[Srijan Wednesday Webinars] Artificial Intelligence & the Future of Business
1 like454 views
Cellstrat is a leading Indian AI startup focused on developing machine learning innovations across various industry verticals. The company emphasizes the importance of AI solutions, deep learning algorithm design, and community engagement in AI-related events. With a commitment to AI research and content, Cellstrat positions itself at the forefront of the AI and deep learning revolution in India.
![[Srijan Wednesday Webinars] Artificial Intelligence & the Future of Business](https://image.slidesharecdn.com/disrupt4-171207140409/85/Srijan-Wednesday-Webinars-Artificial-Intelligence-the-Future-of-Business-1-320.jpg)
![[Srijan Wednesday Webinars] Artificial Intelligence & the Future of Business](https://image.slidesharecdn.com/disrupt4-171207140409/85/Srijan-Wednesday-Webinars-Artificial-Intelligence-the-Future-of-Business-9-320.jpg)
![[Srijan Wednesday Webinars] Artificial Intelligence & the Future of Business](https://image.slidesharecdn.com/disrupt4-171207140409/85/Srijan-Wednesday-Webinars-Artificial-Intelligence-the-Future-of-Business-13-320.jpg)
[Srijan Wednesday Webinars] Artificial Intelligence & the Future of Business
- 2. CellStrat is India’s leading Artificial Intelligence startup specializing in development and research in emerging areas of AI and Deep Learning. • Working on multiple world-class Machine Learning innovations for selected industry verticals. • Our research and content in AI and ML is unmatched in Indian context. • Thought leader in AI communities and among Deep-tech researchers. Intro to CellStrat
- 3. Focus areas • Artificial Intelligence Solutions – AI Applications in areas like ePublishing, marketing automation, text recognition, computer vision etc • Deep Learning algorithmic design – Probability models, Regression, Supervised/Un-supervised learning, Neural Networks • Solution development – Python / Google TensorFlow, Amazon AI and Alexa Skills Set • AI Content – India’s leading research and content program in AI space (www.cellstrat.com/research-blog) • Community Events – Disrupt 4.0 talk series on Business of AI, Basics and Advanced Machine Learning Algorithms
- 4. Business of AI “AI is the new electricity” – Andrew Ng, ex-Chief Data Scientist, Baidu
- 5. Industrial Revolutions Industry 4.0 2000 - Present Industry 1.0 1760 - 1870 Industry 3.0 1960 – 2000 Industry 2.0 1870 - 1960 Shift from hand based production to use of steam engines, electrical communications & chemical manufacturing etc. Technology Revolution came in with telephone and radio getting introduced improving communication modes Switch to Electronics & IT automated production, Automation Connected age blurring the lines between the physical, digital, and biological spheres, Personalisation
- 6. Data economy Data is the new oil Battle for ownership of data as well as deriving benefits from it. 180 zettabytes of data (180 followed by 21 zeros) by 2025, as per IDC Real-time flows of often unstructured data from social media, transportation and all kinds of sensors Data earlier used by firms like Facebook and Google for targeted advertising. Now it is powering n-number of Artificial Intelligence (AI) or “cognitive”services, some of which are revenue-generating. Data- network effect Use Data to attract more users These users generate more data This helps improve services This attracts even more users
- 7. AI: A branch of Computer Science that creates intelligent machines that work and react like humans. AI-based machines can use bigdata that businesses are collecting to identify patterns and insights more efficiently than humans can. E.g. Self Driving Cars, Strategic Game Systems like Go and Chess, understanding human speech etc.
- 8. Web & Mobile Banking (Industrial Revolution 3.0) Intelligent Robotic Assistant (IRA) of HDFC (Industrial Revolution 4.0)
- 10. Meet CONNIE – Hilton concierge robot using IBM Watson https://www.youtube.com/w atch?v=jC0I08qt5VU E.g. Hospitality, Robotic Process Automation (RPA), Transport, Security, Military, Banking, Household etc.
- 11. AI in gaming like Chess, Go, Bridge etc. Eg. IBM supercomputer Deep Blue beat Grand Master and World Chess Champion Garry Kasparov in 1997 Eg. Google’s DeepMind AlphaGo beat global GO Champion Ke Jie of China and Korean Champion Lee Sedol. GO : a game with near-infinite moves
- 12. Applications of AI / ML Image Processing • Image tagging / Image Recognition • OCR or Optical Character Recognition • Self-driving cars Text Analysis • Spam Filtering • Sentiment Analysis • Information Extraction Data Mining • Anomaly Detection • Association Rules • Grouping • Predictions Healthcare • Medical Diagnosis • Imaging Diagnosis • Oncology • Drug Trials Video Games • Reinforcement Learning Robotics • Industrial tasks • Human simulations
- 14. Basics of AI and ML “AI is the new electricity” – Andrew Ng, Chief Data Scientist, Baidu
- 15. Artificial Intelligence Intelligence in machines : simulate human intelligence Train machines to learn from data : Machine Learning Robotics, Computer Vision, Image recognition, Chatbots - Natural Language Processing (NLP), Text Analysis, Data Mining, Self-driving cars, AI in Retail, Gaming, Credit Risk, Fraud Detection, Hospitality, Call Centre Agent Match Healthiply SCAN, Uber self-driving cars, Amazon ECHO product (home control chatbot device), Amazon GO retail store, Baidu AI Medical Assistant, Haptik or niki.ai chatbot, Boxx.ai retail analytics, Hilton using Connie – concierge robot from IBM Watson
- 16. Machine Learning Traditional analytics relied on hard-coded rules. ML relies on learning patterns based on sample data. AI systems learn by extracting patterns from data. This capability is called Machine Learning. ML can learn from labelled data (supervised learning) or unlabelled data (unsupervised learning), though the latter is a more difficult problem to solve. Computers can take decisions that appear subjective – eg an ML algorithm called Logistic Regression can determine when to recommend caesarean delivery. Another algo Naïve Bayes can separate valid emails from spam.
- 17. AI and ML • A Venn diagram showing how deep learning is a kind of representation learning, which in turn is a kind of machine learning. Machine Learning is part of the AI landscape. • ML is used for many but not all approaches for AI. Image Credit : “Deep Learning” book by Ian Goodfellow
- 18. Traditional Analytics Problem at hand Production Review errors Code the Rules Evaluate Pass Fail
- 19. Machine Learning approach Problem at hand Production Review errors Train ML algorithm Evaluate Pass Fail Training Data (x1, y1), (x2, y2)…
- 20. How does a regular program work? Input Data Code (Processing Steps) + = Output
- 21. Machine Learning works a bit differently Input Data Output + = Learned Parameters Training Step Input Data Evaluation Code+ = Output Evaluation Step Training Code+ Learned Parameters +
- 22. Machine Learning and its uses Classify or categorize (A, B or C) Trend analysis (how much / many) Anomaly Detection (odd men out) How data is organized Decide on future action • Data science is the use of statistical methods to find patterns in data. • Statistical machine learning uses the same math as data science, but integrates it into algorithms that get better on their own. • Machine Learning is said to facilitate Artificial Intelligence as it makes machines learn patterns from data. In that sense Machine Learning is what connects AI with Data Science.
- 23. • Function mapping from a set of pixels to an object identity is very complicated. • Deep Learning solves this by breaking the desired mapping into a series of nested simple mappings, represented by layers of the model. • The hidden layers extract increasingly abstract features from the image. • Given the pixels, first layer identifies edges by comparing the brightness of neighbouring pixels. Given the first layer, the second hidden layer identifies contours and corners, by detecting collection of edges. Given the second layer, the third layer can detect parts of specific objects by detecting collections of contours and corners. Given the third layer, the entire object can be detected by checking collection of object parts. Feature extraction with Deep Learning Image Credit : “Deep Learning” book by Ian Goodfellow
- 24. A simple intelligence formula Given input x, predict y : y = f(x)
- 25. Why is machine learning hard ?
- 26. Real life data is messy View Point Difference
- 27. Real life data is messy Illumination
- 28. Real life data is messy Deformation
- 29. Real life data is messy Occlusion
- 30. Real life data is messy Interclass Variation
- 31. Data is ambiguous The movie has great location, wonderful songs, intelligence dialogues. Though I care less
- 32. Data is ambiguous Scientists study whales from space
- 33. Data is ambiguous Boy paralyzed after tumor fights back to gain black belt
- 35. Thank you Vivek Singhal Co-Founder and AI Data Scientist, CellStrat 9742800566 [email protected] Call: +91-9999658436 | t: @CellStrat | #disrupt4.0