AI Agents Fundamentals In 21 Minutes
Description
Improve your AI skills with the FREE Prompting QuickStart Guide I made in collaboration with Hubspot: https://clickhubspot.com/1gg9
Want to get ahead in your career using AI? Join the waitlist for my AI Agent Bootcamp: https://www.lonelyoctopus.com/ai-agent-bootcamp
🤝 Business Inquiries: https://tally.so/r/mRDV99
🖱️Links mentioned in video
A few notebooks to try out from crewAI & Autogen that are easy to follow and get started. All credit goes to these companies and Deep Learning AI. Please make a copy:
Resources I consulted in making this video:
LangGraph course: https://www.deeplearning.ai/short-courses/ai-agents-in-langgraph/
David Ondrej n8n tutorial: https://youtu.be/XVO3zsHdvio?si=AQcMnYn8kJOogqLr
Andrew Ng Snowflake agentic design patterns: https://youtu.be/KrRD7r7y7NY?si=tFtd6wJKB6idtfKb
Andrew Ng Sequoia agentic design patterns: https://youtu.be/sal78ACtGTc?si=2i8Wyy57n8m6TbBK
YC business advice: https://youtu.be/ASABxNenD_U?si=k19a310Tj3USuKNe
🐙 Lonely Octopus: https://www.lonelyoctopus.com/
Check it out if you're interested in learning AI & data skill, then applying them to real freelance projects!
🔗Affiliates
My SQL for data science interviews course (10 full interviews):
https://365datascience.com/learn-sql-for-data-science-interviews/
365 Data Science:
https://365datascience.pxf.io/WD0za3 (link for 57% discount for their complete data science training)
Check out StrataScratch for data science interview prep:
https://stratascratch.com/?via=tina
🎥 My filming setup
📷 camera: https://amzn.to/3LHbi7N
🎤 mic: https://amzn.to/3LqoFJb
🔭 tripod: https://amzn.to/3DkjGHe
💡 lights: https://amzn.to/3LmOhqk
📲Socials
instagram: https://www.instagram.com/hellotinah/
linkedin: https://www.linkedin.com/in/tinaw-h/
discord: https://discord.gg/5mMAtprshX
🎥Other videos you might be interested in
How I consistently study with a full time job:
https://www.youtube.com/watch?v=INymz5VwLmk
How I would learn to code (if I could start over):
https://www.youtube.com/watch?v=MHPGeQD8TvI&t=84s
🐈⬛🐈⬛About me
Hi, my name is Tina and I'm an ex-Meta data scientist turned internet person!
📧Contact
youtube: youtube comments are by far the best way to get a response from me!
linkedin: https://www.linkedin.com/in/tinaw-h/
email for business inquiries only: [email protected]
========================
Some links are affiliate links and I may receive a small portion of sales price at no cost to you. I really appreciate your support in helping improve this channel! :)
My Notes
00:00 — Intro
01:13 — AI agents definition
- 01:17 Defining AI agents is difficult as it's a new field.
- 01:37 What is NOT an AI agent: One-hot prompting (asking AI to do a complex task start to finish in one go).
- 02:10 Example: Asking ChatGPT to write a full essay on topic X in one prompt results in vague output.
- 02:17 An agentic workflow significantly improves results.
- 03:08 Truly autonomous AI agent: AI can completely independently figure out steps, tools, and the iterative revision process.
- 03:26 Currently, we are focused on agentic workflows (level 2), not fully autonomous agents (level 3).
03:45 — Agentic design patterns
- 03:46 According to Andrew Ng, there are four widely accepted agentic design patterns:
- 03:52 Reflection: Asking an AI to review and critique its own output for correctness, style, and efficiency.
- 04:53 Tool Use: Giving AI access to external tools to help break down tasks and execute specific parts.
- 06:00 Planning and Reasoning: AI figures out the necessary steps and tools to accomplish a given task.
- 06:14 Example: Generating an image based on a reference pose, describing the image via text-to-speech; requires using multiple models (image analysis, image generation, text-to-speech).
- 06:48 Multi-agent Systems: Prompting multiple LLMs to take on different specialized roles to work together on a task, similar to a human team.
- 07:21 Research suggests better final results compared to a single AI handling everything.
- 07:29 Mnemonic for Agentic Design Patterns: Red Turtles Paint Murals
- 07:44 Examples of tasks accomplished with agentic patterns:
- 08:36 Prompt engineering is a large component of AI agents and agentic workflows.
- 08:44 Learning prompt engineering is a high-ROI skill.
09:24 — Multi-agent design patterns
- 09:26 Focus on multi-agent patterns from CrewAI/DeepLearning.AI course.
- 09:47 Building blocks: Single AI agent.
- 11:30 More than one agent working together is called a crew.
- 11:50 Different multi-agent design patterns:
- 11:56 Sequential Pattern: Agents work in an assembly line fashion, passing output from one to the next.
- 12:07 Example: Document processing (extract text -> summarize -> extract actions -> save to database).
- 12:32 Hierarchical Pattern: A leader/manager agent supervises multiple sub-agents, delegates tasks, and compiles their results.
- 12:45 Example: Writing a business report (manager delegates research to sub-agents with specialized tools, manager compiles and may pass to a decision-making agent).
- 13:37 Hybrid System: Combines sequential and hierarchical structures; agents collaborate top-down and sequentially; requires continuous feedback loops.
- 13:46 Example: Autonomous vehicles (top-level plans route, sub-agents handle real-time sensors, collision avoidance, road conditions; requires constant communication between agents). Common in robotics, navigation, adaptive AI.
- 14:31 Parallel Agent Design Systems: Agents work independently on different parts of a task simultaneously to speed up processing.
- 14:43 Example: Large-scale data analysis (agents process chunks of data separately and merge results).
- 14:57 Asynchronous Multi-agent Systems: Agents execute tasks independently and at different times; handles uncertain conditions better than sequential/parallel.
- 15:09 Example: AI-powered cyber security threat detection (agents monitor network traffic, suspicious patterns, random sampling independently and flag anomalies). Helpful for real-time monitoring, self-healing systems.
- 11:56 Sequential Pattern: Agents work in an assembly line fashion, passing output from one to the next.
- 15:36 Linking different systems together creates a Flow.
- 15:50 Increasing complexity increases potential chaos; similar to large human organizations requiring structures.
- 16:41 Code notebooks for implementing these systems using CrewAI are linked in the description.
17:04 — Building a no code agent in n8n
- 17:06 Can create multi-agent systems using a no-code tool like n8n.
- 17:15 Example Demo: Building a Telegram-based AI assistant (Inky Bot) for task prioritization using Google Calendar.
- 17:24 Capabilities: Communicate, prioritize tasks, access Google Calendar, create calendar events.
- 18:05 n8n Workflow Breakdown:
- Telegram Trigger: Initiates the workflow.
- Switch: Handles text or voice input.
- Voice Input: Telegram gets file -> OpenAI transcribes -> sends text to AI agent.
- Text Input: Sends text directly to AI agent.
- AI Agent Node 18:31: Configured with TAMT (Task, Answer, Model, Tools).
- Task: Process user query about daily tasks.
- Answer: Provide a prioritized to-do list and schedule calendar events.
- Model: OpenAI GPT-4o mini (changeable).
- Tools: Get calendar events (read Google Calendar), Create calendar events (write to Google Calendar).
- Communicates back via Telegram.
- 19:26 Even a single agent workflow can produce cool results; potential increases with more agents/functionalities, all no-code.
19:46 — Ways to use AI agents
- 19:55 Key Takeaway (from Why Combinator): For every SAS (Software as a Service) company, there will be a corresponding AI agent company.
- 20:13 This provides significant guidance on what kind of AI agent businesses/products to build.
- 20:30 Idea: Take any existing SAS company (Adobe, Microsoft, Salesforce, Shopify, Canva, etc.) and think about how to create an AI agent version of it.
- 20:48 Predicts a "vertical AI unicorn" equivalent for every SAS unicorn.