As AI gets more powerful, the bottleneck in human-AI collaboration can become the input channel between the human and the AI, rather than the capabilities of the AI itself. This bottleneck can get especially severe with longer context windows.
PROMPT ENGINEERING
-
GPT-4 Improves Output With Encouragement and Contextual Feedback
By
–
GPT-4 suddenly understands when you tell it “you can do better”. More context is obviously the only hack you need but “you can do better” is a quick feedback that reminds it that it’s capable. Often helpful in open-ended creative generation where you want different variations in
-
10 Essential Things to Know About LLMs and Fine-tuning
By
–
From #finetuning to the latest open-source #LLM models – 10 things you need to know about #LLMs. Join our upcoming episode of #ML Real Talk to get all of your questions answered. https://
my.demio.com/ref/on74p73dZ7
uUP73o?utm_source=twitter
… -
Self-Instruct: Early Stopping for Minimal Instruction Tuning
By
–
Becoming self-instruct: introducing early stopping criteria for minimal instruct tuning paper page: https://
huggingface.co/papers/2307.03
692
… introduce the Instruction Following Score (IFS), a metric that detects language models' ability to follow instructions. The metric has a dual purpose. -
GPT4RoI: Instruction Tuning LLM on Region-of-Interest
By
–
GPT4RoI: Instruction Tuning Large Language Model on Region-of-Interest paper page: https://
huggingface.co/papers/2307.03
601
… Instruction tuning large language model (LLM) on image-text pairs has achieved unprecedented vision-language multimodal abilities. However, their vision-language -
Prompt and Process with Ethan Mollick AI Miniseries
By
–
Prompt and Process with Ethan Mollick (AI Miniseries)
#AI #GenerativeAI #machinelearning
CC @Khulood_Almani @amalmerzouk @Analytics_699 @MargaretSiegien @baski_LA @Shi4Tech @CurieuxExplorer @KanezaDiane @ChuckDBrooks @FrRonconi @ingliguori @efipm https://
open.spotify.com/episode/4kbOCD
FbFqYLZpmLHEsAsD?si=uesawsgsS22m6izDVucMPw
… -
ChatGPT speeds up development but complicates debugging
By
–
Before Chatgpt Development : 5 hours
Debugging : 3 hours After Chatgpt Development : 5 min
Debugging : 8 hours I hope u have understood : #chatgpt #ai -
LlamaIndex Dismissed as Noise by LLM Practitioners
By
–
LlamaIndex is something i see around so much but never figured out what it does. i've always considered it noise. A general rule of thumb is anyone serious about LLMs should not touch either of these.
-
Code Interpreter Allows You to Visualize Source Code of Results
By
–
Code Interpreter also allows you to see the code that produced the output
-
Code Interpreter Review by Ethan Mollick: Experiments and Analysis
By
–
Here is a nice review about Code Interpreter by @emollick
. Well written and lots of experiments there! https://
oneusefulthing.org/p/what-ai-can-
do-with-a-toolbox-getting
…