DAY 4 part two of dat worlds intro

Page 1: Class Updates

• Complete Class Announcement: - The class is now complete, and vehicle distribution will occur over the weekend for those signing up for a note-taking assignment.- First-come, first-served basis for signing up. - Instructor commitment to accommodate student constraints.

• Walk-In Office Hours: - Available for students but requires advance notice to ensure no scheduling conflicts. - Suggested communication: Email or text if it's a last-minute decision.

Page 2: Current Events in Technology

• DeepSeek Impact on Big Tech: - Emergence of DeepSeek is seen as a turning point for major AI developers.- Initial claims show potential cost reductions and changes in model behavior.

• Assumptions Challenged: - Big Tech companies believed their investments and high-performance resources created a "moat" around their market dominance.- DeepSeek's success demands a re-evaluation of these assumptions.

Page 3: Ethical Considerations

• Schadenfreude in AI Community: - Recent allegations against Chinese use of models raise ethical concerns.- Criticism aimed at OpenAI for perceived hypocrisy regarding integrity in AI use.

Page 4: Investor Concerns

• Investors’ Interest in GenAI: - Questions about when generative AI will start yielding financial returns.

Page 5: Data Theory Insights

• Datafication Defined: - Datafication involves the practices and relations through which data is produced and understood.- The act of rendering data legible and meaningful is highlighted as crucial.

Page 6: AI Productivity Debate

• Discussion on AI Summaries: - AI-generated summaries, such as those from ChatGPT, often criticized for disorganization and lack of detail. - User reflections on finding AI less effective for complex content.

Page 7: Diverse Definitions of Data

• Complexity of Data: - Data encompasses various forms: numerical, textual, behavioral, etc.

  • Sarah Ciston's View: Data as values assigned to entities.

  • Jathan Sadowski's View: Data as a technology-dependent abstraction of the real world.- Recognition of diversity in data use and implications for research and application.

Page 8: Quantitative Evaluation

• Objectives: - Quantitative evaluation focuses on measurable, numerical aspects.- Common metrics: Accuracy, precision, recall, and more.

• Reproducibility: - Results are repeatable due to standardized measures.

• Statistical Analysis: - Importance of statistical rigor in performance evaluation.

• Scalability: - Applicability of methods to large-scale scenarios.

Page 9: Qualitative Evaluation

• Emphasis: - Focuses on subjective characteristics and overall model behavior. - Insights gained through human interpretation and context awareness.

• Human Feedback: - Incorporating user feedback to improve system understanding.

Page 10: Data Positivism Critique

• Critique of Data Salvaging Process: - Data positivism suggests qualitative analysis is overshadowed by sheer data quantity.

• Chris Anderson's Perspective: Claims about the end of traditional theory in favor of data-driven conclusions, which could neglect causal inquiries.

Page 11: Correlation vs. Causation

• Issues with Data Analysis Claims: - Emphasizing correlation is said to overlook the need for understanding the causes behind data trends.- Caution against relying on datasets, which may harbor biases and errors.

Page 12: Data Positivism Overview

• Greg Brockman's Thoughts: - Next-step prediction enhances understanding of underlying data processes.

Page 13: The Shift in Data Theory

• Data vs Datafication: - Differentiating between data as a product and datafication as a process shaping that product.

Page 14: Surveillance and Historical Recordkeeping

• Google's Perspective: - Reflecting on how historical societies weren't 'datafied' per contemporary standards, yet that doesn’t mean they lacked value assignments.

Page 15: Uneven Datafication

• Global Disparities: - Noticing datafication practices reveal inequalities in data access and representation.- Many global populations lack adequate resources for data inclusion.

Page 16: Shift in AI Paradigm

• Deep Learning Revolution: - Reflected growth of deep learning since 2012, including the influence of projects like ImageNet in shaping new AI capabilities.

Page 17: Power Dynamics in Data Usage

• Discussion Prompt: Exploring concentration of power in tech and its implications on data.

Page 21: Tensions in Generative AI and Education

• Education vs. AI Goals: - Generative AI challenges the traditional goals of education in developing articulate, informed students.- Potential economic motivations driving the acceptance of AI in educational settings.