Open Password – Monday April 26, 2021
#914
Digital Corporations – Global South – Michael Kwent – ROAR Magazine – Digital Capitalism – Former Capitalism – Colonization of Education – Labor Exploitation – Wages and Working Conditions – USA – China – Democratic Republic of Congo – Market Leaders – Political Domination and the Means of Violence – Grassroots Movements – Banana Republics – Emerging countries – Liberal public – Mass protests – Code of ethics – European Commission – UN – Machine learning – Image similarity search – Artificial intelligence – Historical library holdings – Klaus Kempf – Markus Brantl – Thomas Meiers – Thomas Wolf – Bavarian State Library – Upload function – User experiences – More general Viewer – Upload option – Weighting of search parameters – CPU cores – Open source – International Image Interoperability Frameworks – Named entity recognition and search – Munich Digitization Center – Streaming – Usage – Civey – E.ON Energy Atlas – Videos – Podcasts – Games – Use of social media – Facebook – Instagram – Facts Office – Toluna – Age groups
1.
Cover story: Artificial intelligence opens up historical library holdings – upload function – limits of image similarity search – conclusion and perspectives
2.
Outside the box: Discuss the influence of digital companies on the “Global South”? There is still no adequate frame of reference for this
3.
Streaming – Use of social media
Outside the box (42)
Discuss the influence of digital companies on the “Global South”? There is still no
adequate frame of reference for this.
Michael Kwet, Digital colonialism: the evolution of American empire, in: ROAR Magazine, “an independent journal of the radical imagination providing grassroots perspectives from the frontlines of the global struggle for real democracy”, http://roarmag.org, March 2021 The author prefaces his article with a short conclusion: “American “Big Tech” corporations are gaining massive profits through their control over business, labor, social media and entertainment in the Global South.” Regarding the relevance of his question, he notes: “While “The expanded dominance of these corporations in their domestic markets is the subject of numerous critical analyses, their global reach is seldom discussed, especially by dominant intellectuals in the American empire.”
To clarify his basic concept of “digital colonialism,” Kwet contrasts it with earlier colonialism: “Under classic colonialism, European seized and settled foreign land; installed infrastructure like military forts, sea ports and railways; deployed gunboats for economic penetration and military conquest; constructed heavy machinery and exploited labor to extract raw material; erected panioptic structures to police workers; … siphoned out Indigenous knowledge for manufacturing processes; shipped the raw materials back to the mother country for the production of manufactured goods; undermined Global South markets with cheap manufactured goods; perpetuated dependency of peoples and nations in the Global South in an unequal global division of labor; and expanded market, diplomatic and military domination for profit and plunder.
In other words, colonialism depended on ownership and control of territory and infrastructure, the extraction of labor, knowledge and commodities and the exercise of state power.”
In his characterization of “digital capitalism,” Kwet refers to the Internet companies’ submarine cables for data transmission, the server farms for cloud services, the attempts to implement proprietary rather than open software in developing countries and – under the heading “Colonization of Education” – the making available Training, hardware and (proprietary) software in schools in South Africa. In the three paragraphs under “Labor Exploitation,” the first paragraph discusses the mining of cobalt in the Democratic Republic of Congo using child labor and disease-causing working conditions. The second paragraph cites the low wages in Latin American countries compared to Western countries. The third section states: “Countries in the South also offer an abundance of cheap labor for tech giants. This includes data annotation for artificial intelligence datasets, call center workers and content moderators for social media giants like Facebook. Content moderators cleanse social media feeds of distributing content, such as gore and sexually explicit material, often leaving them psychologically damaged. Yet, a content moderator in a country like India can make as little as $3,500 per year – and that is after a pay increase from $1,400.
On the question of whether the USA or China are the dominant colonial powers, Kwet comes to a clear conclusion: “In reality, the US is the supreme tech empire. Outside the US and Chinese borders, the US leads in the categories of search engines (Google); web browsers (Google Chrome, Apple Safari); smartphones and tablet operating systems (Google Android, Apple iOS); desktop and laptop operating systems (Microsoft Windows, macOS); office software (Microsoft Office, Google G Suite, Apple iWork); cloud infrastructure and services (Amazon, Microsoft, Google, IBM); social networking platforms (Facebook, Twitter); transportation (Uber, Lyft); business networking (Microsoft, LinkeIn); streaming entertainment (Google, YouTube, Netflix, Hulu) and online advertising (Google, Facebook) – among others.
The upshot is, whether you are an individual or a business, if you are using a computer, American companies benefit the most. “They own the digital ecosystem.”
In the part “Political Domination and the Means of Violence” Kwet writes, among other things: “Governments and law enforcement agencies, in turn, form partnerships with tech giants to do their dirty work. In 2013. Edward Snowden famously revealed that Microsoft, Yahoo, Google, Facebook, Pal Talk, YouTube, Skype, AOL and Apple all shared information with the National Security Agency via the PRISM program.” The author also cooperates primarily with emerging countries in the areas of surveillance systems and prison management systems.
The author recommends to eliminate digital colonialism: “To stamp out digital colonialism, we need a different conceptual framework, that challenge, root causes and major actors, in connection with grassroots movements willing to confront capitalism, and authoritarianism, American empire, and its intellectual supporters.”
Comment: Kwet’s argument can be criticized because his basic concepts do not hold up if one examines the claims implied by them, because his characterizations and arguments of “digital capitalism” consist of seemingly randomly chosen examples and anecdotal evidence (while, in retrospect, colonial structures are listed more systematically ) and because the strategy he recommends is bound to fail.
Nations cannot be politically, economically and socially dominated, let alone controlled, by digital corporations because they compete with each other and are only part of the economy, even in developing countries. Political influence that is exercised on developing countries independently of digital corporations is at least as important as the influence of corporations. In no country can there be any question of a continuous symbiosis between digital companies and politics (as in the former “banana republics”), so that there is no sign of comprehensive rule, as was characteristic of earlier colonialism. The term “digital capitalism” is therefore likely to inhibit rather than promote knowledge.
Given the growing number of emerging countries, the “Global South” (as opposed to the “rich North”) does not exist. In general, from the perspective of emerging and developing countries, it is less about freeing themselves from the influence of digital companies than about not being left behind (further) as global digitalization advances. To ensure that this does not happen, the digital companies are making a contribution (albeit possibly with proprietary software), a performance that, like all other activities of the digital companies, is viewed exclusively negatively by Kwent. According to Kwet’s examples, the areas of responsibility that the digital companies have outsourced to countries with lower wage levels do not sound as if the working conditions were terrible (the miners in the Congo are expressly excluded here). The use of the term “exploitation” is still illuminating when digital companies base their wage payments on the wage level in an emerging country.
Citizens’ initiatives (“grassroots movements”) are far weaker in developing countries than in Western countries because their citizens are less able to afford public welfare-oriented activities or participation in mobilizations for their own sake in the struggle for daily livelihood and a liberal public is a natural ally for Citizens’ initiatives are often missing. Mass protests, in which individual citizens’ initiatives may play a role, are possible, but are usually directed against authoritarian, corrupt and incompetent governments.
Instead of promoting citizens’ initiatives as a solution to the dependence of developing countries on digital companies, what is needed is a more differentiated view of the problem, less superficial argumentation, pressure from the liberal public in Western countries and a commitment from Internet companies and companies in other sectors a globally applicable code of ethics from the US government, the European Commission and the UN. The issue that Kwent raised should be discussed further, but within an appropriate frame of reference.
Machine learning
Looking for the hidden picture
Artificial intelligence opens up
historical library holdings
Upload function – Limits of image similarity search – Conclusion and perspectives
By Klaus Kempf, Markus Brantl, Thomas Meiers and Thomas Wolf
third part
__________________________________________________________________________________
The application with the upload option by the user
__________________________________________________________________________________
In order to keep access as easy as possible for the user, the image similarity search was planned as a web application from the start. The most commonly used entry point into the image similarity search is a pool of partly hand-picked image templates that are assigned to different categories. This selection is intended as an invitation to browse and discover. Here you will find, for example, a selection of portraits from different eras, images of animals and plants, as well as numerous architectural drawings, views of buildings, maps, coats of arms, book covers, title pages, bookplates and much more. Regardless of the categories, the user has the option of displaying images as a pure, unfiltered random selection. The list of results then appears as a colorful hodgepodge.
If you start using the categories or random selection, the actual similarity search only begins from the second step. As soon as the user has selected a motif that interests them, they can explicitly search for similar images using this image. You can then start any number of other similarity searches using the motifs in the hit list. The user gets lost – which is entirely intentional – while browsing through the stock of images. In this way, useful, surprising and sometimes bizarre discoveries are made: Medieval miniatures resemble newspaper photos. Copper engravings can be found in different colors and qualities in other works by the same publisher. You can literally find something “copied” when a motif appears in a slightly modified form in a later work. Despite its visual appeal, the offer represents a certain challenge for the user and often tests their patience. The approach may not be intuitively clear to every user. Not every search hit makes semantic sense, even if the application’s algorithm detects plausible similarities. In some cases, the expectations regarding the accuracy of the hits and the performance capabilities of the software may also be too high.
An alternative entry into the image similarity search was also anchored outside the actual application, namely in the general viewer, which allows users to browse and view all of the BSB’s digital copies. An interface in the background checks every time you browse whether information for the currently open pages is stored in the image index. If this is the case, you can immediately start a similarity search with the content of this page.
For users who are specifically looking for a specific motif, there is the option of using any of their own image templates as a search template by uploading them. This means that he does not have to rely on the search templates only being within the BSB’s image inventory. Instead, it is possible to add your own images to the search. For example, a heraldist or genealogist can upload a scan of a coat of arms in order to find it in the BSB’s numerous digitized volumes of coat of arms collections. A book scholar discovers decorative initials or searches for the use of copper plates in various volumes from the 16th century. If you are interested in old newspapers, you will find photos, but you can also come across advertisements.
The images uploaded by users are not included in the inventory; After being used once for a search, uploads are immediately deleted. The upload option is one of the most popular functions of the online offer.
The frontend also offers a configuration option for weighting the search parameters. This allows the focus to be shifted to the color or edge features. The best results result from a balanced ratio of color and edge features. However, a search for pure color similarity makes no sense, as color alone is not a sufficient similarity criterion.
By specifying a threshold value (between 0.1 and 1.0) you can determine how much the result images should deviate from the search image. The lower the value, the greater the allowable differences. The best hits occur with a value between 0.85 and 0.95.
Figure 3: Similarity search by descriptor comparison
The same number of descriptors are stored in the database for the 54 million images and image segments. For each search query, a comparison is always made across all 54 million descriptors. This process takes a relatively long time, too long for use via the Internet. In order to keep the waiting time for the user as short as possible, the application is designed so that the search query can be distributed across several server processors (CPU cores). For this purpose, the list of inventory descriptors is divided into partial lists of equal size, each of which is run through on a CPU core. The individual results are then combined into one result. A typical search query lasts one to five seconds, but ten seconds at most. The speed depends on the number of search queries running in parallel. If many users are active at the same time, the capacity is divided accordingly.
__________________________________________________________________________________
Boundaries
__________________________________________________________________________________
The image search reaches its limits when it comes to the specific characteristics of the BSB inventory:
Images standing on the page (commonly found in digitized books) are processed without correcting the position.
Low-contrast, i.e. “pale” images are only inadequately recognized or the connection is not recognized. This applies to hand drawings and lithographs and occasionally results in “decapitated” people or otherwise incompletely captured motifs.
Machine learning methods were able to eliminate irrelevant image information. Nevertheless, the proportion of unusable images is relatively high at an estimated ten to twenty percent. This primarily includes tables, sheet music and handwritten works, which are far too often mistakenly recognized as images.
__________________________________________________________________________________
Conclusion and perspectives
_________________________________________________________________________________
The image similarity search established itself as a permanent fixture in the BSB’s digital offering portfolio soon after it went online. With around 4,000 visitors and more than 50,000 page views per week, it is one of the popular and successful offerings of the Munich Digitalization Center. The image similarity search focuses on the wealth of images, illuminations, woodcuts, graphics, drawings and emblems that were previously hidden in manuscripts, incunabula and historical book collections and are brought to the foreground thanks to this technology.
The image similarity search is currently carried out using a technical solution that essentially dates back to 2011 and works stably. The software solution was developed and licensed as a prototype together with a (semi) commercial partner. It has now become outdated and a successor solution is necessary. In contrast to today, the latter should definitely be an open source-based, in-house developed application.
Various approaches are conceivable here. Whoever wins the contract must take into account the fact that the Bayerische Staatsbibliothek has since fundamentally realigned its policy of data presentation and provision for its retro-digital copies. For historical reasons, there are currently over a dozen different online viewers in use for the retro-digital images, so that one can also speak of a “viewer zoo”. This should largely be replaced by a uniform technical infrastructure. We then decided to use the International Image Interoperability Framework (IIIF) [1] . Consequently, the new technical solution sought for the image similarity search must also be one that either results from the IIIF approach, can be combined with it or is at least compatible with it. [2]
This integrative approach also applies to the application of other search strategies based on machine learning, such as named entity recognition and search, which is currently being experimentally tested on the entire inventory of all full texts from the MDZ’s 2.5 million works. Here too, it is of crucial importance to consider that the creation of (even) greater search convenience for the user must only be accompanied by a development and maintenance effort that is medium-sized by the organizational areas of the house responsible for the “digital library” that suffer from chronic personnel shortages. and can be achieved in the long term.
The authors: Klaus Kempf, Bavarian State Library, Munich, klaus.kempf@bsb-muenchen.de – Dr. Markus Brantl, Bavarian State Library, Munich, markus.brantl@bsb-muenchen.de – Thomas Meiers, Heinrich Hertz Institute, Fraunhofer Institute for Telecommunications, Berlin, thomas.meiers@hhi.fraunhofer.de – Dr. Thomas Wolf, Bavarian State Library, Munich, thomas.wolf@bsb-muenchen.de
[2] See Brantl,M., Eichinger,R. , Wolf, T. (2016), A new open standard for image repositories – the International Image Interoperability Framework (IIIF) in use at the Bavarian State Library. In: ABI Technology 2016; 36(2): 105-113. Together with the Fraunhofer Heinrich Hertz Institute, a IIIF Mirado Viewer implementation of the image similarity search was experimentally tested in order to specifically increase its ease of use, see above. ., p.113
Streaming
55 percent of Germans stream
at least once a week
(E.ON) Streaming is booming: More than one in two Germans regularly consumes content such as videos, music, podcasts or games over the Internet – and sustainability also plays a role for many. These are the results of a representative Civey survey for the E.ON Energy Atlas, which is available at https://energieatlas.eon.de .
55 percent of German citizens say they stream online at least once a week. Families with children use such offers more frequently than average, at 70 percent. The streaming rate is highest among 18 to 29 year olds, with around eight out of ten respondents. The proportion is lower among older respondents (50 to 64 years: 49 percent; 65 and older: 31 percent).
The most popular streaming content is videos; 71 percent of streaming fans access them at least once a week. This is followed by music (49.5 percent), podcasts (22 percent) and games (11 percent). Around a quarter of those surveyed consume the content for “two to three hours” a day. Only one in 20 streams “more than five hours”. Furthermore, an urban-rural divide can be seen.
Use of social media
60% of internet users are on Facebook,
but approval among younger people is collapsing
Instagram’s market share increases to 46 percent
( Factskontor) Six out of ten Internet users aged 16 and over in Germany say they use Facebook. The social network’s market share is at a historic low for the second year in a row. Facebook is able to stabilize at a lower level across all age groups – but the decline of recent years continues among teenagers. In the group of users between 16 and 19 years of age, Facebook loses another four percentage points: only 32 percent of online users of this age still use Facebook. For the first time, Instagram is also losing market share in this target group. These are the results of the current Social Media Atlas 2021 from the Hamburg communications consultancy Aktuellkontor and the market researcher Toluna, for which 3,500 Internet users were surveyed representatively.
The proportion of twenty-somethings who say they use Facebook is also declining. It fell by five percentage points to 68 percent. However, among Internet users in their 30s, the social network increased by three percentage points. Facebook is most successful in this generation: three out of four online users between the ages of 30 and 39 use the service – more than any other age group. Until 2014, teenagers held this top spot. In the older age groups, however, there is hardly any movement: between 40 and 49, 65 percent of online users use Facebook (+1), between 50 and 59 it is 61 percent (+2), and from 60 years old 45 percent (-1).
The aging of the network’s user base, which began in 2015, continues. In addition, Dr. Roland Heintze, managing partner of the Aktuellkontor: “Facebook is no longer suitable as a universal solution in social media communication. In order to reach all target groups, their preferred channels on the social web must be identified and used appropriately. And this is especially true for the youngest not Facebook.”
As a company, Facebook was able to compensate for the weaknesses that had emerged by purchasing other social media services that were more successful in this target group. Instagram in particular has proven to be a lifeline for teenagers in recent years. But that time is already over: Overall, Instagram can continue to expand its market position. 46 percent of all online users aged 16 and over use Instagram, six percent more than in the previous year. Usage is also increasing in almost all age groups – except teenagers. After years of continuous growth, the market share there suddenly collapses by eleven percentage points. 80 percent of 16 to 19 year old online users in Germany use Instagram, 82 percent of 20 to 29 year olds. Heintze again: “The battle for the top spot in social media usage among the youngest will be fought out around every six years among completely new competitors.”
OpenPassword
Forum and news
for the information industry
in German-speaking countries
New issues of Open Password appear three to four times a week.
If you would like to subscribe to the email service free of charge, please register at www.password-online.de.
The current edition of Open Password can be accessed immediately after it appears on the web. www.password-online.de/archiv. This also applies to all previously published editions.
International Cooperation Partner:
Outsell (London)
Business Industry Information Association/BIIA (Hong Kong)
Open Password Archive – Publications
OPEN PASSWORD ARCHIVE
DATA JOURNALISM
Handelsblatt’s Digital Reach
MARKET TRENDS
Library Trends 2022
What are the key trends in the library market for 2022?
Online-Publications | 2022 | 2021 | 2020 | 2019 | 2018 | 2017 | 2016 | GPT AI SEARCH
Print Archive | 2016 | 2015 | 2014 | 2013 | 2012 | 2011 | 2010 | 2009 | 2008 | 2007 | 2006 | 2005 | 2004 | 2003 | 2002 | 2001 | 2000
