Now Available: Pumpkin Spice Machine Learning
Recently, High Availability was accepted as a Nvidia East Coast partner to sell and manage their Nvidia DGX-1 GPU platform, as well as reseller of PureStorage's new AIRI (AI Ready Infrastructure) and the Cisco UCS 480 ML M5 GPU Server.
When you think about PureStorage Orange, Winter time, and High Availability, what comes to mind? Pumpkin Spice Lattes
There have been few trends as prevalent as the Pumpkin Spice craze of the last decade. From Pumpkin Spice Lattes to Latkes, Pumpkin Spice PopTarts to PopRocks, Pumpkin Spice Beard Oil to Soaps, for 3 months of the year it's difficult to avoid.
When we think back upon the history of Pumpkin Pie and Pumpkin Spice, we see through the trend. The first recorded recipes were from 1651 by Francois Pierre la Varenne, in 1763 Amelia Simmons put her spin on it by including a recipe in America's first printed cookbook, and in 1934 McCormick & Company made it mainstream by introducing their prepackaged "Pumpkin Pie Spice Blend".
What we are now seeing as a "trend" in AI, Machine Learning, and Deep Learning, is certainly not new. As we can see illustrated below, the foundation math and scientific building blocks for that we now call Machine Learning and Deep Learning have existed for as long as the original "Pumpkin Pie Spice Blend".
For those readers which have never heard of Machine Learning or Deep Learning, or never thought about what AI really is, let's take a step back and look at them at a high level.
- AI/Artificial Intelligence: Give a computer the ability to follow a decision tree to make choices on input. This can be as simple as IF-THEN statements, or Use Cases. A basic rules engine. If a caller presses 2#, transfer a call to Bob.
- ML/Machine Learning: ML is part of AI, but AI may not be part of ML. With Machine Learning, a computer or program learns over time and can be trained in a fashion. As you use daily, machine learning will become better at keeping your mail Inbox free from spam, as you classify what spam is to you.
- DL/Deep Learning: If you wanted to labeled a million pictures of dogs with their type, the program could learn the difference between a Poodle and a Pekingese, or Dachshund and a Hot Dog with mustard. The more input, the better it gets. Deep Learning does not stop there though. Deep learning algorithms may reprocess aspects to get a higher "probability vector" for what it lacks confidence on. OpenAI's FIVE video playing AI team learns as it plays and is constantly trying to best human players in DOTA, learning from new techniques that the humans throw at it.
How does this help your Pumpkin Spice cravings?
Farmers and Transportation companies have been early adopters of GPU accelerated machine learning. Nearly every aspect of planting, harvest, and distribution of our food is a miracle of technology.
- Seed Designers are using machine learning to predict preferred traits for new seed hybrids for different regions and soils
- AI and Drones are utilized to monitor crop needs, and peak harvest times
- Automated visual inspection of crops determines crop health
- John Deere and Blue River have solutions to reduce pesticide and fertilizer use
- GPS and Satellite enabled driverless tractors automatically harvest crops based on crop types and data
- Trucking companies utilize machine learning to determine best routes and maximize transport load distribution
- Self-driving trucks by Tesla, Volvo, Waymo, and Daimler are beginning to be delivered to trucking companies, since there is a shortage of truck drivers. An HA customer is already testing these trucks on the road.
- Maersk and other sea carriers utilize machine learning to route shipments and track shipping containers
- Coffee Shops are using ML for big data analytic, customer preference and targeted advertising.
In the last few years, we have advanced to a level of technology where the hardware to utilize such approaches is affordable to every business and individual. Simultaneously, the software to make best use of this hardware is much more easily integrated into business workflows.
The advancement in the last decade has come from Nvidia's investment of billions of dollars to scale their GPUs from gaming only cards, to the de-facto choice of computational processing. Nvidia has created software and libraries which allow customers to integrate and expand their existing environments. While Virtual Desktops may be the first thing legacy IT organizations think about with Nvidia GPUs in the datacenter, that is just a small piece of the pie.
Today, NVIDIA has a catalog of 550 applications 3rd-party which are GPU accelerated. This does not include the tens of thousands of home-grown applications and systems in use which may use their libraries, cuBLAS to accelerate anything with linear algebra, or one-off algorithms from applications such as R.
The big winner is you. Whether you are looking to use off the shelf analytical software, or develop with your own algorithms via a team of Data Scientists, we have solutions for it. From individual GPUs, to Nvidia DGX-1 or PureStorage AIRI or Cisco UCS 480 ML M5, we can help. If you do not have your own Data Scientists, High Availability has partnered with some of the best in the country. Contact your local H.A. Sales Rep for more information.