Machine Learning Roundup 10/12/2016

Machine learning (ML) and deep learning (DL) content from the past 24 hours.

Note to readers: Big Analytics will rebrand as ML/DL soon.

Good Reads

Another day, another paper from Adrian Colyer. Your assigned 🙂 reading for today: Progressive Neural Networks. Instead of learning each new problem from scratch, PNNs explicitly support the transfer of knowledge across tasks. Thus, they are a step towards generalizability, a key problem in DL.

Babak Hodjat describes how AI changes online retailing.

Zymergen Raises $130 Million

ML-driven biotech startup Zymergen lands a “B” round worth $130 million led by Softbank and previous investors DCVC, True Ventures, AME Cloud Ventures, DFJ, Innovation Endeavors, Obvious Ventures, and Two Sigma. Zymergen uses robotics and machine learning to engineer microbes faster and better. Linkapalooza here.

Health and Medicine

Baidu introduces Melody, an AI-powered chatbot to connect with patients, field medical questions and suggest diagnoses. That’s nice. So now, instead of a person pestering you for your insurance card while they patch you up in the ER, an annoying chatbot will do so. Linkapalooza here.

According to a recently published study, a clinical approach that combines imaging and machine learning successfully identified cancer patients likely to benefit from a particular treatment.

Machine learning helps identify cancer cell types.

Issues

In the Wall Street Journal, Laura Mills profiles NTechLab, a Russian startup that markets facial recognition technology. The company won the University of Washington’s 2015 MegaFace Challenge, beating a team from Google. Mills notes that the firm is currently negotiating sales to state-affiliated security firms in China and Turkey.

AI won’t kill jobs, says Google DeepMind’s Mustafa Suleyman. That’s a relief.

Meanwhile, Daniel Matthews wonders if Google’s pursuit of AI is good for humanity. For starters, we can identify cat videos a lost faster than the old way.

Michael McDonald describes the potential legal issues for hedge funds if they don’t outperform AI-driven algorithmic investing.

Research

Huawei and the Berkeley Artificial Intelligence Research Lab announce a partnership for research into computer vision, machine learning, natural language processing, robotics and research planning. According to TechCrunch, Huawei will invest $1 million in the effort.

Companies

Alex Woodie interviews Databricks’ CEO Ali Ghodsi, who mentions that GPU-based deep learning is on the roadmap.

Applications

Smart cloud provider FittedCloud announces ML-driven optimization solutions for Amazon DynamoDB. FittedCloud uses machine learning to analyze usage patterns and automatically provision the database based on predicted workload. The company previously released a similar service to manage Amazon compute instances.

A startup named Restless Bandit offers a service that mines a company’s database of resumes for qualified candidates.

Lloyd’s Banking Group uses software from Pindrop to analyze voice “fingerprints”, detect fraudulent calls.

Scientists from École Polytechnique Fédérale de Lausanne (EPFL) in Switzerland and Penn State University in the U.S. develop a deep learning algorithm that can diagnose crop diseases from images taken using a smartphone app.

Autonomous Vehicles

Mercedes won’t struggle with tough moral decisions; its autonomous vehicles will prioritize passenger safety over pedestrians. Lawyers who represent pedestrians rub their hands together with glee.

California law now forbids automakers from advertising driver assistance features as “self-driving”, “automated” or “auto-pilot” unless the vehicle qualifies as level 3, 4 or 5 under SAE guidelines. Tesla is now under court order to stop advertising its Autopilot feature as such.

The California Department of Motor Vehicles issues a permit to Wheego Electric Cars enabling the startup to test its autonomous vehicles on public roads.

Hardware

Market Realist explains how NVIDIA threatens Intel’s lock on data center servers.

Semiconductor company CEVA prepares its fifth-generation processor core, the CEVA-XM6 for launch. The new core supports CDNN2, a software toolset for convolutional neural networks.

cevahardwareaccelerator

Software and Services

Dan Rowinski profiles Microsoft’s Azure machine learning services.

Montreal-based Algolux announces CRISP-ML, a product it claims can automate the optimization process for vision systems.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s