Procedural / Generative / LLM Text, Poetry, etc.

anon45924993 · 2023-04-06T14:31:26.108Z

I think a large portion of the population has always regarded computing technology as mysterious and unreliable. Certainly similar to a haunting but I think such folks experience it as mainly irritating.

It’s possible that sense of irritation is the thing that will become more common. If it reduces our collective screen time, sending us back into the real world (also a chaotic complex system with emergent properties, but hopefully more reliable for most of us most of the time, barring climate change…) it might not be the worst outcome.

Cementimental · 2023-04-06T17:37:37.164Z

We still don’t even have a name for the feeling of vague anxiety at having something control-C’d and not yet pasted

synthetivv · 2023-04-06T18:11:58.553Z

This state is known as antepasto.

alanza · 2023-04-10T16:17:56.234Z

Machine-generated code General

I share a lot of feelings in this thread… suspicion and distrust of the corporations behind the new AI tools delight for the folks who find programming accessible for the first time horror at what might happen to communities formed around open, text-based communication now that text synthesis is a thing (for a glimpse into the future, look at Tinder) hope that this might make software development more fun Personally my one ChatGPT success has been, using it to generate an explanation of exist…

@vladimirchicken Somehow this thread, as opposed to Machine-generated code feels like the place to respond to your comment. Not that I have a ton to say, but I imagine we’re about to experience an odder version of Eternal September—just as the influx of new users from September 1993 never slowed enough for them to become acculturated into the norms of the then-existing internet communities they overran, so too are we likely to see a degradation of certain forms of being online.

Which is to say, there’s a sense of loss there, but it’s surprisingly precedented. I remember being a child and getting a sense of, like, archaeological wonder when I came across old Usenet archives. Are those still online? Who knows. It’ll be interesting to see what the internet is like in a few years if things continue apace.

That being said, I’m not convinced they will. AI-generated art hasn’t gone away, but it also hasn’t completely upended the world from what I can see. NFTs sure were a flash in the pan. Maybe SALAMI (you gotta love this acronym) will be too.

terribleben · 2023-04-17T17:31:07.493Z

The New Yorker – 9 Feb 23

ChatGPT Is a Blurry JPEG of the Web

OpenAI’s chatbot offers paraphrases, whereas Google offers quotes. Which do we prefer?

Ted Chiang (scifi author whose work was used for Arrival among other things) with a thoughtful essay. Further in, he touches on how and whether to use such a tool for creative writing. This captures a lot of what I’ve personally felt but had a tough time articulating about the recent LLM hype.

anon45924993 · 2023-08-23T20:10:16.106Z

Like I said, people get bored with the regression to statistical mean that generative AI represents.

honest-broker.com

Ugly Numbers from Microsoft and ChatGPT Reveal that AI Demand is Already...

The only areas where AI is flourishing are shamming, spamming & scamming

Kamil · 2023-09-01T08:41:36.021Z

Thanks for sharing this, interesting read.

swhic · 2023-09-01T13:47:54.526Z

AI might be part of evolution. Species adapt and evolve, Cyborgs and all the in between are possible.

anon45924993 · 2023-09-01T13:59:56.505Z

Sure, in the same way our use of statistics is “part of evolution”. Which, really, is what we are actually talking about when we talk about generative AI. It really should be called “applied linguistic statistics”.

anon45924993 · 2023-09-22T19:17:08.975Z

Continuing some thoughts from the art thread, because this is more specifically about LLMs, seems to fit better here.

From the Wired article Confessions of a Viral AI Writer:

A thought experiment occurred to me at some point, a way to disentangle AI’s creative potential from its commercial potential: What if a band of diverse, anti-capitalist writers and developers got together and created their own language model, trained only on words provided with the explicit consent of the authors for the sole purpose of using the model as a creative tool?

That is, what if you could build an AI model that elegantly sidestepped all the ethical problems that seem inherent to AI: the lack of consent in training, the reinforcement of bias, the poorly paid gig workforce supporting it, the cheapening of artists’ labor? I imagined how rich and beautiful a model like this could be. I fantasized about the emergence of new forms of communal creative expression through human interaction with this model.

Then I thought about the resources you’d need to build it: prohibitively high, for the foreseeable future and maybe forevermore, for my hypothetical cadre of anti-capitalists. I thought about how reserving the model for writers would require policing who’s a writer and who’s not. And I thought about how, if we were to commit to our stance, we would have to prohibit the use of the model to generate individual profit for ourselves, and that this would not be practicable for any of us. My model, then, would be impossible.

staysh · 2023-09-22T20:40:22.785Z

The mere speculation that you might have to police who is a writer and who is not, maybe stay commited to your anti-capitalist stance and the notion of reserving something for “writers”, to me, speaks volumes about the misconceptions everyone has about what everything is and how these immediately obfuscate expectations about encoding or automating information based tasks.

“I imagined how rich and beautiful a model like this could be”

This reads like the first pane of the Donnie Darko psychiatrist meme to me…

I don’t think it’s quite as impossible as stated above, but, I think it fails on it’s reason to exist way before cloud compute costs.

WilliamHazard · 2023-09-24T03:05:29.953Z

last week, my undergrads and I spent some time making found poems out of fragments from Anne Carson’s If Not, Winter: Fragments of Sappho, and after class, a student and I chatted about the possibility of automating our process with a little script. Inspired by that conversation, I threw together this little lua script for shuffling the lines of an input poem. I tried writing it as a shell script first but found that to be a pain, and then I realized that it’d be much easier to write in lua and that I already knew most of what I’d need to know to write it from writing schicksalslied. This experience has given me a new level of appreciation for lua. It’s an elegant, fun language, and I’m grateful for norns having introduced me to it.

robbie · 2023-09-24T03:29:16.037Z

WilliamHazard:

writing it as a shell script

On most* Linux distros (including norns) you can simply do:
shuf input.txt
or…
shuf input.txt --output shuffled.txt

(*shuf is a part of GNU coreutils)

On macOS (and probably many BSD derivatives) you can use sort:
sort input.txt --sort random
or…
sort input.txt --sort random --output shuffled.txt

Edit: see note below on a quirk of using sort vs. shuf

WilliamHazard · 2023-09-24T03:33:45.895Z

I tried this with bash on my Mac, but it didn’t seem to have any idea what shuf was. I’ll have to give it a try once I’ve got my ‘07 MBP back up and running with Ubuntu!

EDIT:

robbie:

sort input.txt --sort random --output shuffled.txt

oh damn, yeah, this works!

robbie · 2023-09-24T03:48:59.782Z

Yeah, on macOS you won’t have shuf (unless you have used homebrew, or similar, to install gnu coreutils) but sort provides similar functionality (see the note above).

As a bonus, which might be interesting for longer input works, you could also use shuf --head-count 5 to limit the number of outputted lines – in this example to a total of 5 – as opposed to including all supplied lines. sort doesn’t offer this built-in, but you can always do sort input.txt | head --lines 5 or similar.

WilliamHazard · 2023-09-24T17:30:30.405Z

the first bonus idea that struck my fancy was iterating through a folder of .txt files, putting them all into one .txt file, and shuffling that file, but I immediately ran into some trouble (which I’ve also run into in other experiments with shell scripts, so if anyone has an answer for this, it would be very much appreciated). When I try to do a for loop to iterate through the folder of original .txt files, it seems to want to treat all of them as one thing instead of moving through each one individually, doing an operation on it, and then moving on to the next one. What I tried is

list=$(ls -r originals/*.txt)
echo "shuffling"
echo "" > shuffled.txt
for file in $list ; do
    cat $file >> shuffled.txt
done
sort shuffled.txt --sort random --output shuffled.txt
echo "shuffled"

and what I’m getting is

shuffling
cat: originals/sappho hymn to aphrodite IV.txt
originals/sappho hymn to aphrodite III.txt
originals/sappho hymn to aphrodite II.txt
originals/sappho hymn to aphrodite I.txt: No such file or directory
shuffled

That’s a list of files I have in the originals folder, as the first line calls for, but it doesn’t seem to be looking at each file individually. Instead, it seems to be looking for one file whose filename is that whole list (and, of course, not finding it). I’ve futzed around with it for a couple hours without any luck. Anyone have any idea what I’m doing wrong here?

EDIT: got this solved with some much-needed help from @alanza. The problem was the ls. This

echo "clearing shuffled.txt"
cat /dev/null > shuffled.txt
echo "shuffling"
for file in originals/*.txt ; do
    cat $file >> shuffled.txt
done
sort shuffled.txt --sort random --output shuffled.txt
echo "shuffled"
echo "result:"
cat shuffled.txt

works as expected

robbie · 2023-09-24T19:43:59.345Z

With shell scripts, it’s good to remember you can always use pipes to efficiently pass data from tool to tool without having to worry about creating intermediary files.
By combining the glob (the *.txt wildcard) and cat together with a pipe to sort you can reduce the script to a single, still readable, line:

cat originals/*.txt | sort --sort random --output shuffled.txt

(Obviously this isn’t so helpful if you want to do more than a single operation on each file, in which case the for loop is probably much more practical.)

Word Cut-up

Made a quick rework of the above code to do per word cut-up, instead of per line.

cat originals/*.txt | tr ' ' '\n' | sort --random-sort --unique | paste -sd ' \n \n \n' -

Script breakdown

cat originals/*.txt
Merges the contents of all .txt files in the originals directory into a single continuous piece of text (as before).

tr ' ' '\n'
Replaces each of the first character with the second. Here every space (' ') becomes a newline character (\n), so there is only one word (plus punctuation) per line.

sort --random-sort --unique
Randomly reorders every line (as before). --unique removes any duplicate lines (see note below).

paste -s -d ' \n \n \n' -
Rejoins all the separate lines together using the characters given after the -d. Each character is used once per line; if all the characters have been used the pattern is repeated from the beginning.
In this example, using spaces and newlines (\n as before), the pattern is:

Five spaced words, newline,
Five spaced words, newline,
Three words, newline

The final - simply tells paste to take input from the previous command (ie. the output of sort)

Example output

around, not – clocks And
A sweeps sleepy nothing of
Who The meets.
wraps quiet, snow only comforts
dear, There’s falling din
’Most everyone away
the that’s city I’ve all
sit, soft watch, In world
slow. known hear.
souls sound at move like
ever and That embrace, makes
it streets, noise

`sort` vs. `shuf` redux

Working through this I realised the behaviour of shuf and sort --sort random are not exactly the same (The Revenge of rtfm). shuf treats every line as unique, no matter the content, completely randomising line order. sort also randomly reorders lines, except any identical lines are individually grouped together as a block. Not immediately obvious when sorting sentences, but it is when sorting single words. There doesn’t seem to be a way to change this behaviour, so as a quick-fix here I added --unique which discards any identical lines and side-steps the problem.

Blueboy · 2023-11-26T12:19:02.194Z

I’m pretty sure that i will never use chatGPT, I’m agnostic but I’m with Nick on this -

Edit to add no disrespect to anyone here who does use chatGPT, i realise it is just a tool just not a tool for me.

JosueArias · 2023-11-26T12:51:47.968Z

Thanks @blueboy for bring that speech, that resumed what I “feel” and somehow I know, about GPT and other so called tools. These are not only tools , they erode something that belongs to all of us. Art, Love, nature, humanity …

oliodnb · 2024-01-21T12:42:41.333Z

Interesting read about the use of generative tools as an aid to unlock creative potential.