Introduction to AI and Privacy

Goals of the Series

• Focus on harnessing AI beneficially (not fear-based)
• Emphasis on privacy and security
• Hands-on tech and practical applications
• Learning to defend against AI threats

Threats of AI

1. Hidden AI on Devices
   • Example: AI on Windows silently sending data to big tech servers
   • Mitigation: Use Linux
2. Communication with Cloud AI
   • Sending private data to external parties
   • Risk of data being used for surveillance or future machine learning
   • Solution: Run AI locally on your computer

Running AI Locally

• Possible to run AI without internet connection
• Demonstration using Linux and local AI
• Addressing concerns about NPUs (Neural Processing Units)

Conceptual Explanation of AI

• Transition from rule-based models to AI models
• AI uses machine learning to adapt and generate novel ideas
• AI models involve neural networks and pattern recognition
• Example: Large models like GPT-4 with billions of parameters
• Training AI involves machine learning and backpropagation
• Importance of validating AI training to ensure correct learning

Large Language Models (LLMs)

• Capable of deep conversation and generating new content
• Distinction between LLMs and small language models (SLMs)

Key AI Functions

1. Machine Learning (ML) Function
   • Learning phase: requires expensive computation
   • Result: pre-trained model
2. Inference Function
   • Use/query phase: can run on standard computers

AI Models and Hardware Requirements

• Smaller models don't need specialized hardware (NPU, GPU)
• Complex models require powerful computers
• Examples of specialized hardware:
  • Microsoft's co-pilot features
  • Apple's neural engine
  • Google's tensor chip
• Importance of having plans to mitigate privacy risks

Safe Use of AI

• AI can be safe if running locally or when communicating with cloud AI under certain conditions
• Privacy risks when sending data to cloud AIs
• Planning to transition to secure operating systems like Linux
• Recommendation of using powerful computers for larger AI models

Demystifying AI Hardware

• GPU and NPU: specialized chips for efficient computation
• The Role of CPUs, GPUs, and NPUs in AI processing

Practical Demonstration

• Example setup: Dell XPS 15 with specific configurations
• Running AI models locally using Llama (open-source model)

Importance of Privacy in AI

• Privacy-focused AI applications in the series
• Combining AI with privacy and security tools for education and demonstration

Conclusion and Next Steps

• Importance of adhering to privacy recommendations
• Encouragement to adopt privacy-first approaches & tools
• Overview of upcoming content in the series