AI and ML for Everyone
1 like396 views
The document provides an introduction to artificial intelligence (AI), machine learning (ML), and deep learning (DL), covering historical milestones, key concepts, and the differences between these technologies. It explores supervised vs. unsupervised learning, popular algorithms, and various machine learning offerings from major companies like Amazon and Google. Additionally, it discusses industry applications of AI and the growing role of technology in customer interactions, e-commerce, healthcare, and ride-sharing services.
More Related Content
Similar to AI and ML for Everyone (20)
![big_data_topic1_[introduction]_[thanh_binh_nguyen].TextMark.pdf](https://cdn.slidesharecdn.com/ss_thumbnails/bigdatatopic1introductionthanhbinhnguyen-250717150423-ba7b6a06-thumbnail.jpg?width=560&fit=bounds)
AI and ML for Everyone
- 1. Artificial Intelligence and Machine Learning Introduction for Anyone and Everyone www.BigDataTrunk.com We will be starting soon
- 2. Agenda 1. Introductions 2. Machine Learning Concepts 3. ML vs DL vs AI 4. ML and AI Offerings 5. AI Use cases www.bigdatatrunk.com2
- 4. Introduction – Big Data Trunk 4 www.bigdatatrunk.com Raju Shreewastava • 21 yrs. in Data & Analytics • Founder of Big Data Trunk • Passion for teaching & Sharing • Presented in several conferences
- 7. Your turn • Working in • Software? • Data? • Big Data? • ML? • AI? 7 www.bigdatatrunk.com
- 8. www.bigdatatrunk.com When did Artificial Intelligence start?
- 10. www.bigdatatrunk.com History of AI • 1950 – Turing Test • 1951 – First Neural Network • 1967 – “Nearest Neighbor” Algorithm is written • 1974 – First AI winter • 1979 – Stanford Cart • 1996 – IBM Deep Blue beats Garry Kasparov Source https://www.bbc.com/timelines/zypd97h • 2006 - Geoffrey Hinton coins the term “deep learning” • 2014 – Facebook develops DeepFace & Google Buys DeepMind • 2015 – Amazon, Google & Microsoft ML offerings • 2016 – AlphaGo beats Lee Sedol • 2030 – Singularity?
- 16. www.bigdatatrunk.com Supervised vs Unsupervised Learning Concepts and Terminology
- 17. www.bigdatatrunk.com Supervised Learning (Train Data - labeled ) Unsupervised Learning (No Train Data – No labels ) KIDS ALIENS
- 18. Supervised Learning 18 www.bigdatatrunk.com Diameter, Thickness à Features Currency à Label Diameter Thickness Currency 1.0430 inches 0.0790 inches US dollar coin 1.0433 inches 0.07680 inches Canadian dollar coin (Loonie) 0.9154 inches 0.09173 inches One Euro coin • Supervised learning uses labeled data to train the model • Forecast an outcome
- 19. Unsupervised Learning 19 www.bigdatatrunk.com Hitters Pitchers Hits Number of Innings • Unsupervised learning is where there is no labeled data, model creates clusters/groupings • Discover underlying patterns and capture useful insights • Used in recommendation systems, anomaly detection
- 20. www.bigdatatrunk.com Main Machine Learning Techniques 20 Classification Clustering Regression
- 21. Classification – Supervised Learning 21 www.bigdatatrunk.com Weight = = heavy? High Mileage Horsepower < 80 High MileageLow Mileage yes yes no no • Classification is about predicting a label or a class • Classify emails as Spam or Not Spam
- 24. www.bigdatatrunk.com Unhappy Customer vs Happy Customer Customer Engagement Predictive Analytics
- 37. www.bigdatatrunk.com The Data Science Process (DIAPERS) Define Problem Ingest Data Analyze Data Prepare Data Evaluate Models Refine Model Ship It What data should I use? Is it labeled? Is data complete, clean, does it have coverage? Which algorithms should you use? What level of performance is acceptable? Deploy the Model and make predictions What are we trying to achieve? Is it labeled?
- 39. www.bigdatatrunk.com Machine Learning Offerings 39 Amazon Machine Learning Google TensorFlow Machine Learning Microsoft Azure Machine Learning Spark Machine Learning
- 44. www.bigdatatrunk.com AI vs ML vs DL AI ML DL 1950 1980 2006 Driverless car, Alexa Recommendation, Fraud detection, Image recognition Color B/W picture, add sound to video Uses ML,DL and repositories of data Works with all sizes of data but needs feature engineering Needs large datasets & compute capability and takes long time to learn Linear Regression, Decision Regression, K – means Clustering CNN, ANN (TensorFlow and Keras)
- 46. Artificial Intelligence - Types 46 www.bigdatatrunk.com Artificial Narrow Intelligence Artificial General Intelligence • AI that is good at one specified task which they are trained on • Examples – predicting home prices based on historical data, categorize email as SPAM • Lot of buzz about the progress in AI, but this is only in ANI (Artificial Narrow Intelligence) • Ultimate goal – make the computer smart or smarter than the humans • AI that can perform intelligent tasks as humans • Raises fears about job loses, “terminator” like scenarios • Still far from reaching the goal of Artificial General Intelligence (AGI)
- 49. www.bigdatatrunk.com AI Offerings 49 Amazon SageMaker Google AI Hub & TensorFlow Microsoft Cortana & Azure Bot IBM Watson
- 51. Customer Interaction 51 www.bigdatatrunk.com • Gartner group predicts that by 2020 over 80% of all customer interactions will be handled by Artificial Intelligence • Chatbot site examples – written and voice Source: https://apiumhub.com/tech-blog-barcelona/artificial-intelligence-ecommerce/
- 56. Ride Share – Uber and Lyft 56 www.bigdatatrunk.com • Machine Learning • Supply & Demand • Route Optimization • Rush Hour Pool • Uber Eats
- 62. Summary Supervised vs Unsupervised01 CCR & DIAPERS Model02 AI vs ML vs DL03 AI and ML Offerings04 Bright Future05