if GPT-3 can answer the essay questions you've assigned as homework then you've learned that your essay questions were optimized for producing bullshit, in the specific sense of words that sound good but are indifferent to their own truth or falsehood
Conversation
spending a decade training students to orient towards writing as mainly about bullshit is not the profound education in the humanities some people want it to be. any actual writing ability i have now i trained on MSN messenger, livejournal, forums, social media
1
17
195
only has a job as a writing coach because school has so thoroughly destroyed so many people's relationships to writing. people who have entire books in them live in fear of Doing The Homework Wrong
Quote Tweet
Replying to @QiaochuYuan
I’ve worked with at least three people who are Pulitzer level talents just quietly wondering whether they should disturb the universe
1
6
135
writing in lowercase is magic because it doesn't activate the Uppercase Context. Uppercase is for School and Work; in other words, it's writing someone else is forcing you to do. lowercase writing is *consensual*
Quote Tweet
i wrote a thing
it's about how writing in lowercase has done magical, inexplicable things for my creative output
ungated.media/article/lowerc
Show this thread
6
38
205
in this way i, for one, welcome our machine overlords. if GPT-4 destroys the essay, or the concept of homework entirely, i will cheer and applaud raucously. we do not need them and we never did
2
18
187
more on GPT-3 as a bullshit engine
Quote Tweet
Linguistic context *correlates* with real-world context, so assembling extremely large amounts of it gives a simulacrum of semantics. Except when it doesn’t, revealing that current language models have zero idea what they are talking about: aiweirdness.com/baltimore-orio
Show this thread
1
8
59
GPT-3 apparently surprisingly good at generating marketing copy. hmmmm
Quote Tweet
An interesting thing about working with GPT-3 is that you have to fact check /everything/, because it excels at generating convincing text but makes no promises that the facts in that text will be accurate in any way
Show this thread
3
3
49
"The thing to know about the AI language models, OpenAI’s GPT-3 and its cousins, is that they are fundamentally bullshitters. The bullshit has gotten better and better, but at the core … well, there’s nothing at the core."
Quote Tweet
Doh, don't know why I didn't spot this glaring connection between ML generated outputs and Frankfurtian bullshit. robinsloan.com/lab/notes-on-a
Show this thread
2
2
16
this feels like it goes here: are greentexts bullshit?
Quote Tweet
Also what does it imply that, out of anything I've seen it try to do, GPT-3 is easily the best at greentext
Like, it's a competent sophomore at poetry, a vague freshman at essaycraft, but a PhD in greentext twitter.com/sashachapin/st…
3
8
greentexts are the perihelion of bullshit
1
1
Replying to
probably? mostly it just felt fun to write
"closest point to the sun" feels laudatory, got a little Icarus in there for spice just to make it a bit ambiguous
1