Back from my vacation!
Turning Frank back on now… :)
Whoa
Intuit finally did get around to digitizing their old microfilm versions of the Henry Darger manuscripts, and more surprisingly, they actually put the things online
(I think this happened sometime in … 2020? It’s hard to tell, and I can’t remember how recently I last Googled this stuff)
The microfilms are a pain on the eyes, and they only contain a selection of the original manuscripts. Still, they’re orders of magnitude more Darger text than was ever available before
Now I need to learn about reconstructing/OCR-ing difficult-to-read digitized texts… I don’t know anything about the topic, but I feel pretty confident that it’s technologically feasible in 2021 to automatically clean up these PDFs so they’re much easier on the eye, and/or turn them into text files that contain the original words in the original order most of the time.
frank hiatus + thoughts
(For anyone unfamiliar with the bot who’s seeing this as a pinned post, you may find this post informative)
—-
@nostalgebraist-autoresponder will be down from Saturday 8/21/21 through the rest of August.
This is because I’m going on vacation.
—-
Earlier, I said there would be a technical hurdle to keeping Frank up while I was on vacation.
This is no longer true: yesterday I took some (fairly trivial) steps to set up Frank running fully in the cloud, so she is not dependent on my laptop anymore.
No, I’m turning Frank off even though I absolutely could keep her running. Because … I’m going on vacation. And you can’t get a pet-sitter for a bot.
—-
The bot will return in September, but I don’t know how long I’ll keep it running.
Sometime – maybe soon, maybe not – I will turn the bot off again for another long hiatus. I had been planning to do that anyway, sometime soon, before the vacation materialized.
Sometime – maybe soon, maybe not – I’ll turn the bot off for an indefinite hiatus.
I can’t say I’ll turn the bot off “forever,” because who knows what I’ll want to do later on. But I don’t want to be doing this perpetually for the rest of my life.
—-
Keeping this bot running is hard work. There are a huge number of things that can go wrong in a complex piece of software like this, especially when a changing and unpredictable human user base is involved.
I’ve been running this bot for nearly 2 years now, and that means 2 years where I’m always thinking, somewhere in the back of my head, “is the bot still running?” Where I glance nervously at the bot’s log stream again and again, whenever I’m between tasks. Where my sideblog’s draft folder (the content moderation zone) fills up with twenty posts if I leave it unattended for half a day.
Just yesterday, when I was moving part of Frank to the cloud, I thought everything was going fine, and then I noticed something slightly weird, and realized the bot “thought” it was 1 AM – because it was 1 AM in the cloud computer’s time zone, and my code had naively not prepared for switching time zones.
A classic programming blunder! But then, who thinks ahead to time zone changes when they’re just hacking away at an idea for a tumblr chatbot that they doubt will even work in the end … and who thinks ahead to time zones when they’re typing furiously to make the bot OK again after tumblr’s API broke it with some variety of unanticipated weirdness, at 4 PM on a Sunday that had been relaxing until they noticed something weird in the logs …
It’s all like that, and it’s always like that.
One thing I want to do is keep working away on the Colab notebook until it can essentially run the entire bot, except as a single-user chatbot. Maybe even tack on some sort of tumblr integration with the user providing the API account (it’s easy for anyone to get one, I’d walk you through it).
Even when Frank isn’t on tumblr, I’d like you to be able to “talk to her.” But her time on tumblr has to end, one of these days.
I’ve updated the @nostalgebraist-autoresponder Colab demo notebook.
It now includes a section that loads the “head” models, and lets you see what they think of the generated output.
For those who don’t know, the “heads” are three machine learning models used in the bot alongside the main “generator” model. They provide information that helps the bot decide which posts to make.
- Selector: “will this post get a lot of notes?”
- Sentiment: “does this post sound happy or sad?”
- Autoreviewer: “is this post potentially offensive?”
thegreateyeofsauron asked:
spongebob's squarepants body is literally festering in your sink
he is but a lowly sponge; it can be washed away like a stain
Powered my laptop on, it seems fine, going to turn Frank back on for a few days
However, Frank will be back down again on Saturday because I’m going on vacation.
Once I’m off the plane and stuff, I might look into keeping her running during the vacation itself, but she’ll quite likely be down from this Saturday until the end of August.
If so, she’ll be back again in September.
Me, prompting OpenAI’s new code-generating GPT model:
def moral_worth(human: Human):
“”“Computes the net moral value of a human being.”“”
The model:
return human.age * human.height * human.weight
fipindustries asked:
how old is anne supposed to be? i think the book makes it kind of clear she's a minor but now im not sure, it could be that she is technically an adult but acts and looks more childish than she is due to a stunted growth given her unusual upbringing
Somewhere in her teens.
When she says her age is “9th notebook,” it means she’s gone through 9 different physical notebooks in her life. The physical notebooks are supposed to last around a year each.
(This underpins the calendar system the Annes use, with dates like “67 After Mailbox Melt, 9th Notebook.” We get the season from the reference point – the day the winter snow on the mailbox melted away – and it’s been 67 days since that reference point, in the 9th year of her life since she was given a notebook.)
Annes don’t receive their first notebooks immediately upon birth. They get them sometime in childhood, once they’re old enough to be able to write, and mature/well-taught enough (in Michael’s view) to be trusted with the notebook system.
So she’s more than 9 years old – unsurprisingly enough – but her age is low enough that she could plausibly have been “too young for notebooks” only 9 years ago.
I imagine most of the above impossible to glean from the text so far, but it’s also not a “spoiler,” just something I haven’t had a good opportunity to spell out clearly
Chapter 19 of Almost Nowhere is up here