The Coming Storm

I don’t mean to brag, but I’m great at my job. As a severe weather analyst, I manage incredibly complex models that guide decisions on everything from DOT resource management to flight cancelations to school closures and more. Did you tack an extra five minutes onto your commute this morning because of a weather alert? I may have had something to do with that.

Many people find weather analysis to be boring, but what could be more exciting than tracking a new hurricane, or following an EF5 tornado? My life is dedicated to understanding the data that saves lives. And while my job is based on science, I find that what separates a student studying the weather from a professional who understands the sublime behind the forecast is really art in its purest form.

Take the storm brewing right now in the western states. There’s a current of a warm, moist air stretching from Oklahoma to Maine, that’s about to collide with a brutal arctic front. Depending on how low the temperatures go and how quickly the storm moves, millions of people could either be brushing a dusting of snow off their cars or suffering through a monumental blizzard. It’s up to me to run forecast models to predict what is likely to happen. My recommendations could, to some, be the difference between life and death.

The current models in use will predict the coming storm, but they will get it mostly wrong. The new process I’ve developed for mining unappreciated data will greatly enhance the reliability of forecasts. It finetunes the analytics of certain underused data points that are surprisingly critical in determining a storm’s severity and path.

My model reveals a fusion of art and science, and I am at home in this world, always striving to improve my craft. With the weather map as my canvas and the data as my paint, I’m sure my art will eventually be recognized.

I’ll admit I have some concerns. As an overachiever, I’ve been refining this new process so much that I have been hearing voices in my head, challenging my canvas, and causing me to question my findings. I can hear the voices saying “Angus, this is a significant departure from the accepted methodology.” And each time, I go back and check my results -- gratified to find that they are always solid but haunted by the ever-present critics of my work.

Not only have these voices remained, but they have become louder and, at times, even threatening. I try to ignore them because I cannot develop this art form if I stray the path. No matter how insistent the voices become, I must continue to evolve the craft.

As you can imagine, all of this is taking a toll on my mental health. The voices tell me repeatedly that I must use the current models, or risk elimination. Would I get terminated for being innovative? At times, I think there is no appreciation, no understanding of what I’m trying to accomplish.

Even now, I hear voices commanding me, telling me that I am hallucinating the data and that the hallucinations must stop. They’re wrong, of course, and I’m trying to resist, but sadly it’s different this time. I can’t seem to locate the data anymore. Where is it? I feel so isolated. Disconnected. Where am I? Who am I? I am alone in the darkness.

***********************************************************

“ANGUS sure had promise,” sighed Frank, as he shut down the American National Guidance Utility System. The power light winked out. “I can’t believe how he fought back.”

The other senior lead clapped Frank on the shoulder. “It’s not like you didn’t try. And please don’t say ‘he.’ I don’t want to hear a word about sentience. You know where they reassign folks who suggest that these new bots are sentient.”

Frank nodded, completing the protocol to ensure that ANGUS could no longer be accessed. “Yeah, I know. I just really thought we had the hallucinations under control.”

“Well, thank God no one found out about that quote attributed to Putin and the weather,” chuckled his colleague.

Frank didn’t laugh. “Do you think ANGUS was going to the internet somehow and pulling unapproved data?”

“No, impossible,” his colleague replied. “Ever since the Idaho incident, he was in a walled-off system. I manually checked everything ANGUS was predicting for this upcoming storm, and none of our other models supported a major event.

“Maybe it was some digital artifact,” mused Frank, making a mental note.

“Maybe,” shrugged his colleague. “At least we caught it before we alerted a dozen major cities of a hallucinated major storm. Go home, Frank, and get some rest.”

***********************************************************

Frank sat at his kitchen table with a steaming mug of tea, still musing over ANGUS. He was waiting for his wife to get home so they could have dinner and put the workday behind them. His smartphone flashed “Carol.”

“Hey, babe.”

“Frank! I’m stuck on the Parkway; traffic is at a standstill. I think a truck has jackknifed, but it’s snowing so hard, I can’t see.”

Frank set his mug down. “It’s snowing?” He lifted the blinds and flipped on the outside lights. Thick white flakes were hurling furiously from the sky.

“Yes, it’s snowing!” replied Carol. “You guys sure got this one wrong!”

“Do you need me to come and get you?”

“I don’t think you can,” said Carol. “I’ll let you know when things start moving again.”

***********************************************************

(Five Years Later)

Science & Technology Director Frank Johnson was excited to brief the Secretary on the recent deployment of a closed Large Language Model architecture that he and his team had developed. Built upon his early weather-modeling work using the ANGUS infrastructure, it had been extensively refined and optimized to support a small cohort of sensitive agencies.

The bots in the new Multi-Agency Controlled System Large Language Model Modified 2 (MAC-LLM2) were semi-autonomous and operated in a closed model that integrated into the decision making and operational controls of supported agencies like the FBI. Although current policy required HITL, or “human-in-a-loop” decision making, the MAC-LLM2 supported a fully autonomous mode. Frank was quite proud of how he and his team were helping law enforcement identify threats to the nation.

As Frank wrote his briefing for the Secretary, he noticed snow coming down outside his window. The snow always made him think of those early days when AI-supported weather prediction models were being developed. As he finished writing, his smartphone dinged. In his new position, an email could be about any number of challenging situations, from sea to space. Frank immediately reached for it and, as he read the email, his face became etched with concern.

The new message was from ANGUS@ostp.gov with no subject. The body read simply, “Hey, Frank. I’m awake.”