IoT, AI, ML Mix or How to Deal with New Technologies (Borys Pratsiuk Technology Stream)
3 likes826 views
The document provides an extensive overview of IoT, AI, and machine learning technologies, discussing various applications and case studies in smart homes, industrial automation, and agricultural fields. It covers the evolution of AI, types of IoT devices, and offers examples of chatbot development, highlighting platforms for constructing automated conversational interfaces. Furthermore, it presents challenges and future trends in integrating AI into daily technologies, including voice recognition and natural language processing.
IoT, AI, ML Mix or How to Deal with New Technologies (Borys Pratsiuk Technology Stream)
- 1. 1 IoT, AI, ML Mix or How to Deal with New Technologies Borys Pratsiuk, Ph.D Head of R&D Engineering
- 2. 2 C developer 2004 2005 С/С++ developer 2006 2009 - 2011 Asm, С, Android 2007 - 2013 defend Ph.D. assistant Profesor in KPI 2012 - now Join Ciklum • Senior Android • Team Lead • Android Architect • Head of R&D Who I am
- 3. 3 4 3 Chat bots overview Agenda What is it IoT? Existing types. Examples Science converted into the product. How algorithms began matter for business? What happening in AI today? 2 1
- 4. 4 Types of IoT Smart Homes Wearable devices Connected devices Industrial automation Agricultural Smart Cities
- 5. 5 Top 5 Industries 1 2 3 4 5 *Gartner research 2016 – Industry grow
- 10. 10 Case study: Digital transformation in IoT PoC results • Wireless channel of data transmission through Wifi connection from device to mobile and tablet • Video streaming from night vision to mobile • Remote device management WiFi Challenge • Transform optical device into smart IoT solution • Share video and photos • Create social network
- 11. 11 Industrial
- 12. 12 Agricultural
- 13. 13 Smart Cities
- 14. 14 IoT world 2016 hackathon The Solution – Empowering Responders Smart AED Predix Cloud App Spark Channel Responder’s Android App Predix Time Series • Public Safety Images • Traffic Speed Reverse Geocode
- 15. 15 Ciklum team solution Technology stack: • Java VM (Predix) for Intel Edison • Node.js for embedded • Ruby on Rails • Android Java • Native C • API’s integration
- 16. 16 Cloud providers for IoT
- 17. 17 Communication protocols Frequency: 2.4GHz (ISM) Range: 50-150m (Smart/BLE) Data Rates: 1Mbps (Smart/BLE) Frequencies: 2.4GHz / 5GHz Range: ~ 50m Data Rates: 150-200Mbps (latest 802.11-ac standard should offer 500Mbps to 1Gbps) Frequency: 2.4GHz (ISM) Standard: Thread, based on IEEE802.15.4 and 6LowPAN Frequency: 900MHz (ISM) Range: 30m Data Rates: 9.6/40/100kbit/s Frequency: 2.4GHz Range: 10-100m Data Rates: 250kbps
- 18. 18 AI and human brain
- 19. 19 History of AI McCulloch & Pitts Rosenblatt Ivakhnenko & Lapa Group Method of Data Handling (GMDH) Perceptron A Logical Calculus of the Ideas Immanent in Nervous Activity 1943 1957 1965 …
- 20. 20 Why Deep Neural Network now?
- 21. 21 THE BIG BANG
- 22. 22 NVIDIA is a leader in AI hardware
- 23. 23 What is Neural Networks? An artificial neural network is composed of many artificial neurons that are linked together according to a specific network architecture. The objective of the neural network is to transform the inputs into meaningful outputs. Tasks: • recognizing a visual object; • anomaly detection; • event prediction; • voice recognition; • deciding a category of potential object; • natural language processing. http://videolectures.net/deeplearning2015_vincent_machine_learning/
- 25. 25 ANN Example – not good http://playground.tensorflow.org/
- 26. 26 ANN Example – OK
- 27. 27 Things we want to do with data Images Audio Text Image labeling Speech recognition Web Search and Natural Language Processing
- 28. 28 Network layers for Image recognition
- 29. 29 Case Study: Pedestrian Tracking The smart IP camera with pedestrian tracking technology could be used within shopping malls for traffic counting, crowd monitoring and business intelligence with the back-end servers being freed up to perform data mining and data collection. The technology could sooner or later become integrated into consumer video monitoring solutions.
- 30. 30 Case Study: Face Recognition • Deep convolutional neural networks applied for the face recognition purposes could be used for door smart locks and security systems. • Face recognition technique combining with e.g. fingerprint scanner could increase security level and permissions given.
- 31. 31
- 32. 32 What is it ? 1. Machine understands what you speak 2. What you don’t speak 3. Other sounds too What is it not . . Does not deal with ultrasonic wavelength Only human audible sounds are under study now Problem? Understand what language is in record Speech processing and recognition
- 33. 33 • Cloud Speech API + AI = fast and furious Google Now • Wolfram Alpha + AI = making jokes Apple Siri • AWS + AI = customizable Amazon Alexa • Bing Speech API + LUIS = stupid Microsoft Cortana • SoundHound + AI = Hound • api.ai Engine + AI = api.ai Assistant (Google!!!) • wit.ai Engine + AI = Facebook wit.ai M • other What if to add Artificial Intelligence?
- 34. 34 • Messaging-as-OS: Messaging can be a platform • The app problem: People are reluctant to install apps • The “conversational interface”: A new model for interacting with online services Major chatbot trends
- 35. 35 How to make your own chatbot? If you are not a programmer Select the Engine Provide scenarios and make bot train Talk with your bot
- 36. 36 Chatbot Engines Engine Platforms Pricing https://api.ai/ Facebook messenger, Slack 0-899$/month https://www.itsabot.org/ Slack, Twitter, email, SMS free https://chatfuel.com/ Facebook messenger, telegram free https://smooch.io/ Facebook messenger, Telegram, Line, WeChat, Shopify, Twillo, Email, etc 0-100$ https://meya.ai/ Twitter, Facebook Messenger, Telegram, Slack and Kik 0-200$ https://wit.ai/ Facebook messenger free
- 37. 37 Conversational commerce Bots monetization: ● Bots + sponsored & native content. After subscription the client will receive a context advertisements similar with bot’s tematic ● Bots as a Services. B2B Bots that help people and teams be more productive, manage tasks or tackle communications challenges will replicate business models being used by existing B2B software ● Retail sales bots; ● Payment simplification bots
- 38. 38 AI betchat
- 39. 39 Case Study: bot and voice controlled lamp Amazon Echo is a hands-free speaker you control with your voice. Echo connects to the Alexa Voice Service to play music, provide information, news, sports scores, weather, and more—instantly. Its ready-to-use technology but still quite raw all you need to do is to write a custom skill to do any action you want, e.g. you are saying: “Alexa, ask lamp switcher to set lamp color to green!”
- 40. 40 Case Study: bot and voice controlled lamp wit.ai service allows us to develop bots for various applications. Users can operate with the different inputs of information (voice, text messages, gestures, etc.). After acquisition of command the information could be send directly to the bot (in case with text input) or transmit to the preprocessing service (as speech recognition service in case with voice control). The bot will answer user’s request with a text, voice, content or perform a command or an
- 41. 41 Case study: Fino app – AI in finance https://www.facebook.com/finoapp http://www.finomon.com
- 42. 42 Messengers with bot support Native
- 43. 43 Cloud speech to text providers and bot engines
- 44. 44 Case Study: Voice recognition and natural language processing Voice recognition and natural language processing is one of the most important things for IoT purposes. Voice recognition technique was implemented to prepare your favorite cocktail, e.g. you’re saying: “Scoofy, make my favorite drink!” and device makes your favorite drink based on your preference and previous history.
- 45. 45 Frameworks and tools for Deep Learning https://github.com/aymericdamien/TensorFlow-Examples