In the past, countries would have narratives to guide public opinion on things like war, famine, natural disasters, and in general major historic events that may or may not have been easy to anticipate. There's thousands of years of history where governments manipulated the facts, edited history books, changed borders on maps, and so on. A fun example of this is the Veritable Records of the Joseon Dynasty, in which a ruler fell off his horse and told them not to add it to the records. Less fun of a fact is that a prior ruler executed one recorder and five other people for what was written in the records, which is why they tried to never again let a ruler see the true records.

If we were to take all of the records of the entire planet, compare them to each other, figure out which ones are more accurate and which were modifications meant to keep the authorities appeased, then we could have not only a more accurate world history, but also a record of every way governments have ever edited the facts, the ways they control the narrative of any world events, and in the end have a system through which global news reports can be added, decoded, and translated back into the probable truth, the likely reasons for the lies, the kinds of lies used, and what it means when a government needs to lie in that specific way.

Once all of that is done, and we have the truth as far as the government understands it, we can then develop another system that takes both sets of information, lies and truths, and compares it against real world events months and years down the line. This can be used to predict future events, whether it be that all governments agree to ignore the truth in favor of a beneficial arrangement (more about strategy than facts), or a disaster that could have been prevented (facts no amount of strategy can avoid).

Like, if a government tears down a bunch of forests for farmland, and thirty years prior those forests were planted so mudslides caused by over-farming would stop, then we can safely assume there will be mudslides in the future, right?

Doesn't take a genius to put these kinds of things together, but you do need to know all of the worlds history in order to connect the dots, which is hard to have every single human do, but it's a lot easier to make an AI that can do this for everyone.

So that's what I think we're going to see eventually, the develop of an AI system that removes all the social and political fluff, and just describes the facts based on past records, likely trajectories in global markets, weather patterns, and so on.

If this works how I think it will work, you could consult this system on whether you should buy a specific house, and it can tell you the likelihood that it'll be torn down within the next century due to any number of reasons. Maybe the house is on the best route for a highway and may be a poor choice in 20 years. Maybe it's the ideal location for wealthy people to live and is just waiting for the land value to rise into unaffordability for the people living there. Maybe the city has been kicking cans down the road and one of them is its lead water pipes.

Alternatively, is there a person you could personally go to and ask for answers to these kinds of questions? You could maybe talk to a historian about old laws, but they may not know enough about regional housing to help you there. You could talk to a librarian, but they may only point you to the books that could help, they haven't read every book in existence and memorized the relevant information. You could do all the work yourself, but it'd be unpaid, and if you neglect to publish your findings then every other person in the world would need to re-tread the same path, assuming you all got the correct, unbiased information through even more work.

There's two ways this gets created. Either some central authority controls it, and can edit what it outputs so they still control the narrative, or it's decentralized and everyone has a copy of it locally, which makes it harder to control the narrative if people don't constantly update it, making it easier to compare the quality of its guesses against new models. An easy test would be having it predict tomorrows news, and seeing which model performs better.