“Amazon Mechanical Turk (AMT) offers a relatively low-cost alternative to traditional expensive survey samples, which likely explains its popularity among survey researchers. An important question about using such samples is whether they are representative of the larger Internet user population. Though prior research has addressed this question about demographic characteristics, little work has examined how AMT workers compare with others regarding their online activities—namely, social media experiences and online active engagement. This article analyzes survey data administered concurrently on an AMT and a national sample of U.S. adults to show that AMT workers are significantly more likely to use numerous social media, from Twitter to Pinterest and Reddit, as well as have significantly more experiences contributing their own online content, from posting videos to participating in various online forums and signing online petitions. The article discusses the implications of these findings for research that uses AMT as a sampling frame when examining questions related to social media use and active online engagement.”
«Tu vois, aujourd’hui, les chaînes de montage de voitures automatisées ? Avant, c’était des gens qui le faisaient, ben je suis un peu cet exécutant-là sur la chaîne, qui un jour sera remplacé par un robot. Sauf que ça sera un robot d’intelligence artificielle. Et ce que je fais moi d’ailleurs, c’est du travail à la chaîne. C’est là qu’ils ont été intelligents chez Google».
Via The Conversation
Machine Learning (ML) is a fast-moving, competitive field. As important as good algorithms are to ML, the state-of-the-art algorithms are reliably available. Compute clusters are also an important ingredient, and they too are easy to access (especially using services like Amazon EC2 and Amazon EC2 Elastic GPUs). What isn’t reliably available are large-scale high quality data sets. The things you need to train your classifiers. That’s where MTurk comes into play.