The AI Hierarchy

AI → Machine Learning → Deep Learning → Generative AI (each is a subset of the previous)

• AI: Broad field — cognitive computing, computer vision, NLP, neural networks, ML, deep learning
• Machine Learning: Supervised (classification, regression), Unsupervised (clustering, dimensionality reduction), Reinforcement Learning (policy)
• Deep Learning: Neural networks — multiple layers, feature extraction (e.g. image → bird)
• Generative AI: Subset of DL using Transformers. Input → Encoding → Decoding → Output. Learns patterns from existing content to create new content.

---

Step-by-Step GenAI Learning Roadmap (20 steps)

1. Grasp AI basics — understand AI/ML/DL differences, types of AI, real-world applications
2. Learn Python — NumPy, Pandas, Matplotlib, basic data preprocessing
3. Master ML Concepts — supervised/unsupervised, classification, regression, clustering, overfitting/underfitting
4. Dive into Deep Learning — neural networks, forward & backpropagation, activation functions, TensorFlow/PyTorch
5. Understand Transformers — self-attention, encoder/decoder, positional encoding. Read “Attention Is All You Need”
6. Explore Language Modeling — tokenization, embeddings, masked vs casual models, next-token prediction
7. Get Started with GPT Models — GPT-2/3/4, OpenAI Playground, ChatGPT experimentation
8. Learn BERT & Encoder-Based Models — masked language modeling, classification, QA, HuggingFace
9. Dive into Generative Models — GANs, VAEs, Diffusion Models (image, audio, video)
10. Practice Prompt Engineering — zero/few-shot, chain-of-thought, prompt structure effects
11. Build with OpenAI & HuggingFace — ChatGPT API, DALL-E, Whisper, HuggingFace Spaces & Models
12. Work with LangChain — pipelines, agents, memory, tools, external data sources
13. Create Real-World GenAI Projects — content writers, chatbots, text-to-image, text-to-code
14. Learn RAG — how LLMs retrieve and generate, LlamaIndex, Haystack, vector databases (Pinecone)
15. Experiment with Fine-Tuning — LoRA, PEFT, domain-specific datasets
16. Explore Multi-Modal GenAI — GPT-4V, Gemini, image-to-text, text-to-image, design/vision use cases
17. Study Ethics & AI Safety — bias, fairness, safety practices, responsible AI deployment
18. Build AI Agents & Workflows — Auto-GPT, CrewAI, OpenAgents, task delegation, memory, planning
19. Join AI Communities — HuggingFace, Discord, Twitter/Reddit, follow researchers, contribute to OSS
20. Stay Updated — arXiv papers, new APIs/tools/models, continuously build and share

---

Generative AI Learning Roadmap (8 phases — ByteByteGo)

Phase	Focus
1. What is GenAI	AI/ML/DL/GenAI hierarchy; GenAI = creates new content from learned patterns
2. Important Concepts	Probability, Linear Algebra, Calculus, Statistics
3. Foundation Models	GPT, Llama, Gemini, DeepSeek, Claude
4. GenAI Dev Stack	Python, LangChain, ChatGPT, Prompt Engineering, VectorDB, DeepSeek, MetaAI Llama, HuggingFace
5. Training a Foundation Model	Dataset Collection → Tokenization → Training → Configure → Evaluation → Deployment
6. Building AI Agents	Memory, Reactivity, Autonomous Action, Human Control, Delegate Tasks, Tools (API, Internet, Code)
7. GenAI for Computer Vision	GAN, Midjourney, DALL-E, Flux
8. GenAI Learning Resources	DeepLearning.AI, Kaggle, Google Labs, Nvidia Learning, GenAI System Design Interview

---

Top Video Resources

Andrej Karpathy:

• Deep Dive into LLMs like ChatGPT — under-the-hood LLM fundamentals
• How I use LLMs — practical usage examples
• Intro to Large Language Models — foundational overview
• Let’s Build GPT From Scratch, Let’s Reproduce GPT-2, Let’s Build GPT Tokenizer

3Blue1Brown:

• Transformers: the tech behind LLMs
• Attention in Transformers, step-by-step
• How LLMs Store Facts

Stanford:

• CS229 (Andrew Ng) — ML foundations
• CS230 — Deep Learning
• Agentic AI Stanford Webinar

Other:

• Generative AI Full Course (Free Code Camp)
• Two Minute Papers (YouTube)
• ByteByteAI (YouTube)
• MIT’s Efficient DL Course

---

26 Best Resources to Learn AI in 2026 (ByteByteGo)

Books

• AI Engineering & Designing ML Systems
• ML Systems (Design Interview)
• ML & GenAI System Design Interview
• Hands-on LLMs

Tech Blogs

• OpenAI Research, DeepMind Blog, HuggingFace Blog, Anthropic Engineering Blog, Google AI Research, AI2 Blog

Courses & YouTube

• Stanford ML and DL Course (CS229/CS230)
• Two Minute Papers (YouTube)
• ByteByteAI (YouTube)
• MIT’s Efficient DL Course

Newsletters

• The Batch (DeepLearning.AI)
• ByteByteGo
• Rundown AI
• Ahead of AI (AI Deep Dives)

Influential Papers

• Attention Is All You Need
• Scaling Laws for Neural Language Models
• InstructGPT
• BERT
• DDPM

---

Pro Tips

• Upload video transcripts into NotebookLM — turn any video into an interactive course. Ask questions, challenge concepts, go deeper in real time.
• The gap between AI-fluent and AI-confused widens monthly. The window is time-sensitive — start from now.

---

See Also

• AI — 30-Day Mastery Mind Map — comprehensive tool/workflow/model overview
• AI — Open Source RAG Stack — tools for building RAG systems
• AI — Agent Frameworks (N8N vs LangGraph) — agentic workflow tools
• AI — Coding Assistants for Financial Domain Evaluation — applied AI evaluation framework
• LLM Wiki Pattern — using LLMs to maintain a compounding knowledge base