[timb07]

> We can't predict the future.

What tosh. Of course we can predict the future... and sometimes our prediction is correct.

Data compression works by modelling the data stream and predicting the next symbol based on that model, and then recording in an efficient way the degree to which the prediction was correct. The comparison between the prediction and the actual outcome is the key point -- just as it is in the article.

[s-shellfish]

Eventually the system you use to model information with will override actual data. That's a feedback loop of confirmation bias. It's thinking in terms of time. How many models of data do you interact with versus how much actual data are you interacting with. I'm not on a soapbox without a point. It's a fine line. Your own mind matters in these systems. It's important to not undervalue it. It's not a computer. It's not a better computer. It's sometimes flawed.

I'm fine with calling it probabalistic inferencing. It's not future prediction, especially when the models one is reasoning about are models of cognition.

Thank you for not down voting though, that's appreciated.