Discover more from hrbrmstr's Daily Drop
Heat; Copy.AI; ✳️2✳️
Happy Monday, fearless readers! Apologies for the unannounced hiatus. I was down and out for just over a week thanks to a double-jab of the latest Pfizer Covid booster (all previous 'ccines were Moderna) and seasonal influenza. It was a real effort just sitting up most of last week, given that it felt like I lost a battle with The Hulk. Still not 100%, but in ship enough shape to drop links again.
This year, historic heat records were set virtually everywhere across the globe, and you — along with hundreds of millions of others — were likely negatively impacted by some sort of excess heat event for at least one day. When these events happen, I'm always curious how much of an aberration any given one is (if at all). If you're a U.S. denizen and have similar ponderings, then this heat records visual essay by Matt Daniels and Russell Goldenberg should be something you keep at-the-ready in your bookmarks.
The section header image is just one chart from this interactive single page scrollytelling piece, and each of them lets you explore and compare heat records for a fairly diverse set of U.S. places.
The data comes from the Applied Climate Information System (ACIS) which is a system architecture and API built and run by NOAA’s Regional Climate Centers (RCCs). It isn't so much a historical weather data archive as it is a system that delivers operational information derived from historical archives and near real-time climate data. ACIS "includes an integrated metadata system for data and observation discovery; national, regional, and local datasets; is used as a data source for RCC websites and delivers climate data-products to supply information to partners and other end-users."
What makes ACIS so special?
Most climate data management systems focus on the storage of climate data and, at best, provide the ability to access data subsets or simple pre-defined summaries from single datasets. ACIS is unique because it focuses on producing customer-driven, value-added products that combine data from multiple data sources. Product quality is assured through rigorous definition of station and individual station element compatibility defined in a metadata database. Data and metadata archives, product access interfaces, and product generation capabilities exist at multiple RCCs. These capabilities are linked via the Internet to provide a seamless, robust and reliable product access system. Identical products can be generated and delivered to a user regardless of which RCC’s system is accessed. Automated 'behind the scenes' switching between centers assures users of dependable and fast access to ACIS products.
One of the ACIS data endpoints used in the heat vis essay is ThreadEx, which is the keeper of long-term weather station extremes for America. Select a city and record type, and you get data and charts back. It's not as fancy as essay, but it gets the job done.
Here's why TreadEx is important/needed:
ThreadEx is a project designed to address the fragmentation of station information over time due to station relocations for the express purpose of calculating daily extremes of temperature and precipitation. There are often changes in the siting of instrumentation for any given National Weather Service/Weather Bureau location over the observational history in a given city/region. As a result, obtaining a long time series (i.e., one hundred years or more) for computation of extremes is difficult, unless records from the various locations are "threaded" or put together. This has been done, but different approaches and combinations of stations have resulted in confusion among data users and the general public about what constitutes an official daily extreme record.
In consultation with NOAA's National Centers for Environmental Information (NCEI) and the National Weather Service (NWS), the Northeast Regional Climate Center (NRCC) has evaluated station relocations and built "threads" for 270 locations that are published in NCEI's Local Climatological Data using NOAA daily data sets. The data sets used for this project include NCEI GHCN-Daily, NCEI DSI3210, DSI3200, DSI3206 and DSI3205. In addition, NRCC has been able to extend over sixty station threads back further in time using daily data contributed by local NWS offices, state climate offices and regional climate centers. An ongoing process of adding daily data from the old NWS Climate Record Books continues. These data have been digitized, passed through quality control and used to extend threads back even further in time for over fifty stations to date (see Revision History for details).
The rendered source of the scrollytelling heat essay was built with Svelte, and if you'd like to try your hand at something similar — perhaps even with the same data — check out this library (demo) by the spiffy folks over at the U.K.'s Office of National Statistics.
The blank page or canvas can be quite daunting, and is often the bane of many creative efforts. In a writing context, one may want to drop a fantastic blog post on to the world, but fail to congeal scattered thoughts and loose notes into something that will enable you to even put one syllable into a new document. Sometimes all one needs is a prompt, but it's also helpful to get some structural elements (e.g. topic sections) in place as well. All that is part of the creative process, but it's also more of a chore and a new service, Copy.AI, aims to help you get past the chore and into the creative part as quickly as possible.
There are various copy contexts (which are likely backed by different AI model flavors) available:
Digital Ad Copy
Social Media Tools
and each one has a series of prompts that enable Copy.AI to generate a scaffold that you can refine and start putting content in. An example might make this clearer.
I chose "blog post" from the Blog Tools and gave it the following answers to requested prompts:
Blog title: The Battle Against Far-Right American Authoritarianism
GOP MAGA authoritarian strongman oppression nationalist
It took Copy.AI a few (noticeable) seconds — I'm uncertain if that was for show or if the model run really took that long — and it came back with:
Authoritarianism is a serious problem in America.
Authoritarianism is a personality and not a political ideology.
Authoritarians have three key features that make them an especially dangerous threat to democracy.
In the United States, authoritarianism results in strong support for Donald Trump and his Republican enablers.
The authoritarian threat goes far beyond just Trump and his supporters,
Republicans and MAGA-land denizens.
There are no shortage of authoritarians among right-wing extremists who might not consider themselves to be on the Republican side of things.
The pro-Trump authoritarians are extremely intolerant, especially regarding Muslims and nonwhite immigrants, but also minorities in general.
There is a significant overlap between American authoritarians, religious fundamentalists and all manner of anti-science attitudes and beliefs.
American authoritarians want hierarchy, obedience, conformity and tyranny, as opposed to liberty and equality.
Authoritarianism is a serious problem both within the GOP and without.
This is just one example, but it seems clear they've trained their models on pretty current corpora, given the level of detail just in the suggested section headers. And, it sure did drop the same talking points I likely would have eventually put forth in such a screed.
It's pretty agnostic, too. The starter plan is free (as are many tools after the trial period), so go ahead and give the blog post helper the following prompts:
Blog title: A Path Back To Conservative Morality, Freedom, And Prosperity
god country morality strength liberty freedom
and see what it drops (it's nice that our AI overlords can help empower evil as well as good?)
It comes with a Chrome extension (insert the usual "caveat pasco" on trusting any browser extension), so you can use Copy.AI in any text box you encounter.
I believe assistive technology like this, or junior cousins such as gDoc's suggested phrase/sentence completions, does have a place in the certain creative workflows. Just remember that these models are trained on words already written, just as this year's "diffusion" visual trend is based on existing images curated with associated text descriptions. Sure, both are creating "new" out of "old" (just as we humans have been doing for eons), but AI will (hopefully) never have the "spark" that occurs in our very human gray matter that truly births something completely and amazingly original into the universe. That's also not Copy.AI's intention, and I can see myself using this tool for ideation and experimentation moving forward.
As y'all know, I'm a big fan of "do one thing, and do it super well" tools. You also know I'm a fan of all things Rusty, and 💙 the Apache Arrow and Parquet format and tooling ecosystem. It's truly delightful when all three of those likes converge, as they have with Dominik Moritz's (@domoritz) series of data format conversion tools:
They're fast, convenient, and absolutely do what they say on their respective tins.
Being waylaid by microorganisms means I haven't gotten to writing the GreyNoise Steampipe plugin yet, but — fear not — that's still on the TODO. ☮