Part 1 - Reflection on Hands-on Generative AI Experience
The uses of AI are numerous, and I believe tool-dependent, as the different types of engines completely differ in what they are designed to do. This was evident in the demonstration from Dr. Humphries as he showed the different versions of AI through the years, with the simple chess program, for example. Generative AI, in my opinion, through using ChatGPT to ask the prompts of the French Revolution, really blew me away with its level of depth and the sources it provided me. ChatGPT is absolutely a search engine that finds information, it gives a lot of information and outlines it very well for you, just on the first prompt alone on the major historiographical debates about the causes of the French Revolution was incredibly robust, it gave me six different schools of thought in chronological order of when it would have been most prevalent and had the most contemporary version last.
It also absolutely does help as a research assistant when you need to organize and summarize sources, in my experience, I have asked it to help group different sources of a specific topic into how the groups should be used, it proved to be very helpful in organizing my sources for a research paper and not be lost spending too much time on sources not pertinent. I believe it can be a distant reading tool to find patterns in large collections of text. I do not see why it would not work for that; it did it with the Amelia Bloomer corpus very easily, identifying the key issues and using quotations to support identified themes.
Finally, I do not see it as being like a co-author because the writing appears to be directly taken from the sources, so I am not sure how much credence it would get as an author. The AI can still make mistakes, as seen in the presentation when it was translating primary sources from images, but it can also learn the writing, so I am torn on the issue. When I think of AI like Grok, which has been proven to be manipulated with an agenda, it does give me cause for pause. When using the Corpus even when I asked for it to be skeptical on the suffrage movement it appeared to be very bipartisan in manner, it emphasized it was not it’s own opinions and was very clear on the context of such skepticism that could have been present at the time, such as the clarification that the historical opponents would frame themselves as defends of social harmony, and complementary gender roles, and not as hostile to woman.
I believe these tools do make historical research more accessible to the masses, as long as the systems in place of the information are credible, which so far it appeared to be as it would source the research with titles of historical work and names of authors, even going as far as always clarifying that there is nuance to such research questions and not attempting to give me one solid answer as being the sole reason. Unfortunately, as previously mentioned, big tech companies also become more powerful and can use these machines non-ethically, such as Grok, where it has been caught spreading misinformation, and the collection of people’s data as well.
Part 2 - Reflection on Scholarly Blog Readings
When reading both articles, I was skeptical the entire time. Perhaps it was due to my exhaustion and weariness from reading/hearing about AI, or maybe it is just due to me being a Luddite. I will begin with the argument that Gan poses about the use of AI in research and the almost redundancy of strenuous research that takes years. I realize that Gan’s is speaking on the topic from an economic and well numbers-focused use of generated research, which has to be noted is far superior in nature to that of writing generated research, due to the nature of the machine processing tokens and understanding numbers is far more conducive than deep archival research or historical nuance (from what I understand).
When I take what Gan argues to be the now (p)research, as in research that is made in anticipation that it will be useful for someone else, but in my mind, this will just lead to a rot of research and knowledge, especially as it comes to history. If AI programs start learning and using information that is created by AI as its main source, and there is no more (p)research, as he would say, eventually sources would run out if all research being done is essentially done already. This links into what Humphries questions about the incredible ability of Deep Research in scouring archival sources and even translating primary sources, such as handwriting, which is very impressive, but AI itself cannot find those primary sources; it has to be found by a physical person to be uploaded onto a database. The tangible aspect of a historian will not change there.
I believe that the mass skepticism seen in Humphries’s comment section is very much warranted. It was all civil discourse, but granted full, of skepticism, but Humphries rebutted quite respectfully and in a very educated manner, which gave me pause for skepticism. The actual historian can use AI to sort mass numbers of files and sources, such as primary sources if the AI is as powerful as Humphries’ states in reading as good if not better primary sources, it should be able to maintain and sustain an archive much easier, which is a great advantage, further more the ability to access sources is a great research assistant for an ethical historian who still wants to write it themselves. Unfortunately, I do not share the optimism of Humphries’ people are becoming over reliant on these tools and it is a threat not only to history but on an anthropological level, for example: some intern adds incorrect health information to a government website because an LLM screwed up, no citation is given and this is now a primary source in the public’s eyes, or AI systems like as seen with Grok can have some sort of an agenda.
I do not see the historian being replaced with AI, as people in the comments emphatically stated, but I do see the quality of research declining if people become over-reliant and lazy with their dependence on AI, as anyone can now “publish” a book that they had an LLM write in an afternoon. If I had to be more positive on the aspect, one could argue that the future of historical “knowledge” could be more accessible to those who before would not digest it or access “knowledge,” but also it can dissipate and become far lower quality with bad actors involved.