Hacker News Reader

Comment by _ke2e | Hacker News Reader

Where do you draw the line between statistics and ML?

When you can no longer transparently explain the reasoning behind a decision made based on the statistics, and instead have to just shrug and say “because the model told us so”. Accountability is very important in this issue.

Apart from neural networks most models are somewhat transparent in terms of reasoning. Neural networks would be the wrong model to use for this anyway.

For instance, decision trees can tell us if the population density between this range and the percentage of single family residence is between this range, the probability of a package being stolen is X.

Or regression could tell us that the square of the population density is positively correlated to this crime .

This law isn’t talking about banning decision trees or banning regression analysis though, it’s talking about ML being used in specific listed applications.

Exactly. We don’t disagree, I don’t think. Decision trees and regression analysis should count as statistics for the purpose of this issue, as they are transparent and accountable. Neural networks are another beast entirely, in my opinion.

I think they are two different things - statistical modelling to me is about understanding relationships between data (i.e. we are highly confident that X and Y are correlated, and can infer a reason and thus put a policy around that based on open reasoning), and ML is about making predictions in new situations based on historical training data from different situations (but we can't necessarily infer the why).

Reply on news.ycombinator.com