L10
Announcement
- Assignment 2 deadline: Mar 21, 11:59 PM
- Optional open lab next week
- Group project participation reminder; penalties for free riders
- Midterm grades available by this weekend.
Overview of AI's Impact
- Generative AI: A tool mimicking human intelligence, used in Google's search, ads, spam filters, etc.
- AI Winter: Past decline in AI interest due to unmet expectations.
- Generative AI creates outputs from text prompts, with examples like ChatGPT 3.5.
Generative AI Characteristics
- User input for prompts influences output.
- Issues: AI hallucination, biases, and inappropriate content.
- Major tech figures herald it as transformative (Gates, Pichai).
Generative AI Tools
- Released by OpenAI; competitors include Google (Bard), Meta (Llama), Amazon.
AI Workings Summary
- Foundation Models: Forms basis for Large Language Models (LLMs).
- Training requires extensive data and computing power.
- Parameters: Affects the complexity and depth of LLM responses.
Key AI Terms
- AI: Software mimicking human functions.
- LLM: For general language understanding.
- Corpus: Data used for training AI.
- Prompt: Input for generating AI responses.
- Hallucination: Incorrect confident AI responses.
Learning Approaches
- Various types: supervised, self-supervised, and reinforcement learning.
- Constitutional AI: Aligns AI behavior with specific rules.
The Transformer
- Technology allowing simultaneous analysis of text elements.
Future Work Implications
- GenAI will automate tasks, enhance cognitive capacity, and aid in learning.
- Generates summaries, assists in creative processes, and improves training.
Generative AI Applications
- Office Tools: Microsoft introduces 365 Copilot with AI enhancements.
- Programming: Aids in coding tasks efficiently.
- Graphic Design: Canva and Adobe leveraging GenAI for creative processes.
Risks of AI
- Issues: discrimination, security risks, bias, and misinformation.
- Importance of understanding and managing AI risks and potential societal impacts.
Addressing AI Risks
- Encouraged diverse hiring in tech, ethical policy development, and system audits.
- Recognition of biases and ethical implications in AI.
Conclusion
- Proactive steps needed to develop responsible AI and balance innovation with ethics.