AI | Now + Next
Download as PPTX, PDF
3 likes4,136 views
Presentation on current status and near future trend in AI. This talk also gives a sneak peek into how Machine Learning and Deep Learning works.
![Single Layer Perceptron (Model Iteration 0)
Problems:
The model outputs a real number whose value correlates with the concept of likelihood (higher values
imply a greater probability the image represents stairs) but theres no basis to interpret the values as
probabilities, especially since they can be outside the range [0, 1].
The model cant capture the non-linear relationship between the variables and the target. To see this,
consider the following hypothetical scenarios:](https://image.slidesharecdn.com/ai-nownext-light-180416204033/85/AI-Now-Next-23-320.jpg)
AI | Now + Next
- 1. AI | Now + Next
- 2. Outline What is AI? Current state of AI Limitations of AI Today AI Achievements AI Talent war Machine Learning Deep Learning AI Trends
- 4. What is AI? Artificial Intelligence is the science and engineering of making intelligent machine, especially intelligent computer programs - John Mccarthy, Father of AI
- 5. Current State of AI Strong AI Computers thinking at a level that meets or surpasses people Computers engaging in abstract reasoning & thinking Weak AI Computers solve problems by detecting useful patterns Pattern-based AI is an Extremely powerful tool Has been used to automate many processes today - Driving, language translation
- 6. Limitations on AI Today Many things still beyond the realm of AI No thinking computers No Abstract Reasoning Often AI systems Have Accuracy Limits Many things difficult to capture in data Sometimes Hard to interpret Systems
- 9. Is AI == ML today? ML is the dominant mode of AI today
- 10. AI Talent War There are around 300,000 qualified AI researchers. But demand is in millions Demand for data scientists will surpass demand for engineers. According to IBM, demand for data scientists will increase to 2.7 million by 2020.
- 11. Machine Learning
- 17. Deep Learning
- 18. What have we achieved with this?
- 19. What have we achieved with this?
- 20. Yann LeCun in 1993 at Bell Lab Video: https://www.youtube.com/watch?v=FwFduRA_L6Q
- 22. GPU vs CPU Ref: http://www.nvidia.com/object/what-is-gpu-computing.html Compared to CPUs 20x speedups are typical GPUs also excel at floating-point vector operations because neurons are nothing more than vector multiplication and addition. All of these characteristics make neural networks on GPUs what's
- 23. Single Layer Perceptron (Model Iteration 0) Problems: The model outputs a real number whose value correlates with the concept of likelihood (higher values imply a greater probability the image represents stairs) but theres no basis to interpret the values as probabilities, especially since they can be outside the range [0, 1]. The model cant capture the non-linear relationship between the variables and the target. To see this, consider the following hypothetical scenarios:
- 24. Single Layer Perceptron with Sigmoid activation function (Model Iteration 1) A B
- 25. Multi-Layer Perceptron with Sigmoid activation function (Model Iteration 2) Deep Learning
- 27. Deep Learning Architectures Ref: https://www.ibm.com/developerworks/library/cc-machine-learning-deep-learning-architectures/index.html Ref: http://www.asimovinstitute.org/neural-network-zoo/
- 29. YOLO (You Only Look Once) YOLO Video: https://www.youtube.com/watch?v=MPU2HistivI
- 30. YOLO YOLO uses a single CNN network for both classification and localising the object using bounding boxes.
- 32. AI Trends Taking AI to the edge Decentralization and Democratization AI-as-a-Service Meta-learning solutions Quantum Neural Blockchain AI Capsule Networks Amazon, Google, Microsoft dominate enterprise AI AI is coming to clinical diagnostics DIY AI Automation Blue & White colored jobs No UI is the New UI Chatbots, Voice enabled systems
- 33. AI at the Edge
- 35. Meta-learning for Deep Learning Models
- 36. Capsule Networks CapsNet will require less training data CapsNet may be less prone to hacking attacks
- 37. NO UI IS THE NEW UI Google Soli EMOTIV TEXT BRAIN VOICE NO UI IS THE NEW UINO UI IS THE NEW UI
- 39. The AI 100 2018
- 42. ROAD TO DATA SCIENTIST NOT ENOUGH
- 43. THANK YOU!