/* ---- Google Analytics Code Below */

Friday, January 24, 2020

Putting Humans in the AI Loop

I argue that they are always in the loop.  Just how effectively are they placed there. Always a very useful question. Starting with existing process helps.

Can AI Put Humans Back in the Loop?
Tiernan Ray

Scientists at Germany's Technische Universitat Darmstadt have developed a process for having a human domain expert review an artificial intelligence model's inner mechanisms during training, in order to catch simple problems and correct errors. Such an expert would check the reasoning offered by a neural network, with the overall goal of building more trust in machine learning. The experimental explanatory interactive learning procedure involves a convolutional neural network classifying the phenotype of a plant as healthy or diseased by analyzing leaf images. The researchers visualize the features the network is using, then a plant biology specialist fixes any network errors. ... 

No comments: