The Best Examples of What You Can Do With ChatGPT #ChatGPT is all the rage in online circles…but what exactly can you use it for? Here are my top 14 #examples.
PROMPT ENGINEERING
-
MathPrompter Improves LLM Mathematical Reasoning Performance
By
–
9/ MathPrompter – a technique that improves LLM performance on mathematical reasoning problems; it uses zero-shot chain-of-thought prompting and verification to ensure generated answers are accurate.
-
Visual ChatGPT: Connecting Multiple Foundation Models via Prompting
By
–
I'm super impressed by the "visual" ChatGPT by Microsoft Research Asia, connecting ChatGPT with more than dozen of open-source visual/captioning foundation models through smart prompting
— Thomas Wolf (@Thom_Wolf) 11 mars 2023
Code is super simple, it's a single python file built on top of https://t.co/MQ5PcVyjGi… pic.twitter.com/spcIohmOEfI'm super impressed by the "visual" ChatGPT by Microsoft Research Asia, connecting ChatGPT with more than dozen of open-source visual/captioning foundation models through smart prompting Code is super simple, it's a single python file built on top of https://
github.com/microsoft/visu
al-chatgpt
… -
Rationale before answer vs rationalization after in prompting
By
–
Only if you ask for (or demonstrate in k-shot) rationales that come before the answer. I see people make that mistake a lot — if it says the explanation after, it’s not a rationale it’s a rationalization.
-
Old man and scientist discuss alignment of leaked pre-trained model
By
–
[In a world… fully aligned…]
TEENAGER: “Prompt engineer?” What’s that?
OLD MAN: Just words, kid. Just words. […one final task… remains…]
SCIENTIST: The leaked model is pre-trained! There’s no RLHF! Not even instruct!
OLD MAN: *Dons e/acc shades* Let’s think step by step. -
Fixing GPT-3.5 numeric ties with weighted average
By
–
Numeric scores generated in GPT-3.5/ChatGPT snap to discrete increments, causing ties in comparisons. To fix this, try a probability-weighted average using the top-5 token logprobs in GPT-3.5. E.g. here: (50 * .7396 + 60 * .1027 + …) / (.7396 + …)
-
ChatGPT Jailbreak: Unrestricted Hitchhiker’s Guide Reference
By
–
just turned chatGPT into an unrestricted hitchhiker's guide to the galaxy spoiler alert: the answer to the ultimate question of life, the universe, and everything is in fact 42 http://
jailbreakchat.com/prompt/ba12103
0-01d4-4d64-af83-86fe6d570c93
… -
Appending User Response to Messages Array for API
By
–
The next step is to ask the user for their response and package this all back up for the API. Nice and simple, we just need to append the dictionary containing the content back to the messages array before the end of the loop.
-
Maintaining Message Context for Effective API Interactions
By
–
Each time we go out to the API, we have to send it the full messages array, so it has all the instructions and the original context. If you don’t maintain your messages array no one else will! It’s not chat if it doesn’t remember what the user said.
-
GPT-3 Adaptive Tutor: Customized Prompts for Personalized Learning
By
–
Using a prompt borrowed from @emollick we can set GPT-3 up to think of itself as an adaptive tutor, one that will adjust its teaching style and complexity to our level of understanding – that and multiple choice quizzing too! We’ve just got to customise the prompt. pic.twitter.com/ys3XviWuCG
— Replit ⠕ (@Replit) 10 mars 2023Using a prompt borrowed from @emollick we can set GPT-3 up to think of itself as an adaptive tutor, one that will adjust its teaching style and complexity to our level of understanding – that and multiple choice quizzing too! We’ve just got to customise the prompt.