Techexpo bigdata ml_ai_hanoi
1 like231 views
This document discusses advances in artificial intelligence, machine learning, and deep learning. It provides an overview of key topics including: - The relationship between big data and machine learning/AI and how machine learning has evolved from programmed systems to deep learning approaches. - How major tech companies like Google, Facebook, Microsoft, Amazon, and IBM are heavily investing in and applying AI/machine learning across many areas from search and recommendations to personal assistants. - Examples of real-world applications of deep learning like computer vision systems, machine translation, and AlphaGo as well as challenges that still exist around training deep learning models at scale.
Techexpo bigdata ml_ai_hanoi
- 1. (Big) Data, (Deep) Learning and AI Phạm Thành Lâm | Founder @ SaigonApps 17.09.2016 When big data hits machine learning
- 2. Big Picture: Big Data - Machine Learning/ Data Mining
- 3. History of AI, Machine Learning and Deep Learning Image taken from:blogs.nvidia.com
- 4. From Program To Machine/Deep Learning Image taken from the Internet
- 5. Image taken from: http://bit.ly/1i4e8oL, kdnuggets.com Gartner Hype Cycle Emerging Technology 2015
- 6. In 2016, is Big Data still a “thing”? - Enterprise Technology = building a data-driven culture, where Big Data is not “a” thing, but “the” thing - The Ecosystem is Maturing (let see the picture) - Big Data infrastructure: Still Plenty of Innovation - Big Data Analytics: Now with AI Credited by
- 8. AI: Artificial Intelligence → Applications and Innovations
- 9. Image taken from: cbinsights.com, econotimes.com Top acquirers AI startups MA Timeline
- 10. Tech giants(FAGA) embracing AI Google Facebook Microsoft Other - TensorFlow DL framework and Tensor Processing Unit (TPU), a custom ASIC chip built specifically for machine learning - 100+ different teams working on Google Today, Street View, Inbox Smart Reply, voice search, Google Play, etc. - Magenta to play music - DeepDream for creative pictures - WaveNets: speech synthesis, music creator - Fblearner Flow the tool, designed to help engineers build, test and execute machine learning assembly lines, is available to every engineer within the organisation like Deep Text - Messenger platform, allowing businesses to create AI-powered chatbots to interact with their customers - Tay, an artificial intelligence Twitter chatterbot, released by Microsoft in March - Cortana – its equivalent to Apple’s Siri and Android’s Google Now – an artificial intelligence-powered personal assistant and knowledge navigator for Windows’ Phones - London-based AI startup Swiftkey is acquired in February Amazon: unveiling DSSTNE, an open-source AI framework developed to run its recommendation system IBM: Watson/Connie, IBM’s AI computer system is able to answer questions posed in natural language, Bluemix apis. Sony: undisclosed investment in Cogitai, a one year old California-based AI startup Info is curated from: techcitynews.com
- 11. The pioneers of AI/ML/DL: (my bias) Geoffrey Hinton - Google Yann Lecun – FB Bengio Yoshua - Montreal University Xavier Amatriain – Quora/Netflix Demis Hassabis – DeepMind Andrew Ng- Baidu GodFather of DL, IEEE awarded 2016
- 12. Real world AI/DL applications Image taken from: Luong’s Machine Translation slide
- 13. CẦM KỲ THI HOẠ Prisma Google Brain Magenta Alpha Go Image taken from:tuoitre.com
- 14. Sample poetry No. he said. “no,” he said. “no,” i said. “i know,” she said. “thank you,” she said. “come with me,” she said. “talk to me,” she said. “don’t worry about it,” she said.
- 15. Image taken from: Andrew Ng twitter
- 16. Limits and challenges of DL/ML Image taken from Internet: wsj.com, twitter.com
- 17. Training DL is painful • Tuning hyperparameters • Network architecture: layers/nodes • Some data preprocessing • Weight initialization: ~N(0,1) • Learning rate, optimization algos • Slowness • Overfitting • More ...
- 18. We know now that we don't need any big new breakthroughs to get to true AI That is completely, utterly, ridiculously wrong. As I've said in previous statements: most of human and animal learning is unsupervised learning. If intelligence was a cake, unsupervised learning would be the cake, supervised learning would be the icing on the cake, and reinforcement learning would be the cherry on the cake. We know how to make the icing and the cherry, but we don't know how to make the cake. We need to solve the unsupervised learning problem before we can even think of getting to true AI. And that's just an obstacle we know about. What about all the ones we don't know about? Yann LeCunUNSUPERVISED AND TRANSFER LEARNING
- 19. How to build ML/DL from scratch Image taken from Internet
- 23. Some Demos
- 24. Find me: @laampt | Github: lampts