What This Librarian Thinks of GenAI
Generative Artificial Intelligence (GenAI)?is nothing more than a tool that predicts or creates an output based on the input (prompts).? It relies?on models that can create new content in the form of text, media, or code.??It uses Natural Language Processing?(NLP)?from the Large Language Models (LLMs) of text and images it was trained on to develop patterns of output: essays, images, slide content, etc.? The LLMs may have a period when?their training stopped and have no new information, and?therefore static.??
Most free generative AI tools are limited in size by how many tokens are used.? Tokens are a breakdown of the words used in the prompts and each GenAI has its own limits.? With each output, you can instruct the GenAI on how to refine the results.? The best prompts come from domain knowledge and that knowledge should also be used in refining the output.? It is known that some GenAI output may contain false facts and incorrect citations; use critical thinking and?check everything?for accuracy before handing in any assignment.
Retrieval-augmented generation (RAG) is for introducing content to be ingested into the LLM.? This can only be done using structured information to which?you own the copyright, public domain sources, and sources supplied by Text and Data Mining (TDM) agreements.? Most database vendors are now adding to their licenses?that their content/articles can not be used to train a?LLM.? Since most GenAI tools use input content to train their LLM, you can not enter most articles into a GenAI for summary, search, or analysis.
Some Library databases to which we subscribe?are beginning to train their embedded GenAI?with the content from?their databases via RAG; this is called GenAI with a?Small Language Model (SLM).? The results will be more authoritative, but limited by the content of that vendor.? You will still?have to search multiple databases to get a better picture while developing your subject expertise.?
Generative AI is not a search engine, it's another way to think about things
Why use GenAI?
Good for: Writing models, Suggesting synonyms,?Copy editing,?Brainstorming, Summarizing (non-copyright materials)
Not Good for:?Domain expertise, Searching propriety databases, Reasoning like mathematics, Original content, Current reality
领英推荐
Warnings
1) Results may contain out-of-date, wrong, and made up information
2) Materials used to train the LLM, do contain?bias
3) GenAI output is NOT?your final project, customize and check everything
Do Use
Concerns for the Future of the Generative AI
A) Depending on court cases regarding Copyright, there may be major revisions in the LLM
B) As the industry matures and consolidates how will the software companies monetize generative AI? Will they turn to advertising with subtle product placement or subliminal messaging?
There is a lot of other information, wisdom, and concerns others have written. I just wanted to add my humble opinion.
GenAI does not have your domain knowledge and doesn't apply critical thinking in its responses.
I agree. It reminds me of Noam Chomsky's thesis/statement "Colorless green ideas sleep furiously" Gen AI is often grammatically well-formed, but semantically nonsensical or just "too much fluff."