Lecture 1 AI

Instructor, context, and course framing

• Speaker: professor in an AI business course; assisted by TAs and a remote TA (Kaden) who will share email for help.
• Personal touch: shows a photo of himself on a magazine cover to relate to being a current student and having been in college once; emphasizes humility and relatability.
• Emphasis on practical, business-focused AI education: goal is to train a new kind of professional who can bridge business and AI (the “business translator”).
• Connection to broader AI discourse: mentions prominent figures (CEO of Google, a Stanford professor, a Toronto professor) and notable milestones (Hinton winning a Nobel Prize in 2024 for foundational neural networks work). Clarifies that neural networks are a component of deep learning; will cover backpropagation and related concepts.
• Context of AI hype vs. reality: acknowledges strong claims that AI could be as transformative as electricity or the wheel, and notes the need for grounding in math and practical application.

Key concepts introduced early

• Business translator: a functional, experienced business person (marketing, finance, operations) who understands enough about AI to contribute meaningfully to AI-driven initiatives.
• Data scientists vs. business context gap: data scientists often lack business domain understanding, and business folks often lack AI deployment context; the translator role is to bridge this gap.
• Industry demand signal: McKinsey chart (not shown here) indicates significant difficulty in filling AI-related roles; 55–60% of companies report trouble finding AI talent.

Why take this course? AI in the job market and strategic value

• AI is reshaping the job market for new grads, particularly in entry-level roles traditionally involved with spreadsheets, M&A analyses, and other routine tasks—much of which is increasingly handled by AI.
• Graduates face a skills gap: employers expect AI literacy and the ability to integrate AI into business processes.
• Note on undergraduate AI minors: only a handful of schools offer undergrad AI minors; having one can differentiate a candidate in the job market.

Capstone course and real-world project

• Capstone design: a 15-week project with a real company sponsor ( Vita Medical ) to solve a live business problem.
• Domain: FDA drug submission data management; vast data volumes and manual spreadsheet processes currently; goal is to apply AI to increase data throughput and automation.
• Deliverables and process:
  • Data gathering, cleaning, feature engineering.
  • Build models to improve throughput and automate data handling.
  • Weekly sponsor interactions; multiple presentations; culminating in a capstone synthesis of classroom learnings.
• Real-world analogy: similar experience to Deloitte, McKinsey projects or other consulting engagements from start to finish.

Real-world context and workforce considerations

• Gen Z and AI adoption: 72% of Gen Z workers fear AI will reduce entry-level jobs; course emphasizes how AI can create new opportunities rather than simply replace roles.
• Optimistic view: technologies like AI can create new roles and opportunities in fields that didn’t exist before; learning curves exist but can be overcome with new skills.
• Goals for students: enhance career prospects with AI literacy; stay ahead of business changes; improve decision-making with data insights; increase operational efficiency through automation.
• Brand new tools and ongoing learning: ethical and regulatory considerations will be addressed; emphasis on staying current as AI tech evolves rapidly.

Course structure and learning trajectory

• Week-by-week outline (highlights):
  • Week 1: Course objectives; intro to AI and business use cases.
  • Week 2: AI foundational skills (data basics, statistics, probability, linear algebra); heavy reading load; prerequisite alignment.
  • Week 6: Exam and core ML models for business applications.
  • Weeks 7–8: Linear regression and logistic regression; core math foundations; other ML/DL algorithms are acknowledged.
  • Weeks 9–11: Model selection and interpretability; deep learning, large language models (LLMs), and GenAI in business.
  • Week 12: Prompt engineering; AI use cases in marketing.
  • Week 13: Ethics and AI use cases in accounting.
  • Week 14: Autonomous agents and AI use in finance and retail.
  • Week 15: Student presentations (five minutes each).
• Each week covers deep topics; the course is designed to be a deep dive that could span a full semester in another setting.

Foundational content and model basics

• Early AI origins and milestones:
  • Alan Turing (1936) groundwork; Turing test (ability to indistinguishably fool humans into believing a machine is human).
  • Dartmouth Conference (1956) launched AI as a field.
  • 1960s: logic theory and early AI development.
  • 1970s: rise of expert systems and knowledge representation.
  • 1980s: hype cycles; data and compute constraints limited progress; later resurgence with new algorithms and data.
• Transformer architecture: key development enabling modern NLP and large models; unlocks effective natural language processing and scalable training.

Market evidence and industry signals

• Market sizing and growth (examples from McKinsey and others):
  • 2020 market size (global AI market) around $6.3 imes 10^9$
  • 2024 market around $6.4 imes 10^9$
  • 2034 forecast around $3.7 imes 10^{12}$
  • Compound annual growth rate (CAGR) around 19.2 ext{ ext{%}}
• Adoption and investment signals:
  • 78% of organizations report using AI in at least one business function.
  • 95% of organizations are investing in AI; 30% plan to invest more than $10{,}000{,}000$ next year.
  • 77% expect AI to increase responsibilities for early-career talent.
  • Data-center investments rising; AWS/Google signals and data spend growth foreseen.
  • US jobs impact: estimate that about 26 ext{ ext{%}} of jobs could be impacted by AI.
• Talent demand signals:
  • Microsoft study: 323% surge in demand for AI talent; 66% would not hire someone without AI skills; 71% would hire a less experienced candidate with AI skills; 77% expect greater responsibility for AI-enabled early-career staff.
• Practical demonstrations of AI value (case studies):
  • Harvard Business School + BCG study: groups with AI support performed better in speed and quality; similar to enhancements seen by analysts.
  • LLMs can outperform some financial analysts on certain tasks; AI-enabled tools provide relative advantages in workflow tasks.
• Autonomous agents: overview and differentiation from basic assistants (e.g., Jenny) – autonomous agents can perform tasks in external environments (e.g., booking travel, making reservations) without human intervention.

Practical use cases and demonstrations

• Finish Line case study (retail):
  • Problem: determine when to discount fashion items with expiration timelines (non-evergreen items).
  • Approach: multi-step AI pipeline using historical data and new-product analogs:
  • Clustering of products to find closest historical analogs for new items (no direct history for new SKUs).
  • Classification (e.g., Support Vector Machine) to assign new products to the closest class based on attributes.
  • Prediction of demand and markdown optimization using a Random Forest model.
  • Key features influencing decisions: price, demand drivers, expiration, attribute similarity, and yesterday’s activity.
  • Outcome: achieved the same sell-through rate with higher margins; margin improvement from ~41 ext{ ext{%}} to ~50 ext{ ext{%}}, representing a ~9 percentage-point uplift in gross margin (significant in retail).
  • Practical takeaway: small, well-structured AI applications can yield outsized business value without massive cost.
• Electronic Health Records (EHR) and clinical data projects:
  • Challenge: data quality and heterogeneity; transforming/normalizing data to enable AI use.
  • Approach: use AI to transform data representations and enable downstream tasks (e.g., fine-tuning a large language model for domain-specific tasks).
  • Learning objective: demonstrates the breadth of AI applications beyond marketing/retail to regulated, data-intensive industries.

Tools, platforms, and practical resources

• Gemini Pro: typically a subscription; free for 12 months for students; needed for assignment 1 (week 2).
• Copilot: integration with GPT-5; recommended to explore as a practical tool in coding and workflow.
• Reading materials: no single textbook; course uses a mix of articles, papers, and cases; some reading is heavy in Week 2; plan to stay ahead.

Ethics, governance, and policy considerations

• AI ethics and regulatory considerations will be touched on (not a full ethics course, but critical aspects discussed).
• Practical caution: AI results require validation; there are real risks in relying on AI outputs without verification.
• The “kill switch” concept and governance: even prominent AI leaders advocate for safety mechanisms and governance controls; awareness of potential existential or societal risks remains important.
• COBOL vs. COBOLT discussion reference: illustrates the tension between historical computing constraints and speculative futures of AI.
• The policy stance: university policy allows AI assistance in coursework but prohibits use during exams; students should verify AI outputs and not submit unverified results.

Student perspectives, questions, and classroom dynamics

• Common concerns:
  • Will AI replace jobs? The instructor frames this as a learning opportunity: AI changes job roles and creates new capabilities rather than simply eliminating jobs; the translator role helps bridge gaps.
  • How quickly will AI adoption outpace skills growth? Acknowledges learning curves but emphasizes proactive upskilling.
  • How to balance AI usage with human expertise? The teacher emphasizes the value of human judgment and domain knowledge alongside AI tools.
• Hands-on practice and presentations:
  • Students will present a topic in Week 15 (5-minute presentations) to practice communication and persuasive explanation of AI topics.
  • The class emphasizes practical projects and skills that translate to real-world business settings.

Ethical, philosophical, and practical implications highlighted

• AI as a tool, not a magic solution: math underpins AI capabilities; understanding math helps demystify AI claims.
• The risk-reward balance of AI in organizations: 5% AI project failure rate; most successful projects happen when human and organizational factors (change management, alignment with business goals) are properly addressed.
• Human-centric AI: even with powerful models, human oversight, alignment with values, and governance are essential.
• Energy and resource considerations: governance extends to responsible usage of computational resources and environmental impact considerations.

Quick glossary and key terms from the session

• Business translator: a cross-disciplinary professional who understands both business needs and AI capabilities.
• AI foundation skills: data basics, statistics, probability, linear algebra; prerequisites for understanding ML/DL.
• Linear regression / Logistic regression: core models used to teach intuition about ML; foundational for more complex models.
• Transformer: architecture enabling modern NLP and LLMs; key to scalable language understanding.
• LLMs / GenAI: large language models and generative AI applications in business.
• Prompt engineering: crafting prompts to elicit useful outputs from AI systems.
• Autonomous agents: AI agents capable of performing tasks and executing actions in external environments without direct human input.
• Selfie (retail context): a term used to describe the starting point of demand-based markdown decisions for new products lacking historical sales data.
• Ground truth vs. model prediction: evaluating model accuracy against observed outcomes to train and improve models.
• Ground truth features: key variables that influence model predictions (e.g., price, demand, expiration).
• Data center spend projections: signals about AI infrastructure expansion and data-processing capacity.

Summary takeaway

• This course blends theory, practical AI techniques, and industry cases to prepare students to become business translators who can lead AI-powered initiatives.
• The capstone project with Vita Medical exemplifies how AI can transform data-heavy business processes in regulated industries.
• Students will gain hands-on experience with foundational math, ML concepts, and practical tools while engaging with ethical and governance considerations.
• The broader market context shows AI adoption accelerating across functions and industries, creating new opportunities for skilled, translator-like professionals who can bridge business and technology.