Simulating worlds

Some time during the past weekend, I ended up in a rather silly but for me entertaining thought experiment: I was musing about that there should be a way to produce the so called “hindsight bias” in advance of an event. What started as a joking idea quickly led me to some more serious reflections.

The term hindsight bias (according to Wikipedia also called “knew-it-all-along effect”) refers to a cognitive bias which brings people to the belief that the outcome of a certain event or situation was the only logical and possible result. Before the specific event, uncertainty about what happens next is widespread and predictions about the future are varying widely. But in the aftermath people experience a feeling of obvious and overwhelming retroactive predictability of whatever happened. Suddenly, everyone claims to always have expected this very outcome.

If you debate the question of how Artificial Intelligence, Virtual Reality or self-driving cars will change human life, you can hear plenty of different theories and predictions. However, in 20 or 30 years, people will point out that whatever will have happened after AI, VR and self-driving cars took over, was the one and only logical scenario.

Now, if you could somehow produce the hindsight bias about the outcomes of AI, VR and self-driving cars today, instead of in 20 or 30 years, that would be pretty neat and useful, wouldn’t it?

Sign up for the weekly email loaded with great things to read about the digital world. Example.

While my thought experiment for apparent reasons was not meant seriously, it made me mentally recapitulate the book about accurate forecasting which I reviewed briefly a few weeks ago. It also made me think of the article about the upcoming, revolutionary video game No Man’s Sky which I linked to last Thursday. This game is built to simulate an artificial universe and, according to the description, will be different from conventional games by that it does not rely on “faked” physics but on a real simulation of physical conditions (such as that planets rotate). You might be wondering what this has to do with my initial thought experiment? Bare with me, I’ll explain soon.

I am not a gamer, but I had periods in which I enjoyed the occasional session of GTA or similar titles. What always annoyed me was the fact that all other characters within a game did not have an actual life within the simulated world. If you created a massive scene of destruction involving other cars and objects, or a large traffic jam, and walked away for a short moment, the whole situation was gone shortly after, as if nothing ever happened. The computing capacity and probably also the type of software required to create a simulated game world in which all objects would be “existing” all the time, even if you as the player hang out in another part of the game, was simply not available. But if No Man’s Sky or the similarly exciting plans of the cloud-based gaming platform Improbable are an indicator, the status quo is changing and full, real-time simulations of virtual worlds are becoming possible.

People will never be able to produce the hindsight bias beforehand. It’s inherently impossible. But instead people will be able to rely on digital simulations in order to make more accurate predictions – and decisions. Here is the mind-boggling thing: Thanks to the advances in computing which now allow for the creation of a full-fledged universe simulation, one day a simulation of the world and its 7+ billion inhabitants might become reality.

Imagine if you could feed a system with a lot of real-world data and then let it generate a real-world simulation of what happens in a scenario in which the UK leaves the European Union. What impact will it have on the E.U. and the U.K.? How will the currency and economies be affected? What consequences to the political landscape have to be expected? How will the public opinion evolve after the split? Will any new political movements appear as a consequence of the split?

Simulations of that kind are already being performed in many places (a discipline dubbed “Modeling & Simulation“, M&S), even if still at a comparatively primitive stage and often limited to specific fields, such as weather, the stock market or science & education. The capabilities and capacities of such simulations will only keep getting better though, and many video games of the future will probably be build around full simulations. I am the wrong person to assess how fast this progress will be and whether near-accurate simulations of planet Earth and all its inhabitants will be possible within a near future. But if you consider the truly exponential advancements in computer science, it might not take too long. Let’s remember Ray Kurzweil’s words: We won’t experience 100 years of progress in the 21st century, but more like 20.000 years of progress.

There is obvious potential for misuse of such simulations by governments, intelligence agencies or certain commercial players. But that’s a topic which deserves more than 1 or 2 paragraphs so I won’t debate that at this point. As usually with new digital technology, it comes with possibilities and dangers.

The idea of simulating life or (virtual) worlds without relying on faked physics fascinates me. It looks like there is reason to be excited.

P.S. If this is a topic that you know a lot about or if you have interesting links which can improve and deepen my and the readers’ understanding of this topic, please share it in the comments.

If you like what you read, you can support on Patreon!


  1. Chaos theory gets in the way of accurate long term predictions of most complex systems. No matter how much computing power you throw at it, no matter how good your models are, if the system is chaotic (i.e. small differences at time a make for huge differences at time b) there is no predicting it. Many systems do have attractors though and statements like this might be deducible from such simulations: with 30% probability Brexit will be for the better, with 70% we’ll be toast.
    You see how this works for the weather. There is still a lot of headroom for improving the weather forecast. A much denser sensory network and more computer power may get us two weeks with useful probabilities and in special conditions maybe more. But we’ll never get a year, no matter what.
    Now, if you want to make meaningful predictions about human societies (like the Brexit prediction above), the prediction itself will influence the outcome … ad infinitum. I don’t know, if there is an exit off such feedback loops.
    One really useful outcome of such simulations may be this though: a simulation mapping probabilistic futures would effectively create a map of multi dimensional possibility space. Such a map could tell us that there is a slippery slope that leads to a horrible abyss and we’d rather go somewhere else – assuming that our future is not so riddled with abysses that “somewhere else” is indeed a more promising alternative.

    Greg Egan, by far my favorite SF author, writes a lot about VR and simulated societies. I recently read a short novelette by him that nicely sums up many aspects he’s covered: “Wang’s Carpets” by Greg Egan, one or two hour’s read, can be a had for a buck in many e-book stores. May require substantial scientific education for full comprehension.

    • Crucial and important points, thanks for sharing. Even if we can’t have accurate predictions based on long-term simulations, this alone what you are describing here would be quite a significant tool:
      “a simulation mapping probabilistic futures would effectively create a map of multi dimensional possibility space. Such a map could tell us that there is a slippery slope that leads to a horrible abyss and we’d rather go somewhere else – assuming that our future is not so riddled with abysses that “somewhere else” is indeed a more promising alternative.”

  2. I love the fact that this has sparked a new thinking perspective and abstract thinking like; this excite me. So thanks!

    The only reason I see this to be near impossible in any sort of foreseeable future is due to the complex array of human emotion and decisions.

    For those who are familiar with the mind body problem I suggest that this issue can be implemented into this debate.

    Those with the position that the brain is incomporial would suggest that the a computer that works through a process of patterns and has a systematic approach can never duplicate a human mind.
    The human mind influences the decisions of any political, investment plan, and so on so forth. If a computer cannot perfectly implement the human mind it cannot perfectly simulate human decisions thus it cannot 100% cogently predict any impact of any act.
    As to how cogent the outcome could be I cannot honestly say that I have read enough into the mind body problem to give a definitive answer but I do feel that the human response, wether in a democratic society or in a dictatorship not one or many human emotion or reaction can be accurately predicted. Obviously only if one holds the position that the mind is incorporeal.

    This bring me to pose the physicalist position.

    If one suggest that the mind is a complex array of human neuronal pathways controlled by external factors then the mind is predetermined thus our actions are predetermined.
    If everything is predetermined one can certainly deduct that everything we do our thoughts, our decisions, our free will is just a complex pattern system.
    If so a computer can predict any actions outcome. Although it is possible, the patterns influencing the outcomes are so complex that until we can decipher this pattern (if human have the capacity to do so) it seems to be in such a distant future I reckon this simulated reality is just a thought experiment for the time being.

    I think it’s exiting to think of such a device and exploring implications is very intriguing.
    This also sparked a thought, I heard (by a third party source but worth researching is) that they have developed ways to stimulate constant pleasure thus eliminating suffering. They tested lab rats and the trials worked to the extent that the rats where experiencing so much pleasure they did not want to eat, mate, have any form of entertainment etc.
    If technology that can achieve constant pleasure can be finalised earlier such a thought experiment seems to be irrelevant.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.