By Tyler Glenn, CableLabs; Jason Rupe Ph.D., CableLabs; Kyle Haefner Ph.D., CableLabs
Enabling technical talent in network operations has been a challenge since the first network was created.
As technicians and engineers figure out how to plan, engineer, manage, and repair a communication technology, the next technology comes around and resets the learning curve. Like Sisyphus, it can feel like the rock rolled back down the hill and our task is to try again to roll the rock back up the hill for the next technology.
Frustration aside, the challenge is long standing, continuously evolving, and always becoming more challenging. As we get better at training and educating the workforce, and get better at managing and maintaining our networks, the network gets harder to manage and maintain as the performance bar is raised too. What is possible improves, so the bar floats above the performance line.
Training the workforce how to do the job is one part of the system. Another part is determining what is the right action to take. Knowing how to use a hammer is one part of training; when and where to use the hammer or not to use the hammer is another important part. The challenge is to train for situations that are highly variable and help them make good decisions.
But good decision making takes time to learn and be reinforced. Repetition is needed.
Operators can't afford to apprentice, meaning have an unexperienced person shadow an experienced person to learn. The usual approach to training is to instruct someone on the how, which will involve some aspects of the where and when, but expect they have learned over time (degrees, experience in related role, etc.) to get the rest of the way there. That's not always easy, possible, or the outcome.
Another approach is to create access to an expert. In the center, that can happen to a degree, but the expert may not always understand the situation and may not always have all the information needed to make the right decision.
Generative AI (GAI) presents a new approach: accumulate the knowledge of experts, encode it for fast access, incorporate situational information, and create the equivalent of an expert assistant to help the person do a better job. But instead of being simply a search engine, GAI provides the information in a way that is immediately useful to the human; instead of providing a likely answer or set of sources to read, it is provided as part of a conversation between the user and an expert.
This is a compelling promise, and as it turns out, it seems very reasonable to expect it will contribute well to this problem.