Humans have anticipated, feared, and revered artificial intelligence (AI) for as long as there have been computers. Today, whether they know it or not, anyone with a smartphone in their pocket is already reaping the rewards of AI and its counterpart, machine learning. As these technologies advance, the possibilities seem virtually endless. But how did we get here, and what lies ahead?
At our latest FutureTalk event at our Portland, Ore., engineering headquarters, a pair of AI experts sought to answer those questions and more. Sergey Razin, chief technology officer at the software-optimization firm SIOS, introduced the audience to the concept of “the self-driving data center,” while Alan Fern, professor of computer science at Oregon State University, explained how an AI system finally became world champion of the ancient game Go—and what that victory portends.
Rise of the “Self-Driving” Data Center
Sergey, who has previously held positions at Kaspersky Lab and EMC, has been researching and working on machine learning for more than 15 years. Now, he said, almost everyone claims to be using machine learning, often falsely. As he puts it, “There’s a lot of fluff in this space.” Still, with speech recognition in our phones, algorithms in our streaming services, and self-driving cars rolling fast down the road, it’s clear that AI has hit mainstream IT.
Why? Because the technology has evolved. Data has gotten bigger and more multidimensional, and computers have gotten infinitely more complex. With so many resources interacting over any given hardware device, the “blender effect” threatens to cripple even the most robust infrastructure. Conventional tools can be overwhelmed by noise, leaving users with plenty of data but zero actionable insights.
Sergey and his colleagues believe they have found a solution in SIOS IQ, a new product that seeks to apply machine learning to software development and operations. Through a process called Topological Behavior Analysis (TBA), SIOS’ tool obtains information from the system’s infrastructure, extracts relationships, and characterizes individual metric behaviors.
That data is then transposed into multidimensional space and fed into root-cause algorithms based on probabilistic models. By casting behavior into visible shapes rather than focusing on specific clusters, TBA can reveal interesting information that might otherwise be lost in the noise.
All of which paves the way for the “self-driving data center.” In the near future, Sergey said, the brains behind any robust IT operation is going to be a machine learning algorithm.
Man vs. Machine on the Go Board
When Deep Blue beat chess grandmaster Garry Kasparov in 1997, it seemed that machines had officially surpassed humans when it came to games of skill. But, as Alan explained, it took another two decades for an AI system (specifically, Deep Mind’s AlphaGo) to defeat a human Go champion.
The challenge was to fuse machine learning with the intuition of human Go experts, and the answer lay in treating the Go board as an image. Through an impressive combination of deep neural networks, the Monte Carlo Tree Search, distributed high-performance computing, huge quantities of data, and reinforcement learning, Deep Mind was able to fashion a truly formidable Go player. Earlier this year, AlphaGo finally defeated international Go champion Lee Sedol in a $1 million challenge.
To discover how Alan wants to apply the lessons of AlphaGo to real-world policy issues (such as stabilizing the electrical grid), and to hear why Sergey thinks you should start brushing up on algorithms ASAP, watch their full FutureTalk in the video below: