Category: Uncategorized

Did you know about stackoverflow.ai? I must’ve completely missed this. It looks like a great alternative to the search function on the site (or using Google to search it). We seem to have come a full circle from LLMs killing StackOverflow to LLMs powering StackOverflow for search and discovery. Recommended.

The intro sections and DAGs for the RD chapter are in. More to come.

I’m looking for interesting datasets for the RD design. I have some candidates, but I’m eager to find more compelling, real data. Ideally, I’d like a business case (rather than policy), such as one on customer loyalty status. The IV chapter already uses policy data (tax on cigarette prices vs. smoking). Please comment with a link if you have ideas beyond the Kaggle datasets.

As a reminder, Causal Book is an accessible, interactive resource for the data science and causal inference audience. It is not meant to substitute for the excellent texts already available, such as The Effect by Nick Huntington-Klein and The Mixtape by Scott Cunningham. This book aims to complement them by focusing on the idea of solution patterns, with code in R and Python, exploring different approaches (Freq. Statistics, Machine Learning, and Bayesian), and clarifying some of the counterintuitive (or seemingly surprising) challenges faced in practice.

Causal Book

This is interesting: Palantir has launched a “Meritocracy Fellowship” to hire high-achieving high school graduates right out of school, offering a paid internship with a chance at full-time employment. The company presents this as an alternative to college.

This is a very limited, transactional view of college. College is more than just training for employment; it is where students gain knowledge and broaden their horizons, learning how to think and ask questions, in addition to acquiring practical skills. I doubt a four-week history seminar will make up for all that.

Source

Is our society increasingly rewarding conformity? Is AI accelerating this process, and is it simultaneously stripping work of deep meaning? After all, thanks to LLMs, many now define creativity as merely a probabilistic recombination of matrices derived from a training set.

Is science also contributing to this potential lack of deviation in culture, education, arts, architecture, and business? Here’s a take on it (not mine):

You can spot this scientific bland-ification right away when you read older scientific writing. As Roger’s Bacon points out, scientific papers used to have style. Now they all sound the same, and they’re all boring.

Maybe science is supposed to be boring? What happened to the style though?

For definitive answers to these questions, look elsewhere. For a compelling set of data (plus plenty of causal speculation) on many aspects of contemporary society and scientific style, check out this compilation and the essay here.

The greatest thinkers in science (and business) are often prolific authors. They write books, blogs, and copious emails to sharpen ideas. Richard Lewontin, E.O. Wilson, and Paul Graham are but three examples. Dorothy Hodgkin’s scientific correspondence and papers, stacked together, extend 25.85 meters in length. Great thinkers, in other words, write all the time.

Researchers are evaluated by simple measures of productivity or influence — number of papers published, citation count, and grant dollars. In such an environment, it has become exceedingly difficult for scientists to take stylistic risks in their academic writing or to devote significant amounts of time to other forms of creative writing.

Here’s an update on using LLMs for OCR without having to use the same hammer (generic model) for all nails. DeepSeek has released an OCR-focused model: https://github.com/deepseek-ai/DeepSeek-OCR

Check out the deep parsing mode, which is parsing images within documents through secondary model calls. Very useful for data extraction. The results are pretty impressive too:

Our work represents an initial exploration into the boundaries of vision-text compression, investigating how many vision tokens are required to decode 𝑁 text tokens. The preliminary results are encouraging: DeepSeek-OCR achieves near-lossless OCR compression at approximately 10× ratios, while 20× compression still retains 60% accuracy. These findings suggest promising directions for future applications, such as implementing optical processing for dialogue histories beyond 𝑘 rounds in multi-turn conversations to achieve 10× compression efficiency.