Top machine learning (ML) and deep learning (DL) stories from last week. Plus some good reads, and special Halloween content. The featured image comes from MIT’s Nightmare Machine.
— Aussie bots that hunt down and kill small animals.
— Bots with knives that cut flesh.
— Machines that hunt and kill humans.
— IBM Watson engines that profile you and sell you stuff in your car.
ICYMI: Top Stories of Last Week
— IBM announces the availability of IBM Watson Machine Learning Service; Bernard Marr and Ed Burns can hardly contain themselves over this great leap forward for machine learning. According to this IBM blog post, however, it’s simply a rebranding of an existing SPSS service:
The rebranding is obvious to anyone who logs in, largely because some IBM worker bee forgot to scrub the docs:
Pro tip for Bernard and Ed: check out what vendors tell you before you write.
— MSFT rebrands CNTK as Cognitive Toolkit, releases version 2.0 to beta. Enhancements include a Python API, Visual Studio support, reinforcement learning and performance improvements. Linkapalooza here. Dave Ramel describes Microsoft’s Cognitive Toolkit and IBM Watson Data Platform.
— Databricks adds support for DL with TensorFrames, a Spark package that enables TensorFlow to run on Spark. The managed service includes support for TensorFrames with GPU-accelerated compute instances. Alex Woodie reports, noting that Spark ML runs 10X faster on GPUs. Serdar Yegualp adds additional detail.
— Top analysts chew over Microsoft’s announcement that it uses Field Programmable Gate Arrays (FPGAs) to accelerate servers in its data centers. Karl Freund of Moor Insights and Strategy dissects Microsoft’s approach. In The Next Platform, Stacey Higginbotham delivers a tick-tock covering how MSFT decided to place its bet. Separately, Scientific Computing World reports on Baidu’s plans to adopt FPGAs to support machine learning.
— Researchers from Google Brain, Penn State, and OpenAI describe a method for semisupervised knowledge transfer from a “teaching” model to a “student” model. The “student” model never sees the original data, and cannot leak it accidentally. Dave Gershgorn reports.
— In Forbes, Peter High explains Gartner’s picks for the top ten technology trends for 2017, of which AI and machine learning rank #1.
— Researchers at University College London develop an algorithm that predicts the outcome of human rights trials with 79% accuracy. While that may sound impressive, bear in mind that if 79% of all trials result in a guilty verdict, an algorithm that declares all defendants to be guilty will be correct 79% of the time. We judge predictive models by how much they improve accuracy over alternative methods and not their absolute accuracy.
— MIT Technology Review reports on Microsoft’s accomplishment matching humans in conversational speech recognition, and why that matters.
— Researchers at Google Brain publish a paper on how neural networks can learn to use secret keys to protect information when they communicate with one another.
— In the Wall Street Journal, Irving Wladawaky-Berger asks: has AI finally reached a tipping point? In the process of answering the question, he identifies six hot trends in AI, including large-scale machine learning, deep learning, and reinforcement learning.
— Tanushri Chakravorty explains how machine learning works.