The TechnologyArtificial IntelligenceMachine LearningDeep LearningTransformersLLMs and other AI ModelsGenerative AIAgentic AIThe Next FrontierImpacts on/of DevelopmentHow the world impacts AI developmentHow AI development impacts the worldState of the ArtA Summary of Whats going on Now (Timeline)
The Technology
Artificial Intelligence
- What is AI?
- Brief history of AI
- 3 levels of AI capability: ANI v AGI v ASI
- Everyday examples of AI
- The AI family tree
- Fields of AI: NLP, Computer Vision, Robotics, Etc.
Machine Learning
- What is ML?
- How ML works (basics of data collection, training, inference)
- Overfitting vs underfitting
- ML vs traditional programming
- ML vs symbolic reasoning
- The 3 types of ML (+2 more)
- Supervised learning
- Unsupervised learning
- Reinforcement learning
- (+) Self-supervised learning
- (+) Semi-supervised learning
- Statistical (classical) ML vs deep learning
- Why ML is taking off now
- Compute power (GPUs)
- Data availability (Internet)
- Algorithm advancements (Transformers, LLMs)
Deep Learning
- What is deep learning?
- Neurons
- Neural networks
- What makes a network “deep”?
- How deep learning works (step-by-step breakdown)
- Random initialisation
- Forward pass
- Loss/error function
- Backpropagation
- Gradient descent
- Repeat over epochs
- Why does more hidden layers = better learning?
- Parameters and hyperparameters
- Types of neural network architectures
Transformers
- What are transformers?
- The attention mechanism (what it is and how it works roughly)
- Why transformers and attention changed everything
- Long-range context
- Parallelizable training
LLMs and other AI Models
- What are LLMs?
- How LLMs work (next-token prediction)
- Key stages of LLM training (look at ChatGPT’s training per the MIT course)
- Pre-training
- Fine-tuning
- RLHF
- AI models more generally (besides LLMs)
- LLMs = autoregressive
- Diffusion models
- GANs
Generative AI
- What is generative AI?
- Key gen AI terms worth knowing
- Types of models used in GenAI and how they’re used
- Autoregressive models (i.e. LLMs)
- Diffusion models
- GANs
- Key gen AI applications
- Text
- Images
- Audio
- VIdeo
Agentic AI
- What is agentic AI?
- Workflow automation vs agentic AI
- 3 core components of every AI agent
- Key examples of agentic AI today
The Next Frontier
- Where is AI technology headed currently? (e.g., robotics, research, world models, intelligence wars such as Fable vs Astra, GPU and microprocessor advancements)
Impacts on/of Development
How the world impacts AI development
- Availability of raw materials and hardware infrastructure (silicon, semiconductors, chip wars, GPUs, data centres)
- Algorithms and software systems (what we currently can and cannot do using AI based on what current software technology allows)
- Human capital; e.g., talent (top scientists and researchers), acquihires (e.g. Google and Windsurf, OpenAI and Openclaw, Meta and ScaleAI)
- Corporate competition and the model development race vs safety (AI safety, outpacing the frontier)
- Laws and regulations (techno-optimist vs regulation-heavy approaches)
- Geopolitical tensions (e.g. China and US race for advanced models, EU regulation-heavy approach, other countries trying to make their mark and getting left behind)
How AI development impacts the world
- Boons (by industry)
- Academia acceleration
- Business productivity
- Job creation for new roles
- Education enhancement
- Banes (by theme)
- Ethical issues (e.g., the rise of deepfakes, biased data, hacking by AI)
- Legal issues (e.g. copyright infringement when training LLMs)
- Environmental issues (i.e., the energy, water and carbon footprint of data centres)
- Economic issues (e.g., job loss to AI)
- Social issues (e.g., impact on human ability to do things ourselves and learn, as well as model collapse and the death of art)
- Geopolitical issues (e.g., China vs US in frontier model development leading to escalating tensions)
State of the Art
A Summary of Whats going on Now (Timeline)
- [timeline from ChatGPT release to key issues and events today]
- (think of brief history section as past to present; this section is present to future)
- News/newsletter feed constantly updated with what’s new