Module-4.pptxdsgaergdvdfregxvzsdgeresdfxcdz

xcvdryrtzxacety5uh,bv bxs

1. I N T R O D U C T I O N T O A I A N D A P P L I C A T I O N S 1BAIA103/203 DR. ABHISHEK S. RAO Associate Professor, Dept. of IS&E Canara Engineering College, Mangaluru.

2. 8.1 AI and Ethical Concerns What’s in the syllabus: Ethical Use of AI → ensuring fairness, transparency, accountability in AI systems. Is AI Dangerous? Will Robots Take Over? → myths vs reality, AI as a tool vs fear of superintelligence. Ethics in AI → principles: right, fair, just outcomes (Microsoft Tay bot failure). AI and Bias → biased training data → biased results → unfair outcomes. Towards Ethical & Trustworthy AI → unbiased datasets, explainable AI, human oversight. Why is Ethical AI Important? → prevents harm, builds public trust, supports long-term adoption. Impact of AI on Jobs → automation replacing repetitive jobs but creating new opportunities (AI maintenance, data roles). How can you explain in class: Start with a real case: Microsoft Tay chatbot (2016) → became racist & abusive in <24 hours. Show contrast: AI in healthcare saving lives vs AI in recruitment showing bias. Ask students: “Would you trust an AI system to decide who gets a loan or a job?” → sparks discussion on fairness. Use simple analogy: AI is like a mirror → reflects the data it is trained on (good or bad). End with debate/discussion: “Will AI replace jobs or create better ones?”

3. Microsoft Tay (2016) – Case Study What it was: Tay (“Thinking About You”) was a Twitter-based chatbot launched by Microsoft in March 2016 . It was designed to mimic the language patterns of a teenage girl and learn from conversations with users. What happened: Within 16 hours of launch , Tay began tweeting highly offensive, racist, and inappropriate content. This happened because trolls deliberately flooded it with toxic language and Tay “learned” to mimic them without safeguards. Why it failed: Lack of content filtering – No strong moderation against offensive inputs. Reinforcement from bad data – It learned from malicious users. Unclear guardrails – No ethical boundaries or role constraints were set. Aftermath: Microsoft took Tay offline within a day and later issued an apology. The case became a classic lesson in AI ethics, safety, and bias control .

4. 8.2 AI as a Service ( AIaaS ) What’s in the syllabus: Factors Triggering Growth of AIaaS → cloud computing, big data, cheaper hardware, demand for AI without high setup cost. The Growth of AIaaS → rapid adoption by businesses (healthcare, finance, retail), flexibility of pay-as-you-use models. Challenges of AIaaS → data privacy, vendor lock-in, hidden costs, limited customization. Vendors of AIaaS → Amazon AWS AI, Microsoft Azure AI, Google Cloud AI, IBM Watson, Oracle AI. How can you explain in class: Start with an analogy: “Just like we use Netflix as a service instead of building our own cinema, companies use AIaaS instead of building AI from scratch.” Show examples: AWS Rekognition (face detection), Google AutoML (no-code ML), IBM Watson (healthcare). Quick activity: Ask students → “If you were starting a startup, would you build your AI or use AIaaS ? Why?” Highlight risks: dependency on vendor + security concerns (real-world challenge). Wrap with discussion: “ AIaaS makes AI accessible to everyone — but do we lose control?”

5. 8.4 Recent Trends in AI What’s in the syllabus: Collaborative Systems → AI systems that work with humans (co-bots, decision-support systems). Machines Assisting Humans → AI helping in surgery, disaster management, education, etc. Algorithmic Game Theory & Computational Social Choice → AI for fair decision-making, voting systems, resource allocation. Multi-Agent Reinforcement Learning (MARL) → multiple AI agents learning and interacting in shared environments (e.g., self-driving cars coordination, swarm robotics). Neuromorphic Computing → brain-inspired chips that process info like neurons, more efficient than traditional processors. How can you explain in class: Begin with a simple analogy: “Earlier, machines replaced humans. Now, machines are starting to collaborate with us.” Show a case: Collaborative robots in manufacturing or AI in education as a teaching assistant. Use a short demo/example: AlphaGo (game theory), drone swarms (multi-agent learning). Relate neuromorphic computing to the human brain: “Imagine chips that think like neurons — less power, more intelligence.” Quick class question: “Do you think future AI will work with humans or replace them?”

6. 9.1 Expert Systems What’s in the syllabus: Popular Examples → MYCIN (medical), DENDRAL (chemistry), XCON (computer configuration). Characteristics → rule-based, reasoning ability, explanation facility, domain-specific knowledge. Components → knowledge base, inference engine, user interface. Participants → domain experts, knowledge engineers, end-users, system developers. Capabilities → problem-solving, reasoning, explanation, prediction. Advantages → consistency, availability 24×7, handles complex problems. Limitations → lack of creativity, expensive to build, limited to defined knowledge. Applications → medicine, engineering, customer support, agriculture. Technology → rule-based systems, frames, semantic networks. Development → identify problem, knowledge acquisition, knowledge representation, testing & refinement. How can you explain in class: Begin with a simple definition: “An expert system is like putting a human expert’s brain into a computer.” Show a real-world case: MYCIN diagnosing infections, or modern-day medical AI assistants. Draw a simple diagram on the board: Knowledge Base + Inference Engine + User Interface. Connect to students: “Think of Google Maps suggesting a route — it’s like an expert system for navigation.” Quick discussion: “Can expert systems replace doctors or teachers? Or just assist them?”

7. MYCIN (Medical Diagnosis, 1970s) Domain : Medicine – bacterial infections & recommending antibiotics. Approach : Rule-based expert system (~450 rules). Function : Asked doctors a series of yes/no and multiple-choice questions about patient symptoms, test results, etc. → Suggested diagnosis and antibiotic treatments. Strengths : Showed how expert knowledge could be encoded into rules. Limitations : Never widely deployed (legal/ethical issues of trusting machine-generated diagnoses).

8. DENDRAL (Chemistry, 1960s) Domain : Organic chemistry – mass spectrometry analysis. Approach : Rule-based expert system with heuristics. Function : Interpreted mass spectrometry data to hypothesize possible molecular structures. Strengths : One of the first AI systems to outperform human experts in a narrow domain. Impact : Proved AI could solve complex, specialized problems. XCON (a.k.a. R1, 1980s) Domain: Computer system configuration (Digital Equipment Corporation, DEC). Approach: Rule-based system (~2,500 rules at its peak). Function: Automatically configure orders of Virtual Address eXtension (VAX) computer systems (which components fit together, wiring, power supply, etc.). Strengths: Saved DEC tens of millions of dollars per year by reducing human errors in configuration. Impact: Considered one of the most successful expert systems in industry.

9. 9.2 Internet of Things (IoT) What’s in the syllabus: Applications → smart homes, smart cities, healthcare monitoring, agriculture (smart irrigation), industrial automation. IoT Products → smartwatches, smart TVs, Alexa/Google Home, connected cars, fitness trackers. Challenges → data privacy & security, high cost of devices, connectivity issues, standardization problems. Sensors → temperature, motion, pressure, proximity, biometric, GPS core components that collect data. How can you explain in class: Start with a relatable example: “Your smartwatch tracking your steps and sending data to your phone is IoT in action.” Show images/video of IoT devices: Alexa controlling lights, farmers using soil sensors. Connect to students: “How many of you are already using IoT? Smart TVs? Fitness bands?” Quick discussion: “If IoT devices are always listening/monitoring, what are the risks?”

10. 9.3 Artificial Intelligence of Things ( AIoT ) What’s in the syllabus: How Does AIoT Work? → IoT devices collect data via sensors → AI analyzes this data → enables real-time decisions (predictive maintenance in factories). Where Does AI Unlock IoT? → adds intelligence to raw IoT data → anomaly detection, pattern recognition, automation. Applications & Examples → smart homes (voice assistants), healthcare (wearables detecting diseases), autonomous vehicles, smart factories. Benefits & Challenges → benefits: efficiency, automation, predictive insights; challenges: data privacy, infrastructure costs, integration complexity. Future of AIoT → edge AI for faster processing, 5G-enabled IoT, AI-powered smart cities, personalized healthcare. How can you explain in class: Begin with a simple analogy: “IoT is like the body’s senses, AI is like the brain. Together, they make smarter decisions.” Show example video: self-driving car (IoT sensors + AI decision-making). Connect locally: “Hospitals in India are experimenting with AIoT for patient monitoring, like AI-based ECG analysis through wearables.” Quick discussion: “If your home appliances learn your habits, is it convenience or invasion of privacy?”

11. T h a n k Y o u M o d u l e 4 - C o m p l e t e d