Starting a Substack: On the Machine Frontier

This is the Substack link. 

Here's what I mean by the Machine Frontier, a concept that was central to my book on technological catastrophes, Inviting Disaster.

An English dictionary of 1721 defined frontier as “the Limits or Borders of a Country or Province.” In the US, the word came to mean an unsettled wilderness that was ripe for exploitation by restless – sometimes reckless – entrepreneurs.

America's geographical western frontier has long vanished, but I see similar dynamics along the machine frontier: a mix of danger and opportunity that offers innovators new, sometimes dramatic, gains. But building out such niches brings a higher risk of error, since safe and reliable methods have not been established.

My posts, then, will highlight cases where new tech needed new thinking. 

Consider what happened in January 1969 at the Hungarian Carbonic Acid Producing Company, at Répcelak, Hungary. The company was in the business of removing CO2 from natural gas and selling it. The liquid was stored in small cylinders as well as in four big storage tanks, cooled by ammonia refrigeration. The gas arrived at the plant with traces of water in it that had to be removed. On occasion this stray water caused gauges, fittings, level indicators, and even safety valves to freeze shut. But the plant kept running.

On December 31, 1968, the plant shut down with the indicators showing at least twenty tons of liquid COz in each tank. The plant opened again late on the night of January 1. Running short of cylinders to store the liquid CO2, operators directed the flow into storage tank C, which was supposed to have plenty of capacity. About a half hour later tank C exploded, and its fragments blew apart tank D.

The twin explosions killed four people nearby and ripped tank A from its foundation bolts, tearing a hole about a foot across. In escaping furiously through the new opening, the pressurized, liquid COz acted like a rocket propellant. Tank A took off under the thrust, crashing through a wall into the plant laboratory, dumping out tons of liquid CO2 across the floor, and instantly freezing five people where they stood.

The deluge left the room at a temperature of -108°F, starved of breathable air, and covered with a thick layer of dry ice.

While popular history tends to focus on seemingly overnight successes, over the long haul the greatest rewards on the machine frontier will go to innovators who proceed on solid facts and exhaustive testing. When experiments are unsatisfactory, they correct and move forward. If that means scrapping a design and starting over with a clean sheet, that's what they do.

One use for AI: Keeping an eye out for forestry hazards

I visited my brothers' tree farm in Missouri last week, in time to help clear storm damage by operating their Cat skidsteer. Storm: as in a likely F1 tornado in early May that toppled many dozens of big trees on their property. First priority was to open the roads, followed by salvaging hardwood saw logs. 

This is me climbing into the cab after chaining a log onto the road, taking care not to trip on the grapple. Tripping here would be a good way to break a leg. 

Here's how their road looked before being fully cleared: A tangle similar to what big woods loggers called "slash." The biggest logs of white oak and walnut were close to three feet in diameter, disease-free and more than 150 years old. 

But logs weighing a ton or two aren't easy to salvage in this location. This was a steep hillside, dense with brush and rock outcrops. The cut log below reflects the slope angle. 

Note the tangled nature of the downed trees. My brothers have years of experience with chain saws and their risks, but commented that clearing trees in this setting poses a bigger danger than the saws themselves. And it can be hard to monitor those dangers, which make logging one of the highest-injury jobs in the country.

Consider that when cutting a log, the operator at the Stihl is watching the saw-cut closely, wary of several things. Often a fallen log requires an undercut, bringing the bar close to chain-damaging rocks. More pressing: is the log under a force that will send something rolling or flying when cut? Will it settle and bind the bar, or worse, catch the tip of the bar, flipping the saw backward?  

Given the need to pay attention at close range, it can be hard to see what might threaten from above or from a few yards to the side. 

Overwatch could be a useful job for an AI- and LIDAR-equipped phone once those are capable of real-time visual processing. Such a tripod-mounted phone could warn about initial shifting of what loggers call a widow-maker, a broken tree that's hung up on a standing tree, and ready to fall during logging operations. With knowledge of tree physics, it could warn that a branch has come under terrific compression and will lash out when cut. 

A person could do this job, but in my limited experience with small-scale logging, there's so much to do that nobody's available to do the overwatch job, however important.