#statstab #537 {hdbayes} An R Package for Bayesian Analysis of Generalized Linear Models Using Historical Data
Thoughts: An interesting approach to priors. I'm not v familiar w this so curious what others think.
#statstab #537 {hdbayes} An R Package for Bayesian Analysis of Generalized Linear Models Using Historical Data
Thoughts: An interesting approach to priors. I'm not v familiar w this so curious what others think.
Scientific Data: Transcribing historical Canadian weather data. “Historical weather journals from across Canada, spanning 1768–1884, have been transcribed from handwritten records into machine readable formats. The NORTHERN (Nineteenth-century Overseas Records Transcribed for Historical Environmental Reconstruction in the North) project transcribed nearly 2 million weather observations from […]
https://rbfirehose.com/2026/05/01/scientific-data-transcribing-historical-canadian-weather-data/USGS: New AI Tool Forecasts Drought 90 Days Ahead Nationwide. “The USGS River DroughtCast uses machine learning models trained on data from thousands of USGS streamgages, some with more than 100 years of continuous records, to forecast when rivers and streams will drop to abnormally low levels.”
https://rbfirehose.com/2026/04/06/usgs-new-ai-tool-forecasts-drought-90-days-ahead-nationwide/Reuters: Belgian museum, US mining company at odds over colonial-era Congo archive. “A U.S. mining company backed by billionaires Jeff Bezos and Bill Gates is in a tangle with Belgium’s AfricaMuseum over who should digitise antique maps of what is now the Democratic Republic of Congo in the museum’s archive.”
https://rbfirehose.com/2026/04/04/reuters-belgian-museum-us-mining-company-at-odds-over-colonial-era-congo-archive/TechCrunch: Google is using old news reports and AI to predict flash floods. “Google researchers used Gemini — Google’s large language model — to sort through 5 million news articles from around the world, isolating reports of 2.6 million different floods, and turning those reports into a geo-tagged time series dubbed ‘Groundsource.'”
https://rbfirehose.com/2026/03/12/techcrunch-google-is-using-old-news-reports-and-ai-to-predict-flash-floods/France 24: Researchers turn to UK archive to solve under-50s bowel cancer mystery. “As growing numbers of younger people are being diagnosed with bowel cancer, researchers in the UK are hoping an archive dating back more than 100 years could provide answers. In a study which began in January, researchers are using cutting-edge technology to compare archived tumours with ones from the present […]
https://rbfirehose.com/2026/02/05/france-24-researchers-turn-to-uk-archive-to-solve-under-50s-bowel-cancer-mystery/Space Daily: Century-old solar records refine future cycle forecasts. “An international team of astronomers has developed a new way to extract solar polar magnetic information from more than a century of historical observations, improving prospects for predicting future solar cycle activity.”
https://rbfirehose.com/2026/01/22/space-daily-century-old-solar-records-refine-future-cycle-forecasts/A student trained an AI only on 1800-1875 London texts, then asked about 1834. The model described street protests and Lord Palmerston - details the student hadn't encoded, but which turned out historically accurate. The "quarantine strategy" may reveal how models encode information in unexpected ways, raising questions about what else they capture without deliberate training.
#AI #MachineLearning #HistoricalData
https://www.implicator.ai/an-ai-answered-a-question-about-1834-it-was-right/
TimeCapsuleLLM: LLM trained only on data from 1800-1875
https://github.com/haykgrigo3/TimeCapsuleLLM
#HackerNews #TimeCapsuleLLM #LLM #1800s #AI #research #historicaldata