“Artificial Intelligence” is all the rage just now, permeating nearly every facet of life, work, and play. Is it useful for living history research?
Traditional search engines, developed as a way to find information on the internet from nearly the beginning, utilize “crawlers”, programs that scan websites and catalog or index key words, images, and phrases in a giant database. When you type specific search terms into an engine, the results returned are links to the specific sources the database indexed.
The “large-language learning models” used for current AI efforts are different. They scrape and store words from millions of sources, and aggregate the word patterns into a mimicry of human communication. The results returned are an artificial expression of language that tries to fit the query provided.
Each type of programming has positive and negative consequences when researching for living history application.
Traditional search engines return specific sources, which the individual must analyze for facts and relevance, as well as determine what inherent biases the author of the source may be expressing, and how that may affect the results and information.
AI bots return an amalgamation without analysis, making it very difficult to track or examine the source material behind individual claims, and increasing the likelihood of promulgating fiction, versus fact. The AI bot has no way to distinguish between, for instance, census records from a government source, and a fan-fiction story with absurd age gaps at marriage (between a fictional immortal and a fictional human teen, for instance).
Because traditional search engines return those specific sources, we can more easily keep notes (an annotated bibliography, source list, etc) on what information was gleaned from which source. When we discover a source should not be on a trusted list, we can eliminate those aspects from our research and understanding.
Traditional search engines are merely indexing information. AI bots scrape, assimilating the unique author’s individual work into the large-language model, effectively stealing it. The intellectual property rights violations are just beginning to explored and litigated, but any reasonable person can see the dangers in co-opting individual rights. Scraping content has been considered unethical since the early days of on-line publishing.
When researching with a traditional search engine, the researcher must employ their own intellect and experiential context to make sense of the results. AI bot usage subrogates the human element, and does not require creative synthesis that brings together facts, understanding over a wide range of vetted sources, and the context of experience unique to the individual (as well as the shared experiences the individual may receive from others.)
The AI bot cannot access the individual’s braod range of inquiry over many topics and multiple years, and has no frame of reference for the experiences the individual may have when applying researched information in historical settings.
When the individual encounters an interesting fact not directly related to their current inquiry, the human brain can store that item as a “Well, Huh!” topic on their individual shelf. One the brain has “pinged” on that topic, it will seek to add to that shelf over time. The next resource that has a similar or closely related fact will “ping” in the human brain and add another item to the shelf. When this repeats over time, the individual brain starts to formulate an entirely new topic to research, backed with multiple documented instances long before the topic can be expressed fully! Using traditional search engines in addition to hands-on and physical research in books, the individual creates a source bank.
AI bots are incapable of aggregating information in the same way. This sharply limits the utility for ever-expanding, unique lines of inquiry that leads to really exciting new facets of understanding in living history!
When we consider the sharp limitations of AI bots for research, and layer on the distinctly negative consequences to the environment from the data center pollution required to “house” the AI bots and stolen physical and intellectual resources that power them, it makes little sense to abandon traditional research methods enhanced by digital research resources accessed through traditional search engines.
One thing AI bots actually can do somewhat reasonably well is tied to their foundational programming. As large-language models written to analyze and mimic human expression, they can be used to create a more nuanced translation from one language to another. Direct translation algorithms sometimes return clunky language; AI bots can often return a translation that takes into account grammatical structure and some ideographic nuance that renders the translation more understandable. Trying to understand a German source text in English, for instance, can be made easier with a good translation model assisting. (The best translation tool is still a fluent speaker with understanding of both languages and the subject matter!)
In general, avoid AI for summaries or “research.” Learning solid individual research skills is worth the time it takes, and will return better information over your years of inquiry.