machine learning – Tracy Durnell's Mind Garden

James G. put together a fun clickable bookshelf using computer vision, OCR, and GPT — this looks far easier than ye olden days of image maps
Researchers have figured out how to extract and decipher text from a scroll carbonized in the Vesuvius eruption using machine learning 🌋
Anna Ridler’s Mosaic Virus is a clever art installation likening the furor over cryptocurrency to the great tulip rush that uses machine learning to render tulip imagery
Alexandra Daisy Ginsberg‘s Pollinator Pathfinder project generates custom planting plans designed for pollinators, and even has a preview of “pollinator vision” 😍 The plants aren’t for North America but the concept is so cool!

Pollinator Pathmaker, Digital render in pollinator vision, 2023. © Alexandra Daisy Ginsberg Ltd

This puts me in mind of Annie Sturdivant’s comparison of AI and plastic; there are some things that plastic is absolutely amazing for (lots of medical stuff), but also the world is drowning in single-use plastic waste.

More interesting uses I’ve saved before:

Analyzing traffic safety with AI

Using GPT-4 to measure the passage of time in fiction by Ted Underwood

Getting creative with embeddings by Amelia Wattenberger

Identifying whale individuals by their tail with machine learning

The siren call of LLMs

We’re misusing LLMs for tasks we want automated

I’ve noticed a lot of people turning to Large Language Models to collect and analyze data for them. I believe relying on LLMs to produce examples, write summaries, and analyze information is an inappropriate, or at least unreliable, use of the technology.*

The LLM sirens whisper our wishes back to us: that we need not waste our time with the boring parts. That we, too, can harness the power of coding or writing without putting in the effort to learn how. That we don’t actually need to read that article to glean “the takeaways” from it. That the only value of research is the final output, and we gain nothing by doing the work ourselves.