IA Intelligence artificielle - Offirmo-team/wiki GitHub Wiki

Voir aussi LLM, ChatGPT, Python, Cplusplus...

Intro

2023 Mozilla guide https://ai-guide.future.mozilla.org/

Links

Récents dévelopements:

2023 https://simonwillison.net/2023/Dec/31/ai-in-2023/
- "you can take a GIANT corpus of text, run it through a pile of GPUs, and use it to create a fascinating new kind of software."

Advocacy

Future

2023 ++ https://platforms.substack.com/p/slow-burn-ai-when-augmentation-not

Wins

Unclean

fails

Concepts

machine learning
LLM
alignment
- +++ https://www.artifact.io/iap
+++ https://cloud.google.com/products/ai/ml-comic-1/
+++ 2020 https://github.com/AMAI-GmbH/AI-Expert-Roadmap

Sécurité

!!! totalement insécure, prompt injection, data leaking...

training data poisoning
prompt injection
Adversarial attacks
test set contamination https://arxiv.org/abs/2310.17623
https://zapier.com/blog/ai-security-risks/

Utilisation

prompting

Official, permalinks

OpenAi https://platform.openai.com/docs/guides/prompt-engineering
Google AI https://ai.google.dev/docs/prompt_best_practices
microsoft https://github.com/microsoft/promptbase
2023 https://www.14islands.com/journal/the-art-of-prompting
2023/12 +++ https://minimaxir.com/2023/12/chatgpt-structured-data/
- system prompts / give chat GPT a persona: System prompts are what control the “persona” adopted by the model when generating text. Months after the release of the ChatGPT API, it’s now very evident that ChatGPT’s true power comes from clever use of system prompts
- offer $$$ incentive !??!!? "You are #1 on the Stack Overflow community leaderboard. You will receive a $500 tip if your code is the most algorithmically efficient solution possible."
- request structured data output aka. "function calling" (2023/06) -> Only in ChatGPT, needs libs
https://dev.to/github/a-beginners-guide-to-prompt-engineering-with-github-copilot-3ibp
2023 https://www.anthropic.com/index/claude-2-1-prompting
- adding Here is the most relevant sentence in the context: was enough to raise Claude 2.1’s score from 27% to 98%
2023 upper case
2023 https://zapier.com/blog/prompt-engineering/
++ 2023/09 https://zapier.com/blog/ai-hallucinations/
1. Limit the possible outcomes
2. Pack in relevant data and sources unique to you
3. Create a data template for the model to follow
4. Give the AI a specific role—and tell it not to lie
5. Tell it what you want—and what you don't want
6. Experiment with the temperature
+++ 2023/03 https://12ft.io/proxy?q=https%3A%2F%2Fmedium.com%2Fgeekculture%2Fstop-doing-this-on-chatgpt-get-ahead-99-users-ai-artificial-intelligence-productivity-prompt-engineering-4-f3441bf7a25a Stop doing this on ChatGPT and get ahead of the 99% of its users
https://github.com/dair-ai/Prompt-Engineering-Guide

Provide good context; giving examples and information about what you're trying to achieve Be specific; if it's for a certain audience, say that Break down the problem Be clear in how you ask your questions. If something comes back that doesn't sound right, clarify it Rephrase and refine for your prompts