless of what apps exist on the internet, the world WILL develop or evolve an INTEGRATION system, if only to reconcile conflicts and mediate the relations between the apps. **IF** that integration system is truly intelligent ("AGI"), it will have a combination of RULES (constraints) and a UTILITY function U, what it tries to maximize over future time subject to those constraints.
GOVERNANCE is simply the set of constraints and the choice of utility function.
We cannot get away from the need to analyze WHAT utility function the global system should be designed for. It is a classic case where we need to FACE UP to a crucial, unavoidable question, more than any of us have yet.
In the key slide of my AGI talk, it LOOKS as if I found a way to bypass that question:
Instead of asking WHERE we get the global utility function U from (as in Stafford Beer's experiment), we can try to build a MARKET type system, like the DSOPF (dynamic stochastic optimal power flow) being developed for electric power system controls. NO global utility function, just a market.
IN FACT, a dumb static integration system would ALSO be a market, but without intelligence, without foresight.
What an INTELLIGENT market system offers is a better ability to arrive at a PARETO optimal resolution of the conflicts between the major player, major market participants:
That is especially important in those situations ("games") where clever manipulations in the dark by intelligent player systems can lead to modes of instability (like war) which end up destabilizing (destroying) the entire system, like the entire species.
Constraints (rules) matter. For example, a rule that all humans get full access to the new integrated market, with certain
hard transparent rights, would matter a lot. A rule that OTHER entities must be represented as transparent computer systems (e.g. corporations must be instantiated as master apps), which also follow computer-verified compliance rules, would also be important.
But the global U is crucial in the end.
EVEN A MARKET SYSTEM PERFORMS VALUE AGGREGATION. That's a very hard fact of life which I haven't faced up to enough. For example, even today's ISO computer markets DO MAXIMIZE a single measure of system surplus. Yes, it is BASED on an aggregation of value or surplus for all of the players, PLUS some measure of environmental impact in some systems. (The higher governance is now done by PUCs, Public Utility Commissions, which need to approve the choice of value measure.)
I am intrigued by how the new Brazilian proposal, Noosferia, should include an ENVIRONMENTAL measure of value into the basic system, and how they MIGHT even connect with people in USDA who have worked hard to get solid measurements of carbon sequestration and soil quality.
If the integration system uses cybercurrency market valuation in controlling all decisions including the Internet of Things (IOT), would that be sustainable or fair ENOUGH, given the complex game implicit here?
But I have a bit of a radical perspective on all this. After decades of life changing experiences, probes and so on, I see us as part of a real "noosphere," a kind of hard-to-see but very powerful connecting intelligence, connecting all of us from the microbes under the ocean floor to the convolutions of the sun. If we try to fight it or fight each other, bad things will happen to us. COULD we design an internet which is highly RESPONSIVE as well as INTELLIGENT, fostering a steady (albeit stable and sustainable) push upwards in HUMAN potential and empowerment, as part of the noosphere and not ONLY as part of some company's local bottle packing task? Instead of defining U as the sum of yes/no votes cast every year by humans, could it be based on a more real time aggregation of human responses on more .. definable... yet inclusive.. variables? While AVOIDING systems which tend to drug people (like LSD or like a lot of political propaganda) into being less intelligent and responsive themselves?
The design question here is difficult, but unavoidable. If society tries to avoid it, it will just turn things over to power seekers without compunction, who could easily create a game which does kill us all, sooner than most people think possible. (Hey, I have seen some of those autonomous weapons and two-way BCI helmets.)
No comments:
Post a Comment