Cliquer sur une vignette pour aller sur Google Books.
Chargement... Algorithmes, la bombe à retardementpar Cathy O'Neil
Chargement...
Inscrivez-vous à LibraryThing pour découvrir si vous aimerez ce livre Actuellement, il n'y a pas de discussions au sujet de ce livre. A powerful explanation of how algorithms and statistical models are used in various aspects of financial and public life in the United States, although some of them have the potential to cause a lot of harm. I don't think there was much in this book that I hadn't read about before, but it was well written and served as a useful reminder. ( ) If data were money, would you let Mark Zuckerberg run for President of the United States? This question jumped out at me while I was finishing Cathy O'Neil's provocative screed "Weapons of Math Destruction." This spring the Zuckerberg family took a Great American Road Trip ostensibly to see how other Americans were faring under technological trends, and presumably under the Trump administration. Zuckerberg himself was quick to dismiss press speculation that the trip was intended to gather data on Zuckerberg's chances in the 2020 US elections. Still the question remains an open one: how "fit" would a Silicon Valley oligarch be considered for the nation's highest office. Based upon my reading of O'Neil's book, the answer is clear: Americans -- and the rest of us (I am not American) -- are rapidly coming under the thumb of big data; that big data's grip is stronger on socially and economically disadvantaged people; and the bigger the data the more valuable the resource. An illustration would do here: an insurance company gathers data on where you drive your car, your propensity for jackrabbit starts at intersections, and how quickly you motor down side streets. The insurance company makes you an offer: manage your car a little more safely and we'll drop your insurance premiums by, say, 6%. In this scenario, providing access to your data -- providing access to your private information -- gives you a clear financial incentive. I know exactly how much money I can save by modifying my driving habits. Here's another scenario: your employer offers you a disincentive to continue smoking, eating pizza five times a week, and guzzling eight beers a day at dinner. Your company-run health insurance premiums will rise and be deducted from your pay. This is the power that big data wields today. Not next week, not next decade. As CEO of Facebook and its gargantuan unique data resources, Zuckerberg is far wealthier than his monetary wealth would suggest. That alone should make him unfit to contest an election. Data is wealth. Big time. But US election laws do not recognize data as having any influence on the outcome of elections even if you and I both know that data is bankable. Sources of money must be documented -- to a degree. Connections with foreign powers must be disclosed. But access to data? Control over algorithms? Nary a word. If we wonder why our political leaders seem to be so weak in the face of economic pressures, social trends, and technology, it's because most of our leaders are data poor. They must beg and pay for data at the trough of the most powerful. O'Neil rightly points out that privacy has a very real price, a price that only wealthy will be able to pay in the future, and even the wealthy participate in the data sharing. Poor people, young people looking for their first job, convicted felons, families looking for financing on their first home, new drivers, high school graduates wanting admission to elite universities, and even school teachers will have no choice. Big data will extract economic rents on them to give them access. O'Neil calls them the new tribes -- may we call them "data tribes"? -- groups who are constructed by algorithms to whom are sold a product, or an idea, or to weed out from exclusive clubs. Even today there are wellness penalties for not participating, teacher metrics, recidivist forecasting, crime forecasting, recruiting screens, and credit scores. Facebook and Google algorithms are massive, powerful, and opaque. Detecting bias in the computer algorithms will be among our greatest challenges in the post-information age. In 2020, the Republicans will have a serious problem with Donald Trump. Commentators have focussed on the trouble Democrats will have unseating Trump, but I think that if Republicans want a primary challenge to Trump they are going to have to go deep or go home. Is a Zuckerberg or Bezos the way they need to go? Well, Peter Thiel backed Trump. O'Neil muses that dumping the electoral college system would be one way to protect the republic and i would have to agree. The electoral college protects the bias built into the US election system toward state interests which are clearly anti-democratic. States shouldn't rule. Should data tribes? aucune critique | ajouter une critique
Appartient à la série éditorialePrix et récompensesDistinctionsListes notables
Qui choisit votre université ? Qui vous accorde un crédit, une assurance, et sélectionne vos professeurs ? Qui influence votre vote aux élections ? Ce sont des formules mathématiques.Ancienne analyste à Wall Street devenue une figure majeure de la lutte contre les dérives des algorithmes, Cathy O'Neil dévoile ces « armes de destruction mathématiques » qui se développent grâce à l'ultra-connexion et leur puissance de calcul exponentielle. Brillante mathématicienne, elle explique avec une simplicité percutante comment les algorithmes font le jeu du profit.Cet ouvrage fait le tour du monde depuis sa parution. Il explore des domaines aussi variés que l'emploi, l'éducation, la politique, nos habitudes de consommation. Nous ne pouvons plus ignorer les dérives croissantes d'une industrie des données qui favorise les inégalités et continue d'échapper à tout contrôle. Voulons-nous que ces formules mathématiques décident à notre place ? C'est un débat essentiel, au cœur de la démocratie. Aucune description trouvée dans une bibliothèque |
Critiques des anciens de LibraryThing en avant-premièreLe livre Weapons of Math Destruction de Cathy O'Neil était disponible sur LibraryThing Early Reviewers. Discussion en coursAucunCouvertures populaires
Google Books — Chargement... GenresClassification décimale de Melvil (CDD)005.7Information Computing and Information Computer programming, programs, data, security DataClassification de la Bibliothèque du CongrèsÉvaluationMoyenne:
Est-ce vous ?Devenez un(e) auteur LibraryThing. |