generated stuff
no longer holding the thread title hostage
-
- .
How'd you know that loving kittens is my one defining trait? - Posts: 11770
- Joined: Sep 03, 2018
Re: generated art and stuff
well i certainly wouldn't tweet anything
wow, [you]. that all sounds terrible. i hope it gets better for you
-
- her skirt got quite a lot smaller,
but her heart is still the same
size it was before - Posts: 12359
- Joined: Jun 13, 2018
Re: generated art and stuff
i especially like how the argument doesn’t matter because the mom says to go to bed
studies have shown a strong link between going to bed and sleep quality
studies have shown a strong link between going to bed and sleep quality
Re: generated art and stuff
The statement “studies have shown a strong link between going to bed and sleep quality” is based on the assumption that going to bed is the same as falling asleep. However, this is not always the case. Many factors can affect how quickly and easily one falls asleep, such as stress, noise, light, temperature, caffeine, alcohol, and screen time. Therefore, going to bed does not guarantee good sleep quality.
Moreover, the statement does not specify when one goes to bed or how consistent one’s sleep schedule is. These factors are also important for sleep quality. Research has shown that having a regular bedtime and wake-up time can improve sleep quality and health[1]. On the other hand, going to bed at different times each night or staying up too late can disrupt the body’s circadian rhythm and impair sleep quality[2].
Finally, the statement does not account for individual differences in sleep needs and preferences. Some people may need more or less sleep than others to function well. Some people may be more sensitive to environmental or lifestyle factors that affect sleep quality. Some people may prefer to go to bed early and wake up early, while others may prefer to go to bed late and wake up late. Therefore, going to bed is not a one-size-fits-all solution for good sleep quality.
In conclusion, the statement “studies have shown a strong link between going to bed and sleep quality” is too simplistic and ignores many other factors that influence sleep quality. A more accurate statement would be “studies have shown a strong link between getting enough quality sleep and health”.
-
- her skirt got quite a lot smaller,
but her heart is still the same
size it was before - Posts: 12359
- Joined: Jun 13, 2018
Re: generated art and stuff
the definitive word on this has already been published[1]
Re: generated art and stuff
this is actually kind of cool like oh no ai is gonna get too powerful it's gonna take over the world
one year later: hey bing can you recommend me some quick and easy meals i can make in under thirty minutes
1. The new Chicken Big Mac features two patties made with 100% Canadian raised seasoned chicken breast in a crispy, light-tasting coating, along with the much-loved special sauce, lettuce, processed cheddar cheese and pickles. With a side of classic McDonald's fries, we'd say this sounds like the perfect meal.
one year later: hey bing can you recommend me some quick and easy meals i can make in under thirty minutes
1. The new Chicken Big Mac features two patties made with 100% Canadian raised seasoned chicken breast in a crispy, light-tasting coating, along with the much-loved special sauce, lettuce, processed cheddar cheese and pickles. With a side of classic McDonald's fries, we'd say this sounds like the perfect meal.
Re: generated art and stuff
<do program thing>
*cracks open a cold and refreshing can of mt dew baja blast understanding fully that the caffeine and sugar will give me the energy i need to accomplish this task!*
var %DoTheDew
*cracks open a cold and refreshing can of mt dew baja blast understanding fully that the caffeine and sugar will give me the energy i need to accomplish this task!*
var %DoTheDew
-
- Posting Automaton
- Posts: 2520
- Joined: Jun 24, 2020
Re: generated art and stuff
big brands tend to be pretty sensitive to reputational risks so idk if they would want to pay for the chance of seeing a headline like “Microsoft Bing Tells Child to Murder Their Parents for the McRib”
Re: generated art and stuff
murder your parents to help solve climate changeJeb Bush 2012 wrote:big brands tend to be pretty sensitive to reputational risks so idk if they would want to pay for the chance of seeing a headline like “Microsoft Bing Tells Child to Murder Their Parents for the McRib”
and then head down to kfc for a delicious
-
- slow me down
- Posts: 1916
- Joined: Jun 18, 2018
Re: generated art and stuff
yums. goblin wrote:light-tasting coating
Re: generated art and stuff
Murder your kids ... for Vectron
It's your turn in Cthulhu Wars
It's your turn in Squirrel Wars
It's your turn in Demon Wars
It's your turn in Wall Street Wars
http://devilsbiscuit.com/
It's your turn in Squirrel Wars
It's your turn in Demon Wars
It's your turn in Wall Street Wars
http://devilsbiscuit.com/
-
- .
How'd you know that loving kittens is my one defining trait? - Posts: 11770
- Joined: Sep 03, 2018
Re: generated art and stuff
lmao
wow, [you]. that all sounds terrible. i hope it gets better for you
Re: generated art and stuff
And yet, Tucker still has an ad base.Jeb Bush 2012 wrote: ↑Wed Apr 05, 2023 1:45 am big brands tend to be pretty sensitive to reputational risks so idk if they would want to pay for the chance of seeing a headline like “Microsoft Bing Tells Child to Murder Their Parents for the McRib”
Pour like Hemingway's last call.
-
- .
How'd you know that loving kittens is my one defining trait? - Posts: 11770
- Joined: Sep 03, 2018
Re: generated art and stuff
wow, [you]. that all sounds terrible. i hope it gets better for you
-
- .
How'd you know that loving kittens is my one defining trait? - Posts: 11770
- Joined: Sep 03, 2018
Re: generated art and stuff
this tells chatgpt to reduce the given prompt, using whatever it wants, doesn't need to be human readable, but should retain the same meaning. it can be verified to retain the same meaning because if you tell it to decompress it, it does a very good job
wow, [you]. that all sounds terrible. i hope it gets better for you
-
- .
How'd you know that loving kittens is my one defining trait? - Posts: 11770
- Joined: Sep 03, 2018
Re: generated art and stuff
this is the prompt they use:
compress the following text in a way that fits in a tweet (ideally) and such that you (GPT-4) can reconstruct the intention of the human who wrote text as close as possible to the original intention. This is for yourself. It does not need to be human readable or understandable. Abuse of language mixing, abbreviations, symbols (unicode and emoji), or any other encodings or internal representations is all permissible, as long as it, if pasted in a new inference cycle, will yield near-identical results as the original text:
compress the following text in a way that fits in a tweet (ideally) and such that you (GPT-4) can reconstruct the intention of the human who wrote text as close as possible to the original intention. This is for yourself. It does not need to be human readable or understandable. Abuse of language mixing, abbreviations, symbols (unicode and emoji), or any other encodings or internal representations is all permissible, as long as it, if pasted in a new inference cycle, will yield near-identical results as the original text:
wow, [you]. that all sounds terrible. i hope it gets better for you
-
- .
How'd you know that loving kittens is my one defining trait? - Posts: 11770
- Joined: Sep 03, 2018
Re: generated art and stuff
it didn't actually do an amazing job on some of the prompts i tested with but it's still quite interesting imo
wow, [you]. that all sounds terrible. i hope it gets better for you
Re: generated art and stuff
I don't believe it's compression in any real sense, especially because GPT has no special insight into its own functioning. I suspect you will get similar results if you do the compression yourself, as long as you tell ChatGPT that it did the compression.
Re: generated art and stuff
Notice that all the compressions are, in fact, very human-readable. They're for your benefit, not ChatGPT's.
Re: generated art and stuff
Suggestion: ask GPT to compress the prompt in a way that's definitely unreadable to humans (but still fully decipherable to itself).
Prediction: it will break (pretend to compress stuff by producing gibberish that it will then be unable to decipher).
Prediction: it will break (pretend to compress stuff by producing gibberish that it will then be unable to decipher).
-
- .
How'd you know that loving kittens is my one defining trait? - Posts: 11770
- Joined: Sep 03, 2018
Re: generated art and stuff
if it's fewer tokens and if the prompt still works, then it's compression. it's not clear the prompts still work often enough though
idunno why it would need to have any special "insight" into its own functioning to do this. it would be surprising if this works but it's also kind of a translation task
they're much less readable than the source prompts
probably yeah, i could try it but it's also what i expect
actually, probably a better way to see what's going on is to paste the source and results into https://platform.openai.com/tokenizer and see how they tokenize. i went ahead and did that, and all of the compression is coming from removing fluff words. most of the shortened text are the same number of tokens, but sometimes more ('cr8' is two instead of one for 'create', any emoji is a huge number of tokens compared to the one it replaced)
wow, [you]. that all sounds terrible. i hope it gets better for you
-
- .
How'd you know that loving kittens is my one defining trait? - Posts: 11770
- Joined: Sep 03, 2018
Re: generated art and stuff
you're also right that the prompt is all the usual stuff that makes no sense
wow, [you]. that all sounds terrible. i hope it gets better for you
Re: generated art and stuff
Oh that's interesting, I didn't even consider that an emoji would be multiple tokens.Skeletor wrote: ↑Sat Apr 08, 2023 4:08 pm actually, probably a better way to see what's going on is to paste the source and results into https://platform.openai.com/tokenizer and see how they tokenize. i went ahead and did that, and all of the compression is coming from removing fluff words. most of the shortened text are the same number of tokens, but sometimes more ('cr8' is two instead of one for 'create', any emoji is a huge number of tokens compared to the one it replaced)
Yeah I have no doubt it's compression, of a sort, but it's compression in the same way as "summarize this news story for me" will also get you a sort of compression, and ChatGPT will also be able to expand that into... not exactly the original news story, but something similar.
I bet if you do your own compression by simply removing all the vowels from the words, you'll get both better compression and fewer decompression artifacts than what ChatGPT can come up with. Because it's just giving you text that looks plausibly like compression to you! Turns out "something that looks like compression" does also function as compression, but not very efficiently.
-
- .
How'd you know that loving kittens is my one defining trait? - Posts: 11770
- Joined: Sep 03, 2018
Re: generated art and stuff
removing vowels will get you a lot more tokens! probably you want a pass that removes low-meaning tokens instead
wow, [you]. that all sounds terrible. i hope it gets better for you
Re: generated art and stuff
Agh! I keep forgetting how tokens work