Every year, technologists wait for the AWS re:Invent 2018 conference in Las Vegas like a four-year-old on Christmas Eve. At this event, Amazon Web Services announces most of their major product and feature releases, and this year was no exception.
At Delta Risk, we pay close attention to these releases every year for ideas on ways we can improve our technology. We do this with three specific goals in mind: gaining more insight into the data collected from our customers’ cloud environments, making decisions faster (both automated and human-based), and processing data in the most efficient way possible.
Several of the headline-grabbing releases at this year’s conference centered on physical things. This included robots, autonomous cars and satellite base station communications. While those attracted a lot of attention in the mainstream press, the Machine Learning as-a-service capabilities that those solutions rely on, announced earlier this year, really stole the show.
AWS Ground Station
Another interesting announcement at the AWS re:Invent 2018 Conference was AWS Ground Station. This service will eventually allow organizations to communicate directly with satellites all over the world. We’ve all heard about miniaturized satellites that are relatively cheap to produce and can be launched en masse. But, without the ability for smaller organizations to communicate globally on the ground with these satellites, there is a missing link.
With Ground Station a developer could use the AWS satellite communication facilities and network to upload and download data from their satellites directly. They could do this using their AWS Virtual Private Clouds (VPCs). This way developers can focus more on their applications instead of on purchasing land, running facilities, and managing their network. Bringing satellite communications to startups is an exciting concept and the possibilities are wide open.
The new service AWS RoboMaker is another item that could be of interest to tinkerers. RoboMaker includes tooling, development, and simulation that support the open source Robot Operating System, or ROS. This tool lowers the entry cost for developers so they can quickly build robots that will do just about anything. But the most fascinating part is that developers can leverage AWS machine learning and new robot fleet management capabilities to build intelligent robots that learn new behaviors and deploy them at a large scale. Think of hundreds of robots in a warehouse constantly improving the efficiency of their tasks, running 24-hours-a-day.
The announcement with the biggest tinkerer buzz, AWS DeepRacer, started as a competition. DeepRacer is a 1/18th scale self-driving car designed to be taught to race around a track. It comes equipped with cameras and sensors just like the latest driverless cars on the road today. The AWS machine learning platform is tightly integrated in with DeepRacer. After many simulations and physical attempts to perform a lap on a miniaturized racecourse, the car operating system will start to learn the quickest and most efficient way to navigate the racecourse. The operator or developer can program the algorithms that AWS SageMaker will then run in preparation for the race. Even though DeepRacer seems to be a toy, it’s really a machine learning tool for developers and SageMaker, and could be a way to lower the barrier for startups looking to develop self-driving cars.
These developments are interesting manifestations of the most important advancements of the past year in Machine Learning and Artificial Intelligence at AWS. Machine learning has historically been a technology relegated to research universities and larger commercial or governmental enterprises. Not only has the hardware and software been expensive and complicated to setup and run, but the skillsets required are not necessarily something that small to midsize organizations have had in house. SageMaker was released in 2017 to lower the bar for organizations to be able to explore these technologies. The releases this year represent an expansion in machine learning based on a year’s worth of feedback from customers, including advancements in Reinforcement Learning, Dynamic Training, and algorithm dataset training. Additionally, ML-related hardware advancements such as new inference GPUs, developer tooling and ML orchestration and workflow capabilities were announced.
Within our ActiveEye native-cloud security platform, machine learning techniques are used to inspect terabytes of our customers’ data to find security anomalies. This data is based on past patterns as well as predictive analytics to determine likelihood of future data breaches. We evaluate every new advancement in machine learning technologies to help our customers secure their environments with better insights, speed, and efficiency. SageMaker allows us to try different algorithms, training, and deployment models without having to invest in hardware or architectures. This approach allows us to quickly experiment and fine tune our machine learning workflows.
Did you know? Delta Risk has just achieved AWS Advanced Partner Status within the Amazon Web Services Partner Network for our ActiveEye Cloud Security Platform. Learn more here.