Figure 1. Powering Artificial Intelligence
In a recent New York Times article, Mark Hung, a VP at Gartner, said that “almost everything Google is announcing these days is AI-related.” He said, “Google has a lead on artificial intelligence over many of its competitors, and it’s going to use that as a weapon to advance their products.” Although “weapon” is a bit strong to categorize Artificial Intelligence (AI), this technology is indeed making eyebrow-raising strides in many areas.
Case-in-point: IBM recently introduced a computer that demonstrates the capability of modern AI. The computer debated a person in front of a live audience. As far as who won? It was declared a tie. But it was a complete win for IBM in terms of unveiling an AI system which can construct a coherent argument and pull together information to make its case. This event was the most recent in a series pitting AI systems against humans: IBM’s Deep Blue beat Gary Kasparov at chess; Google’s AI beat the world’s best “Go” player, and a program called Libratus beat four of the world’s top poker players at no-limit Texas Hold ’Em.
IBM’s debate came a few weeks after Google released a virtual assistant called Duplex, which is capable of booking restaurants and haircuts by phone. IBM’s Project Debateris designed to make coherent arguments as it processes enormous amounts of data, but could it also recommend the right restaurant and meal for you, based on your lifestyle, schedule, and taste?
Applied AI: Chew On This
When TGI Fridays used AI to handle their customers’ questions and send targeted messages to them, the result was greater engagement on social media and more orders for the restaurant. Fridays used AI to double its to-go business in twelve months, increase engagement on social media platforms by more than five hundred percent, and triple the frequency with which guests placed online orders.
AI helped Fridays better target their customers. The result was the simultaneous improvement of customer experience while increasing revenue. As opposed to targeting a demographic such as women between 25 and 45, Fridays could know that Julia comes in at lunchtime for salads during the week, and on Fridays, she often calls late in the afternoon to place an order for pick up. And two of the meals are for kids.
Click image to enlarge
Figure 2. Everything Needs Power
AI utilizes data, collected from all those points-of-sale: social media, credit card transactions, and mobile devices, and analyzes it to make a personal decision for anyone of its four million patrons who has given the company permission to contact them. Information from across the company is sent to Microsoft’s Azure cloud. Technology from Amperity Inc. helps coordinate and organize the data, and AI systems analyze it in order to personalize messages for customers. According to Kumba Sennaar, writing at emerj.com, AI in the foodservice sector falls into four major categories:
Artificial Intelligence is Not An Element
Clearly, AI has embedded itself into our lives - and will continue to do so. However, AI is not an element on the Periodic Table, it must be programmed and fed, and data centers are its kitchen. But can AI help with data center operations while also helping to customize consumers’ dining choices?
Yes. In fact, Gartner also predicts that more than 30 percent of data centers that fail to sufficiently prepare for AI will no longer be operationally or economically viable by 2020.This statement seems to be accurate, especially when we consider this: How many variables can a human process at once to achieve an “optimized” operation?
The beauty of AI is the potential for it to scale and by ingesting massive volumes of data, most AI tools are able to derive a set of operating rules that can be more comprehensive in scope, more finely tuned, and more responsive to dynamic inputs than a person ever could. AI’s ability to comprehend and process more information than a data center IT staff running the facility clearly underscores Gartner's statement.
Google’s use of AI has already produced a 30% reduction in energy usage at its cloud data centers. AI controls the set point for the various heating and cooling systems based on sensors inside and outside the data center, and as workloads are moved into and shipped out of the data center, the AI responds accordingly.
AI will eventually dominate data center optimization decisions and having high volumes of measurement data is the first step towards enabling AI to work its magic. Measuring many facility components will generate lots of data points for AI to process and develop operational rules. After measuring, AI can logically take control of the systems generating the data. However, getting data center owners to turn over the operational keys to an AI system requires a “step-function” in faith that the AI will not quickly devolve into operational chaos.
Assuming operation chaos doesn’t occur, every system could potentially have AI deployed including water, air, heating, cooling, UPS, generators, fuel cells, battery rooms, electrical systems, networking, servers, storage, load balancers, security appliances, video cameras, and hand and retina scanners. Coordination between the systems will have to occur to prevent “thrashing” between disparate systems.
Artificial Intelligence Needs to Runs on Intelligent Power
With time sensitive workloads including certain AI applications now moving from the core and hyperscale cloud facilities out to edge data centers, the vast majority of the power that fuels them will need to be remotely monitored and managed through automation to ensure uptime while maintaining efficiency and avoiding those expensive truck rolls. Intelligent PDUs from Server Technology, such as the recently launched HDOT Cxwith Switched POPs, enable edge data center IT to maximize uptime and at the same time provide a rich data source for the latest AI focused on optimizing between minimizing power consumption reducing latency.
For cloud applications where scale, cost, and commonality are important, having intelligent PDUs also makes sense. HDOT Cxis available with a variety of feature sets and price points to suit even the most demanding of customer applications.
Click image to enlarge
Figure 3. Artificial Intelligence Runs on Intelligent Power
AI will likely take over all data center process controls because it’s a self-fulfilling prophecy. Today, there are servers that do not require manual cabling, instead, they have blind mate networking and power connections. Put a robot in the cold aisle to retrieve the servers, and deploy containers or VMs running on all the servers, and an AI system can:
Conclusion
Keep in mind that not every AI implementation yields the hoped-for outcomes right away and not every AI is going to be a V.I.K.I, HAL or a Colossus (The Forbin Project). However, just as Baby Boomers witnessed the first moon landings and the birth of the internet, today’s youth may never know a day without Alexa, Siri, Watson or Cortana in their lives. In the next several years we will see a continued push to mainstream adoption of AI. This will result in data centers being designed that are further optimized for AI enablement, interactivity, and exploration while minimizing cost, and maximizing efficiency and throughput. So rescind your membership in the debate club and let AI take care of feeding the kids, your optimal lifestyle is close at hand.
PDF
