Update summaries-sebastien-bonhomme.md

_objects/summaries-sebastien-bonhomme.md

@@ -38,4 +38,17 @@ The group who did “Preserving Society Hill” went similarly with their set of
 For the 23rd of October lesson, we were touched upon the subject of networks and their visualisation and how they could be utilised efficiently by historians to illustrate information given by a set of sources, allowing the recognition of patterns or prominence of a certain figure/aspect, left otherwise obscure. We were introduced to this concept of networks via the example of a wedding and the planification of the different guests attending, their attributes (here for example coworkers, singles, friends etc.) and how they are grouped on that network. We were met with other types of occurrences and notions around these social networks like brokers, to give an example, and the general boundaries of a network.
 We used a site called Palladio that allows the visualisation of networks through spreadsheets files like the ones on Excel. We were later tasked with creating our own network to visualize via Excel and Palladio. Initially, I had the idea of creating one related to Rock and Heavy Metal bands and how they could be linked with each other but due to time constraints and not being sure how to make it work, I banded together with Emilie and Jelena to work on a new wedding guest list to have a more familiar basis, as the topic matter wasn’t always easy to follow.
 
+## Summary, review and questions on the 30rd of October 2024 session of Introduction of Digital History
+
+For the session on the 30th of September, we were invited to a conference held by Ludovic Delépine, Head of Unit at the archives of the European Parliament, on the topic of their incorporation of AI for a more efficient search engine for the public wishing to access their documents. The utilized AI, called Archibot, is based on Anthropic’s Claude AI and aims for limited hazards via constitutional AI. The whole process in general for the addition of the documents utilises many different aspects, like for example a tool for the extraction of text data from previously scanned documents. When looking up for documents, the AI can give results based on their metadata, their content and the frequency of given thematic keywords. The prompts and queries done by the user for their research can be done in a multitude of world languages, however some languages, like Romanian for example, have more issues than others and the AI may switch back to English to give an answer. The result shows the 10 results deemed the most relevant with each a brief description beforehand. The AI, to avoid biases or potential errors, only shows results that are indeed from the Archives of the European Parliament and who respect the human rights convention of the United Nations. Furthermore, this corpus of sources comprises documents ranging from 1952 and 1994. The EU frameworks mandate that documents from 30 years ago on must be rendered accessible to the public. After the meeting, some questions were asked and answered. For example, if this process would be made available for other archivistic institutions across Europe. This however is, at the current time, impossible as the search engine and AI is based on software that is not opensource. Perhaps if one day it will be via free software, it may be shared with other institutions. To add up, the utilised softwares are those the EU has given their approval for.
+Upon accessing and contemplating the dashboard of said European Parliament archives, one may make several commentaries regarding the way it is build and presented. Firstly, the dynamic display of the number of available documents and its changes with each filter is informationally appealing, as are the graphical displays of each type and languages for among these documents, their amount each year respectively etc. However, one may criticise the way, in the filter “fond”, how they are named, namely with a set code not indicating at all what they could refer to. This is most certainly not adapted for the grand public the archive of the European Parliament aims to reach. Names rather like those given at the National Archives of Luxembourg for their fonds should be opted for instead as these give some information of what is to be expected inside it.
+
+Is there not a risk or fear that with the advent of the usage of artificial intelligence in archival work, but also in general of such automatization, that more and more aspects of your work will be delegated to the (intelligent) machine and that thus, more archivists aren’t needed anymore, because the machine can now do the task more so and that as a consequence, some of you will lose their employment in the future or that less and less positions will exist and archivists be demanded?
+
+How can you make the AI fully trustworthy? Most often than not, AIs have been fed, sometimes maliciously, with information that ended up harming their learning process or have easily been manipulated by their human interlocutor.
+
+How did you digitalize all these millions of documents and “AI” with it? Must have been an immensely long process over years.
+
+How do you plan on waking the interest of the general public to discovering the European Parliament Archives and use this AI and not just the more academic public?
+