Using a computable approximation of Solomonoff induction (notAIXI, that’s different!) is not some kind of option that can be avoided—modulo some comments about the true razor.
You can warn about its dangers—but we will plunge in anyway.
Ah, I have no idea why I said AIXI. Must have gotten my wires crossed. :|
This seems to leave open the question of what approximation to use, which is essentially the same question posed by the original post. In the real world, for practical purposes, what do you actually use?
The question is: please recommend a model of rationality that a human can actually use in the real world. It’s not clear to me in practice how I would use, say, gzip to help make predictions.
Right, well, the link between forecasting and compression was gone over in this previously-supplied link.
See also, the other introductory material on that site:
That is indeed a correct answer to a reasonable interpretation of the question I asked. I thereby realize that I should have asked differently.
Where examples of rationality usage are given on LW, they tend to be of the straightforward decision-theoretic kind, such as solving the trolley problem; that is, rationality and studied and taught here is mostly about helping humans better make the kinds of decisions that tend to be made by humans.
Suppose I want to take an umbrella to work with me if and only if it will rain this afternoon. How might I go about deciding whether to take my umbrella? And, in particular, is running my own statistical analysis on the weather patterns in my local area over the past hundred years really a better choice than just turning on the weather channel?
And, in particular, is running my own statistical analysis on the weather patterns in my local area over the past hundred years really a better choice than just turning on the weather channel?
Perhaps I am missing something, but the answer is obviously no. This follows from the usual humility and outside view arguments, and from more detailed inside view considerations like the following: the weather station has access to far more data than you over that time period, and has detailed recent data you do not, and can hire a weather statistics expert (or draw on such expertise) who will crush your predictions because they specialize in such problems.
The answer was intended to be obviously no. I wished to refute the idea that esoteric mathematical models like prediction-as-data-compression translated directly into useful advice for the real world outside of a few highly technical cases.
Using a computable approximation of Solomonoff induction (not AIXI, that’s different!) is not some kind of option that can be avoided—modulo some comments about the true razor.
You can warn about its dangers—but we will plunge in anyway.
Ah, I have no idea why I said AIXI. Must have gotten my wires crossed. :|
This seems to leave open the question of what approximation to use, which is essentially the same question posed by the original post. In the real world, for practical purposes, what do you actually use?
Making a computable approximation Solomonoff induction that can be used repeatedly is essentially the same problem as building a stream compressor.
There is quite a bit of existing work on that problem—and it is one of my current projects.
Okay, but what’s the actual answer?
I don’t understand the question. Can you explain what was wrong with the answer I just gave?
The question is: please recommend a model of rationality that a human can actually use in the real world. It’s not clear to me in practice how I would use, say, gzip to help make predictions.
Right, well, the link between forecasting and compression was gone over in this previously-supplied link. See also, the other introductory material on that site:
http://matchingpennies.com/machine_forecasting/
http://matchingpennies.com/introduction/
http://matchingpennies.com/sequence_prediction/
If you want to hear something similar from someone else, perhaps try:
http://www.mattmahoney.net/dc/rationale.html
I understand the theoretical connection. I want a real-world example of how this theoretical result could be applied.
An example of prediction using compression?
E.g. see Dasher. It uses prediction by partial matching.
I also found this thesis, ‘Statistical Inference through Data Compression’, using gzip of all things, quite interesting. (Some half-related background.)
That is indeed a correct answer to a reasonable interpretation of the question I asked. I thereby realize that I should have asked differently.
Where examples of rationality usage are given on LW, they tend to be of the straightforward decision-theoretic kind, such as solving the trolley problem; that is, rationality and studied and taught here is mostly about helping humans better make the kinds of decisions that tend to be made by humans.
Suppose I want to take an umbrella to work with me if and only if it will rain this afternoon. How might I go about deciding whether to take my umbrella? And, in particular, is running my own statistical analysis on the weather patterns in my local area over the past hundred years really a better choice than just turning on the weather channel?
Perhaps I am missing something, but the answer is obviously no. This follows from the usual humility and outside view arguments, and from more detailed inside view considerations like the following: the weather station has access to far more data than you over that time period, and has detailed recent data you do not, and can hire a weather statistics expert (or draw on such expertise) who will crush your predictions because they specialize in such problems.
The answer was intended to be obviously no. I wished to refute the idea that esoteric mathematical models like prediction-as-data-compression translated directly into useful advice for the real world outside of a few highly technical cases.