Living Earth Simulator will simulate the entire world

I honestly don't know what to make of this. There are intelligent people involved in it; I've attended talks by some of them.

Its not as crazy as the sensationalist article headline makes it out to be; one excerpt from the "Project Outline" from http://www.futurict.eu/ shows this:

"Most importantly, the FuturICT project will NOT attempt to collect “all the data in the world”, or to represent each individual on the globe by an identical copy in some giant multi-agent simulation, considering private features and preferences of all individuals. Science is the art of abstraction and approximation. Just as maps do not show all the features of our environment, a scientific model is specified such that a particular question can be addressed in the simplest possible way. That is, models are to be problem-specific, and parameters and variables not expected to be relevant for the answer should be neglected."

I don't know much about the details of the specific project; but I think a large degree of scepticism is healthy, when discussing any proposal to build models with very large scope.

People have put a lot of effort into making economic models, but we still aren't much good at predicting the kinds of phenomena that we are interested in, such as the occurrence of crashes. It can be easy to model large scale social phenomena; but modelling them usefully is very hard.

It sounds very ambitious; but then there are lots of smart people involved; I don't know.

Interesting. Will the LES take into account its own existence (e.g. the change in people's behavior when they are presented with LES predictions)? Of course, doing so would either (a) make the problem undecidable (impossible to solve), or (b) require calculation of fixpoints (very difficult to solve), or both.

The other alternatives are for LES results to be kept secret and used only for policymaking, or for the results to be public and incorrect. I suspect the former is most likely.

I wonder who's gonna develop this system and how will they manage it? Sometimes even domains that appear simple end up reaching unexpected complexity, and this Simulator is already set to handle enormous complexity and massive data.

I can see the value in being an open source of cross-referenced data that would then be digested by proper research. Though as a simulator, I'm not so sure.

Not to mention the bugs. I doubt this complex undertaking will be bug free, and being not so, can the outcome be reliable? Who will debug the predications?

I think, as a rule, that we should not leave to computation what we ourselves cannot fathom. If we had a grand understanding on human behavior, individual and collective, then maybe this could be done, but how can we determine the correctness of a program that does something we can't, right now, fully comprehend?

I think there's much more value in trying to procure all this data in order to enhance understanding of social behavior in face of historical data.

A few years ago, I had an interesting job where I was on a team (of 2) that wrote a simulation of the UK criminal justice system. Police, criminals, courts, prisons, etc.

The idea was that the client could tweak various features (numbers of judges, sentences for specific crimes, number of police on the beat) and look at unexpected secondary effects this would have, like increased time in probation or more prison overcrowding.

It was written in OCaml no less :-)

I wasn't involved in using the simulation that much, so I often wonder how accurate and effective it was in the end.

Somewhat more detailed info here: http://www.futurict.eu/the-project/proposal

Seems they want to wire up every realtime data source they can get their hands on to a massive database, then allow researchers to upload apps to run on the data. So, there would be a variety of simulations running, using different models.

They would also provide access to the simulations at least to important decision makers. It's not clear exactly how open the access would be.

Asimov hashed through all this quite thoroughly in the Foundation series.

But it is always interesting when we catch up to classic sci-if and work towards actual implementations of what was pure imagination only a short time ago.

...and its only a matter of time before the simulated world creates their own simulation...

But I wonder how many simulations deep we are now???

It's interesting to note that the first automated simulator of real-world processes was the water-powered MONIAC, built only 62 years ago. (MIT Whirlwind, the first digital simulator, was finished two years later.)

Articles like this make me wonder if Eliezer Yudkowsky isn't so far off after all. If I ever meet him or Aubrey de Grey, I'll apologize for not taking them more seriously.

Edit: I messed up the links. Sorry.

Let's hope that 1 billion does not give them access to the computer and/or results of certain tests. 1 step closer to full in control government, calculating just how much tax each person will pay without taking action against governments. And much more like that. So I hope its used scientifically and not for political gains.

Can we be far from creating the Total Perspective Vortex?

This sounds like a whole lot of pseudo-science to me.

We're already living in one :)

This is pretty much what wall street does, except with a lot more than a billion dollars of investment. That fact that it's not clear whether the data will be made publicly available means it's just another scam on european taxpayers.