= 27d567bc2d48d9f40e9ccc20ea370b15)
So what kind of metrics are you interested in forecasting? Macroeconomic ones (GDP, inflation, etc.)? Industry-specific things? Interest rates?
We use mathematical models. It does't matter that data use. The issue is that it will give. In any case, interest is the machine that predict the future, wider than those statistical model that people used in the financial markets. I will write an article about my thoughts. If karma will be great.
in the economic environment
Do you mean financial markets?
I'm talking about the economy as a whole, it has a wider meaning and includes not only the financial market but as well as a variety of industry.
on the basis of these 10 future variants, algorithm simulating new probability for 2nd year, then on the following 3 years and other.
That's called chaining the forecasts. This tends to break down after very few iterations because errors snowball and because tail events do happen.
Yes, it's true. It is very difficult to build a forecast of non-automated way. Apparently need to take care of the right frame algorithm and gradually increase it, such as a inventions tree or matrix, for example. Then add the probability, time and further, to increase the amount of data. But what do we when the forecast crumbles in the future? Theoretically, even we having enough data, it's possible to come across a large number of bifurcation points that's just create too much a parallel universes with different futures. What if it does not matter that the forecast is correct. Importance their infinite number, from which you can choose the right future, with the help of special algorithms or be prepared for multiple outcomes.
In response to
Open thread, Jul. 04 - Jul. 10, 2016
AI which predicts the future based on an non-existing past. Imagine an algorithm that is able to predict the future in some of direction (social, industry, etc.), and assume that we have a necessary data for this. For example the algorithm creates a different probability model based on these data, modeling the future to 1 year in advance. Let's say what we’ll have created 20 models, including 10 models that have a true probability of 85%. What if on the basis of these 10 future variants, algorithm simulating new probability for 2nd year, then on the following 3 years and other. The philosophy of the system is that the algorithm lives in the present but have accessing data from the past to the simulation of the future, which is created by the analysis of the data has not yet come last. If he has a kind of infinite power, it is able to predict the future for thousands of years in advance, and have the experience even non-existents events.
So, are there any researches of this subject?
In response to
Welcome to Less Wrong! (8th thread, July 2015)
Hello everyone
I represent a small team whose goal is predicting the future based on artificial intelligence in the economic environment. Some time ago I was looking for a place where I could be useful with our arguments and this community is seems to be the best place for it.
I hope to be useful, to discuss if we missed out something, to test new ideas with a competent audience.
The essence of my questions will be: management system based on artificial intelligence; risk prediction using parsing systems; quantum models predict the likelihood; ideas of the frame prediction systems for various events.
Thank you for attention.
= b715d02e85f7b3b4ae65ca3d3e450868)
Subscribe to RSS Feed
= f037147d6e6c911a85753b9abdedda8d){kind=logo}
The right algorithm doesn't give you good results if the data which you have isn't good enough.
What do you mean?