Introduction of machine learning at Reflen
0 likes107 views
The document introduces machine learning and discusses how it has progressed from early successes in image and voice recognition to more complex areas like language understanding and reinforcement learning. It then describes a company called Reflen that is building algorithms, software, and platforms to support augmented intelligence and accelerate human cognition, with their goal being to create AI systems that act as "e-bikes for the human mind" to amplify human capabilities. Finally, it outlines Reflen's roadmap from minimum viable products to integrated solutions incorporating features like behavior understanding, intention prediction, and optimized decision making while protecting user privacy.
Introduction of machine learning at Reflen
- 1. Introduction of Machine Learning By Homer Quan
- 2. Our understanding of todays ML
- 3. OPTIMAZATION DATA UNDERSTANDING Two groups of problems
- 4. Breakthrough so far Image recognition Voice recognition Voice synthesizer Video 2. Structured Behavior understanding 1. Raw data understanding (in timeline)
- 5. Underneath CNN, RNN, LSTM, GAN, VAE GPU, TPU, FPGA Facebook, Google, Amazon, Apple Practice and technologies in Training method, hardware, and data
- 6. Alchemy or electricity? Mostly an engineering problem
- 7. What is emerging Reinforcement Learning
- 8. How about Language Its far more complex Source: PBS The Brain with David Eagleman 6 times reverse data to thalamus than input to visual cortex Internal model matters Memory(knowledge) matters
- 9. Opportunity Bicycles for the mind steve Jobs used to describe computers. We believe, today, A.I. system is an e-bike for human mind. First you need a good bike, then a good motor and battery, 鍖nally a good design to put them together.
- 10. We build algorithm, software and platform to support Augmented Intelligence. Re鍖en is an e-bike to accelerate Augmented Intelligence
- 11. How we do it Composite Learn by mimicking
- 12. Where are we
- 13. Roadmap MVP 1st Stage 2nd Stage Feature RT Sensing RT Engagement Knowledge management Product Convospot (Personal and Education) Feature Behavior understanding Intention prediction Interactive mimic learning Product Convospot Cognospot (Small business and Enterprise) Feature Mobile, IoT SDK Optimized decision Task allocation Plugins ecosystem Privacy protected learning
- 14. Thanks Re鍖en stands for re鍖ective learning. Q&A