Archives · janvier 2017 · Revue de liens

Veille hebdomadaire : du 31 décembre 2016 au 6 janvier 2017 (revue de liens)

Par Christophe Deschamps · 6 janvier 2017

Ce texte a été publié le 6 janvier 2017. Il est conservé tel quel dans les archives d’Outils Froids : certains liens, outils et informations peuvent ne plus être d’actualité.

Dans cette revue de liens vous trouverez des logiciels, services en ligne et extensions pour navigateurs regroupés sous le terme générique d’ « Outils » et des articles (universitaires, de journaux) et documents divers (présentations, tutoriels, vidéos…) sous le terme « Contenus ».

Vous pouvez retrouver ces découvertes « au fil de l’eau » sur mon compte Twitter ou sur mon profil Diigo (auquel vous pouvez vous abonner en créant votre propre compte ou en utilisant le flux RSS).

Et puis j’ai rajouté une catégorie musique. Parce que.

Bonne(s) lecture(s)/expérimentation(s)/écoute(s)

This is known as Transfer Learning, a field that helps to solve these problems by offering a set of algorithms that identify the areas of knowledge which are “transferable” to the target domain. This broader set of data can then be used to help “train” the model.
Inglehart and Norris find many similarities between the populist rise in different countries; the same effects of economic insecurity in post-industrial economies and a backlash against diversifying societies have driven the same groups of voters to the ballots.
Transfer learning thinking suggests that using the 2016 Brexit voting data from the UK could have allowed statisticians to better understand current global turnout and voting trends. A model that considered data from beyond the U.S. thus might have predicted more support for Trump, especially in demographics that share the same anti-immigration views as was recently seen in the UK.
Transfer learning methods can help the model to overweight the similarities between the U.S. and the German markets, such as population groups that share similar demographic and economic characteristics, and to underweight the dissimilarities. From a business perspective, this will enable decision-makers to simulate the performance of the company in an environment similar to that of the target market.
Instead of the common techniques of solely using the historical data of the same problem for making predictions, political statisticians and business predictors should also start to use data from similar problems occurring more recently — even if they might not be directly connected. To make the connection between the two problems, transfer learning algorithms help focus the learning process on the more relevant parts of the historical training data.

Des fonds de cartes géographiques éditables pour Powerpoint

Rapport de la Maison Blanche sur l’impact des IA sur les inégalités et l’emploi

it does warn that with improper handling, automation could drive further inequality in this already deeply divided country.
In part that increase has been because of a technological fact: that technological innovation, more recently, has helped complement people with higher skills. So we now have a few decades of experience with technology helping to contribute to inequality.”
Three general strategies are suggested for making the inevitable automation of millions of jobs less impactful on the people doing those jobs.
Invest in AI.
“We’ve called for the inclusion of ethics in data science and computer science education to make sure that the technical professionals who are making these decisions are aware of the implications of what they’re doing and are equipped with tools to address these issues,”
Educate and train for the jobs of the future
If the United States fails to improve at educating children and retraining adults with the skills needed in an increasingly AI-driven economy, the country risks leaving millions of Americans behind and losing its position as the global economic leader.
Reinforce the safety net
With a new technology threatening to produce huge numbers of displaced workers, it behooves us to invest in unemployment and healthcare to make sure these people can stay on their feet while finding or training for the next opportunity.
The winner-take-most nature of information technology markets means that only a few may come to dominate markets. If labor productivity increases do not translate into wage increases, then the large economic gains brought about by AI could accrue to a select few.
In other words, if we don’t make sure that AI is working for everybody, you can be damn sure a handful of people are going to make it work for them.

Musique

B.B. King – Blues Boys Tune (From B.B. King – Live at Montreux 1993)

Led Zeppelin – Kashmir – Celebration Day

1 commentaire

thebloom · 6 janvier 2017
Merci de m'avoir cité (2 fois en plus !) dans cette revue web très intéressante