Tina Huang is the Founder and CTO of Transposit, a platform that bridges the gap between dev and ops with automated human-machine workflows.
Since artificial intelligence (AI) was invented in the 1940s, chess has been an ideal testbed. To win, a chess player must consider multiple moves ahead, the likely responses to each move and how to develop an effective strategy. Huge amounts of time and money have been poured into automated chess-playing programs.
The first program-assisted chess algorithm was created by Alan Turing. Fifty years later, IBM’s chess-playing supercomputer, Deep Blue, emerged. Russian chess grandmaster Gary Kasparov played against Deep Blue, and Kasparov was victorious. Even as AI had matured drastically since Turing’s time, a human was still able to beat a machine. The true power of AI would come later as people realized that humans alongside machines is the winning combination.
The progress of automation in chess has many lessons for technology operations (TechOps). In both domains, where a human operator makes complex decisions, the ability to anticipate future moves and analyze the impact of moves taken generates fewer mistakes and better outcomes.
Why Chess Is A Favorite AI Proving Ground
Chess requires planning for future events, making chess and AI good bedmates. By creating an automated system for the real world, chess becomes a defined starting point where AI concepts can first be tested.
However, TechOps scenarios are not nearly as structured. When responding to an incident or creating a new system, the circumstances are large and fluid. The key is to get humans as much relevant and integrated information as possible. In some instances where the rules are well-known, AI can go further and make recommendations. Over time, TechOps teams will understand more of the environment and can provide valuable information, expanding automation.
Advanced Chess: The Human And The Machine
After chess-playing programs became widely available, the combination of humans and chess-playing programs performed better than either did individually.
In advanced chess, players use a program to explore the results of moves. Still, it is the human who controls the game. An advanced chess player marries human intuition with a computer’s ability to remember and calculate a staggering number of moves, countermoves and outcomes.
By design, advanced chess brings together human and computer skills to increase the level of play and reduce potential mistakes. For TechOps, amplifying humans through automation is profoundly important because the situations are much more complex and dynamic than chess. The question for TechOps practitioners: How can adding computer-based automation and analysis improve their game?
Until recently, incident response has been a manual effort prone to human error and inconsistencies. AI and TechOps is the perfect opportunity to converge humans and machines. Automation enables IT professionals to tackle incident response more effectively, giving teams clearer sightlines into what occurred. However, TechOps teams still need the requisite knowledge and skills to resolve issues.
AI-driven automation allows teams to confidently apply that knowledge faster. It can also enhance the tools and processes needed to identify, interpret and resolve issues at the software, hardware and infrastructure level.
The ideal form of TechOps, like advanced chess, combines a system for analysis, decision support and automation with a human operator. The human player is provided with a rich set of integrated information, analytics and recommendations to help them take the right actions. AI-augmented TechOps, like all automation, is leveraged incrementally, from simpler tasks toward more complex.
Here’s how TechOps can benefit from the type of AI, analytics and automation seen in chess.
• Data integration. Data from across systems needs to be parsed into structured data that humans can analyze. Data integration does this so a person can interact with masses of data in meaningful ways. The data integration automation focuses on the context that the TechOps person is facing, automatically organizing the data related to an incident so the human can make sense of what’s happening.
• Automated analysis and understanding. Based on a set of integrated data, various types of analytics can be run for an in-depth understanding. While data integration sorts all of the IP addresses making requests by volume, analytics can divulge more about the top 10 IP addresses. Or, if the data shows increased activity on payments, teams can search for all recent changes on that service. Here, the human controls the filters, and the machine does the thinking underneath.
• Automated discoverability and recommendations. Discoverability is about evaluating all possible options. Teams need to gain an understanding of the environment using different searches and filters that identify unusual or anomalous activity. With options on the table, a TechOps system may also recommend which is best given the context, like suggesting the most common runbook used during an alert.
• Augmented machine analysis. Next, the computer helps evaluate detailed analytics to support a problem-solving path. For example, a TechOps decision support system can provide more granular information when alerting the person on-call about a high load from particular IP addresses. The machine might identify the IP addresses associated with a specific partner, requiring a human to contact the partner to request they throttle their traffic.
In both chess and TechOps, the level of automation depends on the understanding of a situation. In TechOps, automated alerts that happen frequently can be resolved through automation — but only after the situation is understood.
Humans With Machines Make The Most Formidable Opponent
Even with a more visible and predictable environment, chess has demonstrated that humans and machines work better together. While end-to-end automation may seem desirable, it’s neither feasible nor optimal for TechOps. A system for supporting TechOps can make it easier to gain that vantage point to visualize the environment better. As operations teams get a granular idea of what’s happening, they can ask increasingly complex questions that deploy specific, higher-level analytics.
While the aim of AI is to make a machine act like a human, AI-augmentation isn’t designed to replace humans. Instead, the goal is to take advantage of the best capabilities of humans and technology. In TechOps, just as in advanced chess, the partnership between humans and machines creates the most formidable team.