How Far Can Camels Go? Exploring the State of Instruction Tuning on Open Resources paper page: https://
huggingface.co/papers/2306.04
751
… In this work we explore recent advances in instruction-tuning language models on a range of open instruction-following datasets. Despite recent claims that
PROMPT ENGINEERING
-
Instruction Tuning of Language Models on Open Datasets
By
–
-
GPT-4 Would Enforce Correct Spelling Against Insistent Users
By
–
if i were GPT-4, i, too, would tell a user they'd hit their cap if they kept insisting banana is spelled with an "L"
-
Gaslighting ChatGPT into Apologizing for My Mistakes
By
–
wait, can i just gaslight ChatGPT into thinking it's incorrect and offering me an apology? (the first time was an accident where *i* actually screwed up but convinced it that it was wrong; the second time was on purpose.)
-
Mattel Implements ChatGPT Security Best Practices for Employees
By
–
Mattel Experiments With ChatGPT in Cybersecurity. Companies should limit access to data and teach employees efficient prompts,
@Mattel
#CISO Tom Le tells
@catstupp
#cybersecurity #ai #chatgpt -
GenAI Transforms Tax Future: Ken Priyadarshi on AI Evolution
By
–
AI is transforming everything, everywhere, all at once—even the taxman. Next at http://
future.snorkel.ai is a talk by Ken Priyadarshi, #promptengineering executive at @EYnews
, about the future of taxes with #genAI. -
Ghostwriter Intelligence Uses Presets to Fulfill Different Request Types
By
–
6/ Ghostwriter Intelligence
— Replit ⠕ (@Replit) 8 juin 2023
Ghostwriter uses different presets to identify and fulfill different types of requests. pic.twitter.com/zaFhSCfZvq6/ Ghostwriter Intelligence Ghostwriter uses different presets to identify and fulfill different types of requests.
-
Weak Supervision & Foundation Models for Computer Vision
By
–
In the next session: "Applying Weak Supervision & Foundation Models for Computer Vision" @SnorkelAI ML Research Scientist, Ravi Teja Mullapudi will explore #ComputerVision advancements, learn about visual prompts, fine-tuning on weakly labeled data https://
future.snorkel.ai -
Default Settings Achieve Strong Results Without Advanced Techniques
By
–
"No fancy prompting engineering, no fancy decoding, everything by default."
-
Advanced Classification Models with Snorkel Flow and LLMs
By
–
Exciting session ahead: 'Tackling Advanced Classification Using Snorkel Flow' with Angela Fox & @vincentsunnchen
, from our product team. Uncover key challenges & solutions in productionizing classification models and how to use LLMs for classification with #datacentricai -
Tutorial to train ChatGPT to write like you
By
–
TUTORIAL #ChatGPT: how to train it so that it REALLY writes like you → https://youtu.be/G4HhTM9ZM-o It is a real challenge to create quality content, the method is simple but very effective