Drop #275 (2023-06-12): Back In The Saddle

BLE; Pax GPT; WTHeck Is Happening?; Moar Perplexity

Programming Note: The weather $DEITIES were on our side last week as the early Summer Down East Maine climate turned out to be better than expected. Your friendly neighborhood hrbrmstr is rested up and raring to get back into the swing of things. He also completely lost track of time, forgetting that he ha[ds] a five-hour CMU lecture to do tonight, so the Bonus Drop will still be forthcoming (Sunday was an unforeseen lecture refresh/prep day).

After tomorrow's Typography Tuesday edition we'll be spiraling into some solid tech-heavy Drops for the next few weeks. So, today I thought I'd ease us back into the routine with a cool CLI tech find, GPT stochastic preaching, and a fairly alarming weather trifecta (with tons of datavis!). We'll close with me continuing to wax poetic about Perplexity.ai.

BLE

Yes, I deliberately obfuscated what's in this section, and some of you most certainly thought we'd be talkin' Bluetooth Low Energy despite the “CLI” nod in the setup block.

BLE, in this case, stands for Bash Line Editor, a “full-featured line editor written in pure Bash! Syntax highlighting, auto suggestions, vim modes, etc. are available in Bash interactive sessions!”. It aims to provide a similar editing experience as Zsh's line editor (but, obviously, within Bash). The author started developing it as an experiment in 2013 and released the first version back in 2015. The latest version (0.4) adds programmable highlighting.

Ble.sh is implemented entirely in Bash shell scripts and relies on POSIX utilities. No Rust. No C/C++ compilers. No LLVM. It can be installed either by cloning the GitHub repository and building from source, or downloading a pre-built tarball from the releases, and provides a basic infrastructure for users to configure their own prompt, aliases, and functions. Key bindings and highlight colors can also be customized using ble-bind and ble-face functions.

The section header is an MP4-ified version of the gif at the GH URL.

While I have pretty much switched to Zsh (thanks to Apple), I do keep my Bash setup current and this dropped in super well.

Pax GPT

ChatGPT (et al.) is super good at generating content, especially if you provide decent prompts and aren't expecting the next blockbuster young-adult fiction novel to be auto-generated without editing. I've had it take a stab at emails, summarize Slack threads (be sure you don't put sensitive content into ChatGPT from $WORK without permission or even with permission), and have even used it to augment some teaching preparation.

Not that I am a preacher or anything, but one thing that I would not have LLMs/GPTs do is write a sermon for me.

It appears that a decent number of humans recently attended an experimental Lutheran church service in Germany where most of the content, including the sermon, prayers and music, was generated by the AI chatbot ChatGPT. Obviously, ChatGPT alone can't play music or create human-like avatars with voices to deliver said content, so other tools were also used, and IRL humans had to do all of said curation and orchestration.

The AI reverends “preached” about overcoming fear of death and trusting in Jesus Christ through an avatar on screen. While some attendees found the experiment interesting, others felt the AI lacked emotion and spirituality essential for a sermon (This is my shocked face.).

I take a bit of umbrage with a bit of the commentary sourced for the article. Some researchers noted that while AI can help make religious services more accessible, it cannot replace the personal connection and knowledge that human pastors have with their congregations; and, that “Artificial intelligence has limits in implementing it within religion”. Having spent many a Sunday and Wednesday in pews/seats at a diverse array of church liturgical styles, I can 100% see ChatGPT being [ab]used by sermonators to provide content for an entire service. I mean, we already had SermonGPT, and there are tons of (for now) human-curated sermon and service content generation farms like this one.

For the past few CMU lectures (the courses are aimed at budding chief information security officers and chief data officers) I've been stressing the need to educate humans in organizations about LLMs/GPTs; including their limitations and potential areas of trouble. I'd like to make the same charge to readers who may also be pew (etc.)-dwellers to have a chat or three with folks in faith leadership with the same education focus in mind. I fear that there will be even more misuse of our nascent AI tools to do more harm than good, especially from some things I've seen in various forums.

WTHeck Is Happening?

I will let Dr. Jacobson's post speak mostly for itself, and further note that I have not had a chance to hit the data sources on my own to reproduce his results. If you're in a TL;DR frame, he describes three concerning climate events currently happening:

• record low Antarctic sea ice

• global temperatures exceeding 1.5 degrees Celsius above pre-industrial levels for multiple days, and

• unprecedented ocean warming, especially in the North Atlantic.

One fascinating posit by Jacobsen is that a reduction in sulfur content in shipping fuels since 2020 has led to less sulfate aerosols. This was supposed to be A Good Thing™ for the environment, but it turns out that it has also reduced global dimming, which has helped fuel ocean warming, especially in shipping routes (see below).

I try not to be a climate “alarmist”, but I've been an amateur weather person for a few decades, now, so the post has passed the initial “smell test”. Many of you may have seen this already (I heard it “went viral” but I'm not plugged into the ZOMGOSH zeitgeist as much anymore); but, whether it's your first view or your re-view, I'd 💙 to hear your thoughts on it in the comments.

Moar Perplexity

As usual, I get nothing for making notes about a freemium offering, but I thought it might be helpful to share that I'm thinking of switching my $20.00 AI Tax to Perplexity.ai from ChatGPT-proper. They keep adding new features, and now have a “pro” tier (that costs the same as ChatGPT+) with far better internet-sourcing and citing capabilities. As I was out of the loop all of last week, I asked it to catch me up on what was happening in the context of my work-work, and it did a bang up job. It even has a new feature that lets you provide some background information on yourself to help it tailor answers.

Their iOS app also seems to be a bit less skeezy w/r/t “permissions” than ChatGPT's is.

FIN

Two weeks of June is almost gone! ☮