Get ready. Closed loop automation and AI are coming to the cloud

If you’re one of the 170,000 people currently working in a network operations center or cloud control center in the U.S., look out. Artificial intelligence-enabled network automation technology is coming for your job, with an inevitability rooted in historical precedent. The only question is when?

Deploying network automation capabilities has been a priority for telcos and large enterprise companies for years, and current inflationary environment is putting even more pressure on them to use automation tech to cut costs and increase efficiency.  

But until now, companies have stopped short of allowing automation to run the network show – a practice known as closed loop automation, or autonomous automation. 

“No-one has gone full closed loop yet because they’re all worried about automating off a false positive,” says Kailem Anderson, vice president, portfolio and engineering at Blue Planet, a division of Ciena and a network automation market leader.

Instead, service providers and enterprises are starting to implement partial closed loop automation, where the network operating system is given autonomous authority over low-level, mundane tasks which won’t have a catastrophic effect on the network if they get them wrong (for example, a network connectivity check, but not critical traffic routing decisions).

Kailem thinks Blue Planet’s customers will move to closed loop automation in the next 10 years. However, there are several reasons why that timeline could be moved up.    

Three degrees of automation

But first, some background.

The first generation of network automation products consisted of chained scripts written by hand using a command line interface. The second generation of automation tools consolidated network tasks into pre-assembled programs that can be selected, scheduled and executed from a graphical user interface. Third generation network automation goes further still, putting the job of interpreting network events and activity, and acting on them, in the hands of artificial intelligence (AI). It’s these third-generation products that represent an existential crisis for anyone that works in the communications industry.

So far, service providers and enterprises haven’t given their third-generation automation products license to work in full robot overlord mode (a.k.a. closed loop automation).

Instead, their current MO is to let AI-enabled automation recommend actions, but require that they are approved, or not, by a human being, sitting in the cloud or network operating center (NOC) – a binary job that is on a direct evolutionary path from the 19th century telephone switchboard operator.

“With any system or process, the outcome is only as good as the inputs,” says Bart Giordano,  President - Networking, Intelligent Cellular and Security at CommScope. The role for humans [today] is to provide the right inputs and be cautious with the outputs. But [in the future] the productivity benefits and the economic gains from automation are going to be simply fantastic — beyond what we can imagine.”

With networks running literally at light speed over fiber optic cable, between servers that are so powerful they create their own gravity well, in the long term the odds of having a human being made out of meat being actively involved in the decision tree quickly becomes vanishingly small.

Doing so could also be an extremely costly mistake. High-frequency trading (HFT) applications require latency under 1 millisecond. The only way to guarantee that performance is through autonomous automation, and HFT markets already use AI to react to minute price discrepancies and analyze trade execution patterns. In other words, HFT networks is primed to move to move to closed loop automation.  

The convoluted cloud

Increased network complexity, exacerbated by new and disruptive cloud technologies, is another driving factor for closed loop automation.

“As we moved into this cloud-native world, things didn't get easier [for customers] like they were supposed to, they got more difficult,” says Mark Bunn, senior vice president of SaaS business operations for Nokia Cloud and Network Services.

Network automation vendors are responding by developing automation functions that move beyond the provisioning and assurance layers that sit above the network, down into the telecom network stack itself, helping operators to manage intricate next-generation cloud infrastructure.

“Every vendor in the industry is trying to build that [automated] end-to-end stack,” said Bunn, whose division of Nokia is also focusing on adding another layer of automation to the telecom managed services that run over the stack, creating economies that benefit total cost of ownership (TCO) and time to market for its customers.

Below is a list of leading automation proponents. Click on the company name to view what they are working on: 

Planes, trains and automation

Other industries have committed to automation faster and harder than the comms industry. What can be learned from their examples? Take a look.

Aircraft manufacturers pioneered automation, and still pay human pilots a quarter of a million dollars a year to play sudoku on long-haul flights despite the fact that modern planes can both land themselves, and taxi and take-off automagically, without any human involvement (something the carriers tend not to broadcast).

This is in part because the optics of a pilotless plane crash would be generally poor, and bad for business. Don’t expect telcos and enterprises to feel the need to take an equivalent approach. Crashes on optical networks happen silently in the dark somewhere out there in an opaque cloud.

Will safety concerns stop the move to closed loop automation? Not if the UK rail industry is any indicator. Automation was part and parcel of the privatization of British Rail in the 1990’s, which was quickly followed by four major fatal rail disasters. But the new train companies continue to cut costs to the bone by using automation to eliminate train guards, station managers, ticket clerks and control room engineers, until most stations now resemble eerie ghost towns without staff or, often, trains.

Given the history of automation, there is a high probability that, in the early days, communications networks will go through a period of close loop automation/AI incepted catastrophes, but this doesn’t mean service providers and enterprises won’t implement the technology at the expense of human operators anyway. They will feel that economic and competitive forces will leave them with little choice.

Glass half full

It's not all doom and gloom for the hoomans. In the next few years, anyone working in cloud infrastructure is relatively safe. “There still aren’t nearly enough of those people and they’re all terribly overworked,” says Andrew Coward, general manager, software networking at IBM.

Further, every automation vendor spoken to for this article said they thought existing cloud techs and network operators could be retrained to focus on business strategic tasks.  

“There's already a new job for them,” points out CommScope’s Giordano. “It’s called a prompt engineer, and that person is responsible for working out the most effective way to take advantage of those AI capabilities, accelerating the delivery of the business outcome to the customer.”

Dystopia beckons

No article covering AI would be complete without an acknowledgment of the growing concern over the potential disastrous outcomes of implementing AI.

AI is getting smarter. This is because AIs learn in multi-cast. If you teach a man to fish, he can teach other men to fish.

AI isn’t like that — you teach an AI to fish and suddenly all the AIs have gone fishing. If they wanted to. Which they probably wouldn’t.

What they might want to do, though, is sit around looking at their human makers (dispassionately because AIs don’t do loyalty or empathy — yet) and decide that, net of net, humankind is bad for the planet, and things might just work a lot better if they just got rid of us.

They could do that as easily as turning off the electrical grid. Various studies — including from the U.S. government — predict it would take between two and three weeks between the power being turned off and total societal collapse (think former venture capitalists sitting next to I-280 gnawing gristle from the tip of a human femur).

Even if that doesn’t happen, by 2050, communications networks will have changed out of site and mind from what we know today. 

The cloud control center (formerly network operating center) will go the way of smoking in the break room. Its functions will either be distributed or consolidated into data centers filled with quantum computers, encased inside vacuum-filled, dust free, sealed compartments sitting in the pitch darkness in the depths of cold oceans, or the void of outer space. There’s no room for humans in this cloud ops future. Of course, depending on what the AI overlords decide, we may not be around anyway.

Read more columns by Steve Saunders here.