4 ms·
In many cases, you may want to optimize one of the inputs to achieve the desired outcome. If you think of classic linear regression, you get a rough approximati
by adamhp 4y ago
In many cases, you may want to optimize one of the inputs to achieve the desired outcome. If you think of classic linear regression, you get a rough approximation of how each input affects the output. So, for example, if you built a linear regression model on house pricing, you could be relatively confident that an X increase in number of bedrooms, or proximity to public transport, or some other variable, has a direct affect on your output Y (home price). And in the case of linear regression, you get weights and power for each coefficient, so as a builder you can say "I'm going to build houses closer to the metro to fetch a higher price". Greatly simplified, but I think this is is what is missing from AI. If we're having AI classifying things, it's only one piece of the puzzle to actually do the work of classifying (and that's useful to a lot of people). But the other half of the equation is to explain why things are classified the way they are.
- Perihermeneias 4y agoThanks for the thoughtful response. What would this look like in something more complicated? In deep learning would this be something like “if you can get these neurons to fire more you can get a better output” and “you can increase these neurons firing by changing the inputs in this way”? Also, I feel like this sounds a lot like generative AI, and explainability might not be necessary. Or maybe one could say that in the real world people want to have better assurance that things they do are based on a good underlying model. Images and text are easier to evaluate than predictive, speculative decisions about housing markets.
- adamhp 4y agoCorrect. The latter point is important, especially in heavily regulated industries, or the public sector. You can't just tell the public you decided to do X because an AI told you so.