Fwd: Beyond Money ?

[Dante-Gabryell Monson]

---------- Forwarded message ----------
From: Shann Turnbull <stur...@mba1963.hbs.edu>
Date: Thu, Apr 3, 2014 at 1:09 PM
Subject: Re: Beyond Money ?
To: gbr...@listserv.vub.ac.be
Cc: "Evolution, Complexity and Cognition group" <evol...@listserv.vub.ac.be>, 

Ben

Weaver bring up points that also concern me.

I see your offer network as a variation of the Local Exchange Trading Schemes (LETS) that are located at hundreds of locations around the world as documented by the Complementary Currency Research Center located at http://communitycurrenciesinaction.eu/online-resources/

Your Offer Network raises the question raised by a Bank of England research paper that asked the question if "money will disappear from use?”  Refer to: http://ssrn.com/abstract=530686

E-Barter Versus Fiat Money: Will Central Banks Survive?

Abstract:      

New technology in computing has led some to suggest that the ability to settle transactions electronically will develop to such an extent that money will disappear from use. Two versions of this belief exist. One maintains that there will be 'e-money', issued conceivably by many organisations, and that this will replace central bank money. The other, on which this paper focuses, suggests a further development - that the very concept of a medium of exchange may become redundant, as assets or goods can be exchanged directly for other assets or goods through use of computing. In this paper we argue that the information-economising properties that allowed money to develop will also allow it to survive, despite actual and hypothesised technical progress which reduces the cost of electronic barter.

They concluded that money would still exist but perhaps not central banks.

The BitCoin technology allows money to be managed without central banks or interest.  Alderman Professor Michael Mainelli has co-authored an excellent paper on the topic. Refer to: 

'Building Bit – What’s A Poor Government To Do About AltCoins?' posted at: http://www.zyen.com/index.php?option=com_content&view=article&id=1393&Itemid=359

I have just posted my working paper on the topic: ‘Might supplementary tethered currencies reduce financial system risks?’ 29 March, posted at: http://ssrn.com/abstract=2417826.

I will be presenting a variation of this paper as a co-founding member of the Sustainable Money Working Group with the title: ‘Might Tethered Money Reduce Market Failures?’ to International Finance & Banking Society Confeence (IFABS) 2014, Lisbon, Portugal, June 18-20^th.  Details of the SMWG is posted at: https://sites.google.com/site/smwgorg/

Cheers

Shann

On 1 Apr 2014, at 10:31 pm, Weaver <silken...@gmail.com> wrote:

Hi Ben and all,

I finally found the time to read the article and the various responses.

Of course I like very much the idea of a distributed peer to peer economy.

Here are two preliminary reflections:

1. It seems to me that the offer network idea bundles together two related problems.

The first is basically a mathematical problem of extending the the unidimensional

value represented by money to a multidimensional one. This involves the definition of a space of values with an open ended dimensionality and a very complex metric necessary for it to serve its function. The second problem is designing and implementing a viable economy that utilizes multidimensional value system and works at least as good as the current one (that with all the obvious drawbacks seems to enable the operation of a complex civilization). It might be a good idea to decouple the two problems and discuss each of them separately.

2. Understandably the article focuses, for the sake of simplification, on the exchange of what I would call atomic units of value e.g. an Hamburger, and hour of programming etc. This simplification however hides and perhaps overlooks the vast complexity of the current economic machine and its transactions. I tried to play with the idea a bit and could not come up with an obvious or less obvious way of how can the mechanism of an offer network be scaled up to meet the complex challenges of today's economy. For example how can an offer network deal with the following complex operations:

1. Designing producing and distributing silicon chips (including the machines involved in the process). I use this example because of the criticality of this specific category of products to the future information society.

2. Building an Highway from city A to city B (same goes for an overseas shipping route, flight routes etc).

3. Run a biotechnological facility capable to research and produce a vaccine against say Ebola virus (lately on the news) or a medicine for Alzheimer (including setting research priorities etc).

4. Building AGI agents capable of operating in physical space and interacting with humans (that's for you Ben :-)).

These are just a few examples where the exchange process is extremely complex, involves a highly coordinated operation of hundreds and sometimes tens of thousands of agents, legal operations, budgeting, risk taking and long term planning in conditions of uncertainty.

In as much as we do not like the unidimensionality of the current value system and its hierarchical structure, I think it is far from easy to devise a scalable offer network capable of coordinating the more complex economic operations a progressive civilization requires. A possible approach to the problem of scalability is to create a mechanism that reduces the dimensionality of value as a function of the complexity of coordination inherent in increasing the scale of the exchange operation but this idea is no more than hand waving at this point. It certainly involves inventing new mathematical tools and not only smart algorithms.

Weaver 

 

On Sat, Mar 29, 2014 at 7:07 AM, Ben Goertzel <b...@goertzel.org> wrote:

Francis etc.,

> Before I would even start to think about implementing such a complex system,
> I would like to understand its pros and cons better on a conceptual and
> mathematical level. Then, I would like to simulate it in order to get an
> ideas of its dynamics in a more realistically complex setting.

Makes sense...

>The
> simulation environment that we are developing at the GBI, once fully up and
> running, should be able to do that.

Sounds plausible, though I don't have a strong sense of the GBI
simulation environment at the moment...

> But the first stage should be to formulate some algorithmic or mathematical
> specification of the central "clearinghouse"  function that matches offer
> and demand without relying on a quantitative measure of value such as money.
> One good approach may be the formalism we have been exploring already for a
> while at GBI, Chemical Organization Theory (COT, see references below). COT
> is inspired by chemical reactions in which an input condition (joint
> presence of certain "molecules") is transformed into an output condition
> (joint presence of a different set of molecules). E.g.
>
>  a + b -> c + d + e
>
>
> This is similar to an individual who wants a and b, and is willing to
> produce c, d and e in return, as in your offer network.

Hmmm....   Indeed it's *similar*, but I'm not sure it's similar enough
that one would want to use a chemistry simulator (even an abstracted
one) to simulate an offer network....

But here we get deep into the details and I'd need to look at COT more
closely to have a strong opinion...

One could straightforwardly set up a simulation of agents that have
the capability to provide various goods and services, and have
needs/desires for various goods and services.    One could also give
each agent a certain limited capability to predict its future
capabilities, needs and desires....   Using such a simulation, one
could "stupidity test" an Offer Networks framework, and see if it
behaves OK for the simulated agents.  This would be a good idea.   But
of course, setting up a *realistic* agent population is infeasible (as
there is insufficient data about the actual probability distributions
of real human capabilities/needs/desires, for one thing), so there
would still be lots to learn from trying such a framework with real
humans as well...

> At present, we are considering dynamics that basically mimic "natural"
> processes, such as reinforcement learning. But if you have in mind a
> specific clearinghouse algorithm that would distribute the tasks and inputs
> in a better way, this could be tested out as well in such an environment...

I haven't thought about the clearinghouse algorithm in detail.

Here are two possibilities, though.  I like the second one better but
will elaborate both for completeness/ interest

OPTIMIZATION APPROACH
--------------------------

1)
Formulate an evolutionary learning problem, where each genotype is a
set of pairs (person, action-set), indicating a set of possible
actions undertaken by the people involved in the network; and the
fitness of the genotype is: 0 if anyone in the network is doing
something they're unwilling to do; otherwise, the total sum of the
weights that each person assigns to the (actions of which X is giver,
actions of which X is recipient) set in which they are involved.

2)
Fitness evaluation involves calculating the barter chains involved in
an optimal implementation of a given genotype.  This can be done via
some sort of heuristic search algorithm.   Dynamic programming is too
expensive, so some approximative approach must be used.  The fitness
evaluation therefore will just involve an estimate of the actual
fitness of the genotype, indicating the fitness according to the best
corresponding barter chains that the heuristic search algorithm can
find.

3)
Fitness estimation will be critical, and will involve caching of
information about barter chains found during prior fitness
evaluations, to speed up heuristic search (e.g. by allowing rapid
discarding of genotypes containing combinations previously found
infeasible)

...

This seems a difficult, but very interesting, optimization problem;
and there may well be a more efficient approach than I've outlined
above....   Evolutionary learning is rarely the best approach to
solving a problem, though it's often the easiest approach to specify
;)


SIMULATION APPROACH
--------------------------

An alternative would be a Monte Carlo approach, wherein you have a
simulated agent for each member of the Offer Network, and you let them
trade with each other for a while and see what results from all the
trading.  But you run the whole simulation many, many times and record
the results.  Then, from all the simulations, you measure the fitness
of the end result (via the same method suggested in Step 1 of the
evolutionary approach, above).   And you have the clearinghouse
recommend the assignments that would result from the best simulation
world found...

Of course, the effectiveness of this approach depends on how dumb the
simulated agents are.   Caching information about barter chains found
in prior simulations, as in Step 3 of the evolutionary approach
mentioned above, would make the simulated agents less stupid as the
series of simulations proceeds, and improve the ultimate result found
in most cases...

...


I think the simulation approach is better , as it seems easier to
tune; both approaches would be computationally expensive and it's hard
to compare them in this regard without going a lot further along in
the direction of implementation/experimentation...

...


Alternative suggestions would also be valued ;)


-- Ben

--

"It is not guilty pride but the ceaselessly reawakened instinct of the game which calls forth new worlds."
(Heraclitus Reloaded)

Shann Turnbull PhD; Principal: International Institute for Self-governance
PO Box 266, Woollahra, Sydney Australia, 1350
Ph:+612 9327 8487; Cell: +61 (0) 418 222 378
Skype and Google ID: shann.turnbull

Bibliography/CV: http://independent.academia.edu/ShannTurnbull/CurriculumVitae