This article is from the public number:L Mr. said (ID:lxianshengmiao), author: Lachel, title figure from pixabay

Liu Cixin has a short story called “Mirror” that tells a story:

A software engineer at the Weather Bureau accidentally developed a simulation software using a superstring computer to simulate the process of a universe from birth to evolution.

He experimented with more than a thousand universes, and finally, one night, inadvertently discovered a set of data, using the universe it generated, all physical constants are exactly the same as our universe. In other words: he simulated the universe we are in now.

So, he saw the buried “real history”: Troy has never existed, Marco Polo has passed China from the future, and many unknown little people are actually true heroes…

Friends who are interested can go to the novel on their own, very short.

But in Mirror, even this versatile cosmic simulator has a fatal flaw: It can only be used to look at the past and cannot be used to predict the future, otherwise it will fall into “infinite recursion” Among them.

Predicting the future is probably the limit of human imagination.

Either Marvel’s singular Dr., or the Doctor of Manhattan in The Watchman, once it is related to the future, it will always bring complex thinking traps and paradoxes.

If I can predict in the future, can I change it? Can I predict the future that I have changed?

If the future can be foreseen, does that mean that the whole world is essentially deterministic? Has everything been destined for a long time from a distant past?

This may be a forbidden place for human beings, the last thing you want to get involved in.:Because it will have a huge impact and impact on our “free will.”

1

But everything is going to be a reality.

A technique called ABM (Agent-based Model) may really enable us to “predict the future “.

What is its principle? For a simple example:

Imagine an open space with 10 dots on it. We set the following instructions for them:

1) Walk randomly on this open space and cannot walk outside;

2) When two dots meet, there is a 50% chance that they will “align” and 50% chance will attack each other;

3) When attacking, there is a 50% chance to win between every two dots.

So, will there be a few dots that will “survive” in the end?

We can also add rules. For example, some dots have “brave” attributes, and the attack has a higher winning rate; some have “reconnaissance” attributes that can avoid danger; some have “friendly” attributes, and the chances of alliance are higher… and so on.

This is an extremely simple example. Just take a computer and write a few lines of code to solve it.

We let it run a few times and do a few simulations to get different results. By making a large number of these results, you can know the general tendency.

But if you change this open space into a city; replace these dots with one “person”; replace these instructions with some simple rules of conduct that govern humanity?

In a way, we can predict the general trend of a small group by simulation.

If you’ve played a game like “The Sims,” ​​you’ll be very clear.

Of course, the principle of “simulation” is not new – it has been in use since the last century. The weather forecast is like this.

Similarly, there have been many cases of simulations of human society.

For example, the Ebola outbreak broke out in West Africa in 2014. At that time, the relevant experts spent seven months building a model that introduced real-world data such as the number of cases, infection rate, mode of transportation, population distribution… and finally concluded that if uncontrolled, it may 1.4 million people will be infected.

Finally, under the guidance of this predictive model, the medical team went to the core of the intervention to successfully control the number of infections within 28,000.

But whether it’s weather forecast, Ebola simulation, or The Sims, there is a serious problem: The amount of data is too small to simulate “reality.”

For example, in the Ebola simulation, each “person” has only a few behaviors: being infected, running away, and acting randomly. Rather than being a living “person”, it is better to say that it is a “small point” that is highly simplified and abstracted.

This has led to the stagnation of ABM technology development. After all, to restore the real world, the amount of data needed is too large, and it is not something that can be achieved in a short time.

2

However, in recent years, the development of a technology has made it possible to “simulate real society.”

What is this technology? You must also be very clear, it is artificial intelligence – The development of artificial intelligence, solving the “input” problem for data.

What do you mean?

What is the hardest part of the old analog technology? If you want to restore a group as much as possible, you have to enter enough data for everyone. Including: birth, family background, preferences, habits, work, skills, financial situation, and so on.

This is almost impossible.

But for artificial intelligence, this is only a minor issue.

Artificial intelligence has a very powerful ability called self-learning: you can learn, self-test, self-correct, and self-improve by giving it a certain amount of data.

When it “learns” a pattern, you can really “simulate reality.”

For example, AlphaGo is very familiar to everyone. In 2017, AlphaGo Zero does not require any knowledge input. Just give it a board, tell it the rules, let it play against itself, and only train for three days, it can beat AlphaGo.

Similarly, the VQ-VAE model released by DeepMind only needs to “feed” 70,000 face images as a training set, which can generate a clear “face” of 1 million pixels, which is completely indistinguishable to the naked eye.

In life, we generate data every day. These constant data are the best training sets for artificial intelligence.

Enter them into artificial intelligence, let artificial intelligence learn the “human” model, it can easily generate a “real person.”

These people, like real people, have personality, faith, family, and work. They will get sick, grow old, interact, learn, and respond to external stimuli.

The difference is that they are only in the computer and not in reality.

This is MAAI, multi-agent artificial intelligence 3

This is a very new technology.

What is the new level? Nigel Gilbert, director of the Center for Social Simulation Research at the University of Surrey in the United Kingdom, said that some policy analysts have noticed it, but most politicians know nothing about it.

However, it does not only exist in the paper.

According to Saiku Diallo of the Virginia Modeling Center, “The technology is now able to accommodate tens of millions of ‘agents.’” In short, “we can model a city like London” .

Imagine: Building a city in your computer with tens of millions of “residents.” They are exactly the same as real people. Everyone has detailed births, resumes, personalities, traits… Although they have no self-awareness, from the perspective of the observer, they are just like real people.

In theory, when we put some of London’s resident data into artificial intelligence, use it to generate enough “agents” and put them into a similar environment, this constitutes a virtual “London 2” .

It is very similar to the real London, and it is not too much to say that it is a simplified version.

Then, we can try to control the external conditions to observe the changes in the entire “London 2”.

Example:

  • Implement a new policy to observe changes and trends across the city;

  • Initiating a disaster to see if it can be effectively curbed and controlled;

  • Enter some unexpected events, such as “Brexit” and “Refugee Influx” to observe long-term trends.

As long as you let it run a few more times, we can get enough data to have a bigger grasp of the future.

In a way, this is not to be “predicting the future.”

And the development of this technology is likely to be faster than we think.

It is currently London, what’s next?

F. LeRon Schults said the next goal would be 320 million- the population of the entire United States;

The next goal, 1.4 billion, China.

In the end, the world.

4

In reality, such a model is running.

According to NewScientist’s report: A model that simulates a Norwegian city is trying to solve the refugee problem.

In the past eight years, 1 million Syrian refugees have fled to Europe, and 20,000 of them have settled in Norway. What impact will these refugees have on the local social environment? What policy should I take to deal with it? How can we minimize risks and damage? no one knows.

Thus, Schults is leading a model that simulates a possible future for a Norwegian city under the impact of foreign refugees.

This model is small, with only 50,000 agents, but will simulate the time of up to 3 generations to observe the consequences of various measures over the longer term.

A model like this volume takes a few hours to several days to run.

You must have a question: Does the modeling “generated” from real data really represent the reality? How high is its fit?

This, of course, modeling scientists have also considered. In Schults’ Norwegian refugee model, they spent a year setting up initial conditions, modeling, and conducting real-world tests.

The research team used the Norwegian government’s social survey data to observe the results of the modeling and whether the social survey data matched. If not, fine-tune the initial parameters and re-model.

What is the key? These are all one-off jobs. Once you pass the reality test, your modeling will be successful. You have a “sandbox” that is big enough to experiment freely and freely, and correct it based on the feedback you get.

In addition to the Norwegian refugee model, other applications have surfaced.

British economists are considering introducing MAAI technology into economics to better understand real-world economic changes, and predicting like 08The impact of the financial crisis.

A team at the Boston Center for Psychological and Cultural Affairs has established a MAAI model to deal with criminals who are being trafficked by children. But according to the group’s director, criminals are likely to be adopting similar techniques to counter government actions.

In the near future, many scholars are concerned about the 2020 US election. Will the teams on both sides use the technology to experiment with various campaign strategies to maximize the voter’s vote? If so, how do you define this behavior?

It cannot be said that this will happen, but at present, the possibility is not small.

5

If this technology can be popularized in the near future, the impact will be very large.

I have mentioned in this set of “creator thinking” that I have benefited a lot: In modern society, everything is highly sophisticated, methodical, fully industrialized and streamlined. Everything you see and perceive is the product that someone presents to you after processing.

With the blessing of this technology, this phenomenon will only intensify.

What you perceive is not only the products that others have presented to you, but also the “optimal solutions” that others have undergone numerous trials and exercises. Every movement, every rhythm, is just right, appears at the most appropriate time, and falls in your most comfortable place.

This is a bit like “reading the heart”.

Of course, from another perspective, this technology will also bring a huge positive result: The uncertainty we face will be reduced to a small extent.

Either an emergency, a decision, a crisis, with the help of MAAI, is equivalent to countless “restarts” – risk will be minimized, all our possible reactions They are already anticipated by others and have some plans.

We will live in a “more comfortable” and “safer” society.

As for how to treat this issue, it is left to everyone’s thinking and discussion.

6

About this technology, you can actually extend it.

Many people know that the randomness of quantum mechanics destroys determinism. But in fact, the randomness of quantum mechanics is only manifested at the micro level. When it shifts from the microworld to the macro world, this randomness does not exist—everything becomes a sure, controllable, and stable thing.

Therefore, our existence is stable, not suddenly “collapsed” into a pile of particles, and will not suddenly appear on the moon.

And human society is the same.

Everyone is extremely complex, and without enough information, it is almost impossible for us to “predict” a person’s behavior. But when a group of people form a group, it is much simpler to predict the behavior patterns of this group.

Why? Because the group has a feature: it will erase all “outliers”. Thus, stable features and patterns are presented.

Simply: We say that a drug is effective for a disease. Does any person who has this disease apply to this drug? No.

Everyone’s situation is different. Some people’s symptoms may not apply to this drug, but putting everyone together, these become “cases” that will be overshadowed and erased by larger data. . As a result, on the whole, it has shown “effectiveness.”

Similarly, put together 10,000 very smart people, and the decisions and actions they make together must be better than the results of any one of them? Not necessarily. Unless they are absolutely rational, they may attack each other, gain benefits, reach compromises, and so on. What is finally presented is probably just an average.

This is one of the cornerstones of MAAI: we can’t “predict” a person’s future because there is too much uncertainty and too much data to consider; but when so many people form a group The data we need is reduced, but the uncertainty we face is reduced.

Further, as I said before: thinking is an entropy increase. Do you want every cell in your body to have independent thinking skills? Certainly not. right?

Questions about “groups” and “individuals” may also be an issue that needs to be considered.

This is also what I hope to discuss and think with you.

7

Finally, a few more words.

The breakthrough of technology is coming, in the face of the wheel of history, praise or opposition, most of the time, the meaning is not great.

A lot of things, once created, may not be able to be transferred by human will.

What we have to do is to be mentally prepared to meet a change that may seem undulating but may change society and change human civilization.

It may come in a few decades. Maybe never will come.

And maybe, just after.

This article is from the public number:L Mr. said (ID:lxianshengmiao), author: Lachel