Machines and the magic of fast learning
summary
Connects live data pipelines to machine learning and shows concrete wins: a nonprofit Thorne scans 100,000 public ads per day and faces 5,000% growth in exploitative images since 2007, using facial vectors to match victims. MemSQL adds high-speed vector dot products and Intel optimizations to cut positive-match time from 20 minutes to 200 milliseconds, a claimed 1,000x speedup, enabling law enforcement interventions that helped recover 2,000+ children in 2016. Architecture uses Kafka + TensorFlow + in‑memory DB for real‑time scoring.
AI-written summary · about
chapters (6)
- 0:00Start
- 2:02THORN SIFTS THROUGH MASSIVE AMOUNTS OF DATA
- 2:53THE MISSING DATABASE SYNTAX
- 3:32REAL-TIME HAS REAL WORLD IMPACT
- 4:24MACHINE LEARNING FOR IMAGE RECOGNITION
- 4:42BUILDING FAST LEARNING FOR IMAGE RECOGNITION
related talks
05mDriving the On-Demand Economy with Predictive Analytics
40mMachines are Learning
05mTeaching databases to learn in the world of AI
36mVideo analytics at scale: Deep Learning, Computer Vision, ML & Spark
27mMachine Learning's Missed Opportunity in Visual Data Management
01h 03mUsing Machine Learning and AI to serve people and communities
24mMachine Intelligence gets real
58mAI-Enhanced Big Data: Integrating Private LLMs and Vector Databases