92 92. TOPICS (ignore this section - it is just for me to develop the book)

92.1 Sessions 1,2

Survey of current AI capabilities and shortcomings.
Popular AI tools and models.
AI tools for data analytics.
Running AI models locally vs remotely - privacy and security concerns.
Legal and ethical considerations.
Expert’s projections on trajectory of AI capabilities in the near and medium‑term future.

training date cutoff
trained on “Internet”
objective - predict next token - if wrong modify weights slightly to make the correct token slightly more likely to be chosen next time
- repeat this training over entire “Internet size” data many many timess
probability distribution of next most likely token
tokenization
temperature -
model sizes - parameters
LLM vs other modalities - eg pictures, audio, video, special purpose (e.g. recognizing fraud),
multi-modal models

Survey of Web development technologies - HTML, CSS, JavaScript, NodeJS, web development frameworks (eg. React).
Deeper dive into browser-based Developer Tools (e.g. the debugger)
If there is time - survey of advanced web development concepts - SCSS, bundlers (e.g.. webpack), etc
Homework: Update the initial project to include additional features. Respond to feedback.
Session 6
- AI agents (as related to coding)

Session 7, 8
- Overview of relational databases and SQL
- NOSql - Document database (e.g. mongo db)
- Integrating a database with your software
- Homework - expand on features of previous submission, track everything in git and github. Include database features if that hasn’t already been done.
MCP Servers
Sessions 9,10:
- Survey of Internet architecture and infrastructure technologies - e.g. servers (web, database), “containerization”, virtual servers, DNS
- Deploying a web product
- If there is time: CI/CD using github
Sessions 11,12:
- Survey of programming languages and related technologies. How to determine a good fit for your project? Survey of tech product categories (web, mobile, data analytics, embedded systems, high performance e.g. algorithmic trading, etc)
- Survey of Python language for non-Python coders.
Sessions 13,14:
- UI development tools – e.g. figma, sketch, framer, webflow, canva
- charting languages, eg. mermaid
Sessions 15,16:
- Creating and using automated test cases

Sessions 17,18:
- Integrating with 3rd party services via APIs.
- Adding credit card processing to a web service (e.g. via Stripe)
- If there is time: Creating your own APIs (possibly to be included in a follow on course)
Sessions 19,20:
- Developing mobile apps
- Using phone emulators for app development (e.g. Android Studio, Xcode)
Sessions 21,22,23:
- Overview of no-code/low-code workflow automation tools – e.g. make, n8n, zapier, gumloop (new AI first approach), connect common tools without writing full software applications.
- Rapid development when compared with custom coded apps but can be more expensive to run. Great for prototyping or getting out version 1 of complicated apps. Can also be used in production but at much higher (sometimes prohibitive) cost.

Homework: develop

Session 24:
- High level survey of architecture of deep learning models. Using this knowledge to inform best practices for effective use of AIs (e.g. why some AIs cannot count the number of r’s in “strawberry”). vector embeddings, tokens, the perceptron, deep neural networks, training data, objective function, backpropagation, sizes of models (measured in parameters) and compute requirements, pretraining, post training, reinforcement learning, RLHF, reasoning models, multi-modal models (“modality conversion models” vs “native multi modal”)

Sessions 25,26,27:
- How to customize an AI with your own data:
- Different approaches to customizing an AI – discuss pros/cons of each approach
  - Add data to context (simplest approach), discuss context sizes of models
  - Retrieval Augmented Generation (RAG), vector databases
  - Survey of “Fine Tuning” concepts
(If there is time or perhaps for a 2^nd semester course) -
- Using advanced AI coding tools. The following are some current examples, but these will most likely need to be updated later. We will obviously not cover all of these. We will just introduce the types of tools and what they enable you to do). e.g. openRouter, langchain.com, langflow, flowwiseai , auto-gpt, agentGPT, llamaIndex, GradientJ, Haystack, orq.ai
- Developing AI agent software.