Artificial Intelligence - Common Myths and Reality #DDD12

Transcript

1. Artificial Intelligence Common myths and reality *as of June 2017 @GaliyaWarrier
2. Sponsors
3. About me ENTERPRISE, Azure, Data (Azure) Data Solution Architect @ Microsoft Previously: BI Consultant Previously: Software Developer galiyawarrier (Double ;)) MSc. in Computer Science BSc from Samara Aerospace University
4. Agenda Myths REALITY CURRENT CAPABILITIES
5. Myth: AI will replace our jobs
6. Artificial Intelligence vs Machine Learning •Artificial Intelligence (AI): • A broad concept of machines being able to carry out tasks in a “smart” way •Machine Learning (ML): • An approach to achieve AI based around the idea that we should just give machines access to data and let them learn for themselves
7. Machine Learning Process Image source: https://mapr.com/blog/parallel-and-iterative-processing-machine-learning-recommendations-spark/
8. Deep Learning 101 Image source: https://hackernoon.com/log-analytics-with-deep-learning-and-machine-learning-20a1891ff70e
9. Microsoft CaptionBot https://www.captionbot.ai/
10. Autonomous cars
11. Tesla Autopilot Full Self-Driving Hardware (Video)
12. Reality: Some jobs will be replaced in 20 years
13. Myth: AI can think like humans
14. Narrow AI vs. General AI vs. Artificial Superintelligence
15. Microsoft Translator https://translator.microsoft.com/
16. Research project by MIT’s Computer Science and Artificial Intelligence Laboratory • Take a still image and create a short video (of the future) • “Adversarial” deep learning approach to train 2 competing neural networks (NN): • 1st NN generates video • 2nd NN discriminates between the real and generated videos • Over time, the generator learns to fool the discriminator • Trained on 2 MILLION unlabelled videos • Paper [pdf]: http://carlvondrick.com/tinyvideo/paper.pdf
17. Creating Videos of the Future (Video)
18. Reality: WE CAN make sense of images, text and sounds BUT LOTS OF DATA IS REQUIRED
19. Myth: AI MEANS full automation NO HUMANS REQUIRED
21. Citizen Science Projects
22. Humans are everywhere • Who builds a data set? • Who labels a data set for supervised ML processes? • Who interprets unsupervised ML results? • Human in the Loop ensures “high risk” conclusions are not leading to a real-world impact • People need to customise the AI tools to deliver a marketing promise
24. Reality: WE AS HUMANS ARE STILL IN THE GAME
25. Myth: AI will end our privacy
26. Data security & privacy issues are not specific to AI Source: Cloud Security Alliance – “Top Ten Big Data Security & Privacy Challenges“
27. Reality
28. Myth: AI will rule the world
29. AI is used a lot already • Virtual Personal Assistants • Home devices • Smart Cars • Fraud Detection • Online Shopping • Online Customer Support • Movie / Song Recommendation services • Smart Home Devices • Security Surveillance
30. Microsoft Video Indexer https://www.videoindexer.ai
31. A Wizard Hat or Implantable Brain-Machine Interfaces
32. http://waitbutwhy.com/2017/04/neuralink.html
33. Reality: AI is (AND WILL BE) A Part OF OUR LIFE
34. Thank you! galiyawarrier