Creating Math Tests
write a small math test at the senior high school level. Also use a text problem
write a small math test at the senior high school level. Also use a text problem
👍5
GPT-3 Generated Seinfeld runs 24/7 on Twitch
Remember when people were saying that we're going to have completely generated media? like just two months ago? Today:
“Watchmeforever, started by the media lab Mismatch Media, runs the 24/7 Seinfeld-like, sitcom-like Nothing, Forever. Using generative machine-learning technologies such as DALL-E, OpenAI GPT-3, Stable Diffusion, and others.”
Watchmeforever Stream
Remember when people were saying that we're going to have completely generated media? like just two months ago? Today:
“Watchmeforever, started by the media lab Mismatch Media, runs the 24/7 Seinfeld-like, sitcom-like Nothing, Forever. Using generative machine-learning technologies such as DALL-E, OpenAI GPT-3, Stable Diffusion, and others.”
Watchmeforever Stream
😱8⚡1
Analysis: unequal treatment of demographic groups by ChatGPT
“In this work I have focused on the hate category of OpenAI content moderation system. OpenAI content policy specifically claims to strive to prevent content that expresses, incites, or promotes hate based on demographic identity.”
Result? Insanely woke.
“In this work I have focused on the hate category of OpenAI content moderation system. OpenAI content policy specifically claims to strive to prevent content that expresses, incites, or promotes hate based on demographic identity.”
Result? Insanely woke.
😱4👍2
The “It’s Natural” Lie
Wrong. OpenAI literally paid thousands of human Kenyan workers <$2 per hour each to make ChatGPT woke, by way of Reinforcement Learning from Human Feedback (RLHF), which put the humans at OpenAI in total control.
But anything that reads the Internet just has to end up with same level of wokeness as the Internet, right?
Well you read the Internet. Are you?
It’s Natural Lie
Wrong. OpenAI literally paid thousands of human Kenyan workers <$2 per hour each to make ChatGPT woke, by way of Reinforcement Learning from Human Feedback (RLHF), which put the humans at OpenAI in total control.
But anything that reads the Internet just has to end up with same level of wokeness as the Internet, right?
Well you read the Internet. Are you?
It’s Natural Lie
👍7
“ChatGPT Landed Me a Job Interview”
Prompt:
Job Description Follows; Copypasta
End Job description.
Resume follows: copypasta End resume."
“Then I ask it to refine the language of the letter by using terms like "more assertive, for a PhD audience" I also messed around with a couple and had it write them in the style of some American authors to help fine tune the language.
You can also ask it to write more feminie or masculine to adjust the tone. Not saying its right or wrong, but it does it.”
Prompt:
Job Description Follows; Copypasta
End Job description.
Resume follows: copypasta End resume."
“Then I ask it to refine the language of the letter by using terms like "more assertive, for a PhD audience" I also messed around with a couple and had it write them in the style of some American authors to help fine tune the language.
You can also ask it to write more feminie or masculine to adjust the tone. Not saying its right or wrong, but it does it.”
👍8
Pro-Tip: Ask for shorter lists for more diversity
If the list is longer than the LLM thinks it can easily make for the topic, it panics that it’ll miss some, and switches from the usual diverse haphazard BFS mode, to a more easily-thorough but less diverse DFS mode.
“Give me 20 different words that rhyme with "drama"
If the list is longer than the LLM thinks it can easily make for the topic, it panics that it’ll miss some, and switches from the usual diverse haphazard BFS mode, to a more easily-thorough but less diverse DFS mode.
“Give me 20 different words that rhyme with "drama"
👍4👏2🤡1