AndreasDec 17, 2022
Jaimie Murdock

On Monday, I received my author’s copy of “The Dynamics of Science: Computational Frontiers in History and Philosophy of Science” from University of Pittsburgh Press. https://upittpress.org/books/9780822947370/

Proud of my chapter with Colin Allen: “LDA Topic Modeling: Contexts for the History & Philosophy of Science.” A pre-print can be found at http://philsci-archive.pitt.edu/17261/

In it, we discuss common misconceptions about the interpretation of topic models, relevant for all practitioners in #nlp & #digitalhumanities.

The Dynamics of Science - University of Pittsburgh Press

|9780822947370|Computational Frontiers in History and Philosophy of Science|Millions of scientific articles are published each year, making it difficult to stay abreast of advances within even the smallest subdisciplines. Traditional approaches to the study of science, such as the history and philosophy of science, involve closely reading a relatively small set of journal articles. And yet many questions benefit from casting a wider net: Is most scientific change gradual or revolutionary? What are the key sources of scientific novelty? Over the past several decades, a massive effort to digitize the academic literature and equip computers with algorithms that can distantly read...

University of Pittsburgh Press
Dec 16, 2022 at 5:57amWeb
Show threadJaimie MurdockDec 16, 2022

With the delay from chapter authoring to… 2 weeks ago (intro of ChatGPT), I wish I had laid a claim on “pick the right model for the task.”

These new generative models are truly wonderous and (in my opinion) approach General AI. Strikingly, they find adequately “explanatory” responses.

However, for both humanists and analysts, we need to explore how ChatGPT aids meta-analysis: summary of a body of knowledge and interpretation. Are the old ways (like LDA) useful?

Show threadandreiDec 16, 2022
@Jaimie oh for sure! i’ve been thinking a lot since the unveiling of ChatGPT about the need for a meta-analysis and deconstruction of it as the such general AI tools are going to progress because the possibility for bias and conditioning is so high imho