Artificial_Intelligence_Updated_SH25.pptx

This presentation outlines topics around artificial intelligence

1. Artificial Intelligence

2. Agenda The Unseen Revolution The Ghost in the Machine From Knowledge to Numbers Finding the Hidden Patterns

3. A.I. in daily life

4. The U n seen Revolution

5. From Giant Calculators to Your Pocket There was a time when a 'computer' was a machine bigger than this room, just to solve basic math. Today, we knowingly—and unknowingly—use powerful AI every single hour. Examples of AI in our life" Unlocking your phone with your face ( obvious ) Your email filtering out spam ( invisible ) Getting a movie recommendation on Netflix ( obvious ) Your phone's autocorrect fixing a typo ( invisible ) Key Takeaway: AI has evolved from a massive, specialized tool into an invisible engine that powers our everyday world.

6. Evolution Of AI/ Ml

7. Brief History Of AI’s Evolution Key Milestones: Precursors (Ancient - 1940s): The dream of automated logic and calculation 1943: The First Spark: The first mathematical model of an artificial neuron 1956: The Birth of a Field: The term "Artificial Intelligence" is officially coined 1980s: Expert Systems Boom: AI goes commercial with rule-based systems that mimic human experts Late 80s-90s: The "AI Winter": Progress stalls, leading to reduced funding 1990s-2000s: The Rise of Machine Learning: Shift from programming rules to learning from data 2010s-Today: The Deep Learning Revolution: Massive data and powerful computers fuel today's AI breakthroughs Key Takeaway: AI's history is a story of shifting ideas: from trying to program rules of intelligence to creating systems that learn from experience

8. Modern Era Of AI/ Ml Key Milestones: 1980s - Expert Systems: Systems based on a large knowledge base of "if-then" rules. They couldn't learn beyond their programming. 1986 - Backpropagation: The breakthrough algorithm that allowed neural networks to learn from their errors efficiently. 1989 - CNNs for OCR One of the first practical, successful applications of a Convolutional Neural Network to read handwritten digits. 2017 - "Attention Is All You Need": The landmark paper introducing the Transformer architecture, which powers modern LLMs. 2018-2022 - The LLM Race Begins: The release of GPT-1 and later ChatGPT sparks the current boom in generative AI.

9. The Ghost in the Machine

10. But how do we define AI ? Artificial Intelligence is about mimicking traits of natural rather specifically human level Intelligence. But, what is Intelligence then…? What are Traits of Intelligence. Is a Sunflower and intelligent entity ? Or what about Sparrow or a Chimpanzee ?

11. Traits of Intelligence Decision Making Being able to take actions from based on some intuition. Interpreting Information Understanding data observed from the environment and getting actionable insights from it

12. Traits of Intelligence Adaptability and Generalisation: The ability to adjust actions and strategies in response to new or changing situations . Pattern recognition The ability to identify patterns or regularities in data or situations..

13. The Illusion of Intelligence It Seems Intelligent: Can track you, dodge, heal, and change attack patterns But It's Just a Complex Rule-Based System: Healing: IF Boss_Health < 20% THEN Execute_Heal_Ability () Dodging: IF Player_Starts_Heavy_Attack THEN Execute_Dodge_Move () Phase 2: IF Boss_Health < 50% THEN Switch_To_Attack_Pattern_B ()

14. The Illusion of Intelligence Siri and earlier Assistants The 3-Step Process Pattern Recognition Listen for pre-programmed keywords and phrases Intent & Entity Extraction Identify the user's goal ( Intent ) and key detail ( Entity ) Execute a Pre-Programmed Action Trigger a hard-coded function or pull from a database of responses Example: Prompt:"Set an alarm for 7 AM“: Intent: SetAlarm Entity: Time = 7:00 AM Action: Call the phone's Clock app with the correct command

15. Why the Old Way Was So Limited The Brittleness of Rules: The Missing Ingredients of Intelligence: 1.   No Real Adaptability: The system was rigid. If your command didn't precisely match a pattern, it failed. It couldn't handle typos or new ways of asking. 2.   Very Low-Quality Generalization: Knowledge was siloed. Knowing how to "set an alarm" gave it zero insight into how to "start a stopwatch." 3.   Zero Learning from Experience: It made the same mistakes repeatedly. The only way it "learned" was when a human manually added more rules. Key Takeaway: Early chatbots failed because they lacked the core ingredients of intelligence: they couldn't adapt to novelty or generalize their knowledge.

16. Finally , how do we define AI ? AI is not about building a system with a bigger, better rulebook. It's about building a system that can throw the rulebook away and learn from experience. It’s the shift from... A Cookbook to a Chef: We're moving beyond systems that just follow a recipe (rules) and creating systems that can actually taste the ingredients (data) to invent a new dish (solution). Handling the Expected to Mastering the Unexpected: A calculator can only solve the problems it was built for. An AI learns to generalize, making intelligent guesses about new problems it has never seen before. In one sentence: It’s the difference between a tool that follows a map and one that can draw its own .

17. Finally , how do we define Intelligence ? 1.   Learning from Experience Improving performance based on past outcomes. 2.   Reasoning / Problem-Solving: Using logic to make decisions or solve new problems. 3.   Adaptability: Adjusting behavior to handle new or changing situations. 4.   Perception / Understanding: Making sense of sensory data (images, sound, text). 5.   Generalization: Applying learned knowledge to completely new, unseen situations. Examples: *   Human Analogy:** Studying for an exam—you learn from mock tests and generalize your knowledge to solve new questions on the final. *   AI in Action: Self-driving cars perceive the road, learn from millions of miles, adapt to pedestrians and generalize to drive on a new street.

18. From Knowledge to Numbers

19. How does AI understand Data ?

20. Final tips & takeaways Consistent rehearsal Strengthen your familiarity Refine delivery style Pacing, tone, and emphasis Timing and transitions Aim for seamless, professional delivery Practice audience Enlist colleagues to listen & provide feedback Seek feedback Reflect on performance Explore new techniques Set personal goals Iterate and adapt

21. Speaking engagement metrics Impact factor Measurement Target Achieved Audience interaction Percentage (%) 85 88 Knowledge retention Percentage (%) 75 80 Post-presentation surveys Average rating 4.2 4.5 Referral rate Percentage (%) 10 12 Collaboration opportunities # of opportunities 8 10

22. Thank you Brita Tamm 502-555-0152 [email protected] www.firstupconsultants.com

23. Data Representation in AI AI only works with *numbers* Text, images, audio → converted into arrays/matrices Example: Image → pixel grid Text → word encodings Categories → integers

24. Feature Engineering Cleaning & transforming raw data → usable form Examples: Sentiment (Positive=1, Negative=0) Age group → categories (Teen=0, Adult=1, Senior=2) Normalization → scaling between 0–1

25. Encoding Categorical Data Label Encoding: Cat=0, Dog=1, Fish=2 One-Hot Encoding: Cat=[1,0,0], Dog=[0,1,0] Prevents AI from assuming categories have order

26. From Data to Models Once data is numeric → AI applies math models ML → Finds patterns in data DL → Extracts meaning through many layers Always math, never magic

27. How ML Processes Information ML models learn rules from data Linear regression = best-fit line Decision trees = splitting conditions Learns relationships, not memorization

28. How Deep Learning Understands Neural networks: layers of math functions Input layer → raw features Hidden layers → extract patterns Output → prediction/classification Example: cat image → edges → shapes → 'cat'