Hybrid intelligence learns to predict unpredictable / SurprizingFacts

Race of artificial intelligence has begun: neural networks are mastering more and more unexpected and interesting tasks that they could solve, and more companies and start-ups are trying to bring them to a qualitatively new level, behind which lies Passing the Turing test and the singularity. While some lavish generous promises for those who dream of everyone around ("The first trillionaire is the one who deals with AI"), others are terrified by the singularity and unpredictability of relations between humanity and AI. However, both of them argue within the framework of one paradigm: AI is inevitable – the only question is whether it will bring us ills, prosperity or a complex mixture of both, then in what proportions?


But there is a third way, an alternative to the world of uncontested superiority of AI, in which we surrender to him at the mercy – hoping that this, and truth, will be mercy. This is the possibility of the human brain itself. In the article "People do not even understand how they became cyborgs," I outlined the idea theoretically: neuron interfaces, which are being developed with equal success, but with less aplomb than the development of neural networks, can lead man and humanity as a whole to a new level of intellectual power that Is quite capable of postponing for us the moment when we will turn into pets at the careful care of robots and Skynet, and prolong our time when we can still afford to feel ourselves crowns of creation – now we have our own And not evolution, gods or Engineers.

2017 already deserved a place in the future description of the history of human cyborgsation, even if by the fact that Ilon Mask, a man whose ideas always become the number one novelty in the technology world, launched Neuralink, which should weave a "neurocrust" implantable electronic shell of the brain. Electric cars, autopilots, Mars exploration or new public transport technologies – Mask stands for the main trends of progress one after another. But this does not mean that before the mask there was no life – just Ilon can disperse any topic he drew attention to by one fact of getting into the orbit of his most powerful media influence.

So, the team of Russian mathematicians and traders Cindicator suddenly found itself at the forefront of progress: the guys just did their job until the launch of Neuralink did not indicate the unification of the properties of the human brain with the computing power and communication capabilities of computers as one of the top trends. And Cindicator, suddenly for himself, was among the leaders of this trend.

What does the Cindicator do?

Cindicator is a platform that combines the capabilities of the collective mind in the most direct sense (the so-called "wisdom of the crowd") with the analytical power of computer systems to predict future events, primarily in the financial sector.

The wisdom of the crowd is a repeatedly observed and confirmed effect, when the average of the guessing results by any large number of people turns out to be close to the correct answer. Do you remember how Raymond Babbitt, played by Dustin Hoffman in the movie "Rain Man", at a glance exactly counted the number of toothpick scattered by the waitress? A normal person is not able to do this, but if a hundred people record their estimates made at a glance, then the average arithmetic is likely to turn out to be close to the truth.

If you go completely from afar, then any mass will expressed correctly is the expression of the collective unconscious that can be cast in granite: the voice of the people is the voice of God.

But to go beyond the amusing laboratory experiments and find a practical application of the "wisdom of the crowd" has so far been unsuccessful. Firstly, for understandable logistical reasons: the technical possibility to involve an unlimited number of people in solving problems in this way arose relatively recently. It was the boom of smartphones and the mobile Internet that made the crowd as accessible as possible for those who are interested in its "wisdom" – and the process of guessing itself made it possible to give the game form.

Secondly, in accordance with the mathematical principle "the correct formulation of the problem Contains half the solution, "one must be able to raise the question before the collective mind – so that it does not work out like in the" Hitchhiker's Guide to the Galaxy ":

The answer to the main question of life, the universe And all this: after careful reflection, I can with azat – a 42

Research wisdom of crowds say that the best use of it is for the problem to a specific unambiguous answer – math or geography. In other words, places or numbers. And since the place is just a point in the coordinate system, it all comes down to numbers. Paradox: very few people in the school liked mathematics – but collectively we turn out to be good accountants. I wonder what will happen if the schoolchildren start taking exams, voting for the various answers by the whole class?

And thanks to the "connection" to the collective intelligence through the mobile application, Cindicator achieves full decentralization of its collective mind, which is one From the fundamental conditions – the members of the group should avoid any communication and exchange of opinions within the collective, in order to avoid the influence of some individuals on others.

How to make money on the forecasts

When the character of the movie "Back to the Future" decided to earn some money on knowledge of the future, he decided to earn on sports betting. Cindicator, however, choosing the sphere of application of its efforts, started with the financial market, which gave the most promising feedback: there was also an acute, everyday need for predictions (what to say, when some financial instruments themselves say that they require Nostradamus) And clear feedback, which allows us to assess the quality of our forecasts – and, of course, the ability to effectively monetize the data.

Strengthening the team of mathematicians and data analysts with experienced traders with serious experience managing Cindicator began to grope "a special way" in the brutal world of financial trading.

Practical application of the wisdom of the crowd in trading in stock and currency markets in reality looks like A little more difficult than daily attempts to collectively guess tomorrow's dollar rate. Cindicator constantly tests various types of questions, data, new trading strategies and models to optimize the use of the data obtained to create trading robots and analytical products (indices, price levels, sentiments, etc.);

In addition to working with the results of the crowd forecasts, Cindicator tunes the predictor machine itself: as the Cindicator experience shows, fine tuning is quite capable of overclocking the efficiency of its engine at times. One example of this is a joint experiment with the Moscow stock exchange, during which the trading robot Cindicator made transactions for three weeks based on the collective opinion of 863 application users, of whom 40% had never traded on the stock exchange. In total, during the experiment, the robot modeled 27 transactions, 17 of which were profitable. As a result, during the contest, the model portfolio grew by 2.8% in currency (47% per annum).

To improve its performance, Cindicator electronics need to analyze not only the accuracy of each collective forecast, but the efficiency of each of thousands and Tens of thousands of forecasters, its strengths and weaknesses are the same big data, which has been talked about so much in recent years, but barely learned to approach it.

The work of Cindicator with large data includes a system and methods for determining Weight of the forecast of each user Based on a personal time sheet for different types of questions (decision time, forecast history for each type of investment instrument, the relationship between the answers, the type of analysis method used, etc.), dynamic feedback after each transaction, taking into account both The fact of profit or loss, and its size according to the results of each of its forecasts.

A dream of reason gives rise to monsters. Analysis of collective intelligence – superforkaster

Thus, the computing power of Cindicator not only tunes the machine itself of collective predictions. She singles out the individuality among the crowd. At first glance, this may seem to run counter to the very idea of ​​crowd wisdom, when a large amount of noise, errors, subjectivism is compensated by its heterogeneity and multidirectionality (Gauss distribution is the simplest example), but the Cindicator practice shows that a small dirigism performed with a high mathematical Accuracy, allows not only to adjust the spontaneous force of collective predictions, but also, which can not be expected in advance – to allocate people who are consistently in fact strong forecasters themselves By themselves – superforkasterov.

Superforkasters – superheroes of the real world

Until recently, the phenomenon of "wisdom of the crowd" remained, in fact, deeply unexplored. Experiments confirming its presence have been conducted sufficiently. But there were simply no computational and organizational capacities to dig deeper.

Cindicator is not a research project, but the side effect of the practical application of "wisdom of the crowd" is just a deeper understanding of the principles of this phenomenon – and new discoveries. As soon as the platform was launched, the confirmation of the theory of superforcasters (super-forecasters) – people who in most cases make more accurate forecasts for certain assets and issues than others. The share of superforcasters among all users of Cindicator is about 2%. Further analysis showed that they are also heterogeneous – some superfoccaster are universal as forecasters; in others, the accuracy above the median is manifested in the prediction of only a certain type of events. The reasons for both the uniqueness of these 2% and the specialization of some of them are still to be determined – it may be natural talent, and the acquired competence – in the previously mentioned experiment with the Moscow Stock Exchange, the author of the most accurate predictions became an investor with 10 years of experience. However, this does not depreciate the very idea: his personal trading result would be + 1.48% for three weeks (25% per annum), which is almost twice lower than the yield provided by the collective wisdom of Cindicator forkasters (2.8% in currency for 3 Weeks / 47% per annum.)

In a comedy show is often found geg, when the wife-housewife guesses the results of sports matches better than her husband-fan. In real life, the most unexpected people can be the bearers of talent to predict exchange rates or weather conditions of the future without even knowing it – as well as about what they could achieve by using this superpower.

Next, to Noosphere

Corrected by the machines, the wisdom of the crowd Cindicator calls "hybrid intelligence" (GI). The project team is not even right to call developers – because they are developing only part of the system. Its main force, literally – is alive. So Cindicator would be more suitable definition of managers of hybrid intelligence. Located at the very beginning of the road to its development, study, domestication.

If you combine the knowledge of the Cindicator about the wisdom of the crowd and the experience in directing the spontaneous power of collective intelligence to solve applied problems with the same Neuralink developments, if Ilona Mask will be accompanied with them with the same success as with electromobiles, the hybrid intelligence Cindicator, in combination With a neural interface, can even provide people with a breakthrough to a new technological, intellectual, social level, which until then was associated exclusively with the invention of AI.

The physical (more precisely, bioelectronic) interface created by Neuralink will require software developments that allow you to directly realize the opportunities that have appeared – just like a modern smartphone without a rich set of programs, it turns out to be nothing more than an expensive flashlight. The development of iron is more complicated and expensive, but the winner is, in the end, the one with the best software – let there be an example of a never-soaring Windows Phone as an eternal reminder of this.

By deploying its technologies on the Neuralink platform or any other similar project , Which will come to the finish line before, Cindicator will be able to overcome the boundary of convention in the definition of "hybrid intelligence" and create a hybrid of the human mind with the computing power of machines without any quotes. In fact, this will mean a transition to the noosphere of Vernadsky. And so far, from all the alternatives available to us, it is the possibilities of the human mind, overclocked by integration with the electronic brains, combined, perhaps, into some kind of super-mind – they seem to be the best protection against AI threats, real or imaginary. At the very least, this will preserve the possibility of the human mind to cognize reality – postponing for some time the onset of singularity – that is, the moment when the speed and complexity of the development of technology will outstrip a person's ability to realize what is happening. Hybrid intelligence is still our best chance to preserve our place at the top of the food chain.