AI - Andrew Ng, Generative AI for Everyone
READING APPLICATIONS FO AI
! Proofreading 2. Summarizing 3. Email routing 4. Restaurant reviews 5. Sentiment reviews
TIPS FOR USING FOR PROMPTING
1. Be detailed and specific. Give sufficient context for LLM to complete the task. 2. Guide the model to think through its answer. 3. If you don't like the response, ask AI to clarify or rewrite it.
WHAT ARE SOME LIMITATIONS OF LLMs?
1. Knowledge of LLM is frozen at the time of its training. 2. Hallucinations. Make thing up. Especially Quotes. 3. The input length of the Prompt or output of the LLM. 4. They do not work well with structured data. 5. Text on the internet can reflect biases that exist in society.
AI IS A SET OF TOOLS
1. Supervised learning (labeling things) 2. Generative AI 3. Unsupervised learning 4. Reinforcement Learning
1. Generative AI or GenAI SYSTEMS
A category of artificial intelligence (AI) algorithms that generate new outputs based on the data they have been trained on.
BING CHAT
An Open AI developed by Microsoft
BARD
An open AI developed by Google
MULTIMODAL MODELS
Can generate either text or images.
WEB-BASED INTERFACE APPLICATIONS
ChatGPT, Bard, or Bing
SOFTWARE-BASED APPLICATIONS
Email routing, document search
DIFFUSION MODELS
Four steps of supervise learning
HALLUCINATION
LLM can sometimes make facts up. Double check.
LLMs
Large Language Model - Natural language processing computer programs that use artificial neural networks to create high-quality text and images.
START DATE AND RISE OF GENERATIVE AI
Started in November 2022 when Open AI released ChatGBT which can follow instructions.
WHAT CAN AN LLM DO?
What a fresh college grad can do by following only the instructions in the prompts. Works best with text, images, and data.
COMMON USES FOR GENERATIVE AI
Writing, reading, and chatting tasks for both web-UI and software-based implementations of LLMs.