NOCMACOC is pleased to report preliminary results of research into applications of artificial intelligence (“AI”) in white-water paddle-sports.

Self-driving cars have already proved so successful that public clamor has risen for similar advances in whitewater paddle-sports.  Road vehicles guided by artificial intelligence use fuel more efficiently, eliminate operator-induced accidents, and raise prospects of vehicle sharing which will eliminate misallocations of societal capital -- a Win-Win-Win. While traffic congestion and fuel use aren’t problems on rivers, a different set of problems beg solution: (i) injury and loss of human life, (ii) expensive equipment losses, and (iii) anxieties created by stressors of river obstacles in remote environments. As well: we at NOCMACOC were interested in this complex mathematical situation.

Can you tell that this press release is itself written by a machine? No? Good!  NOCMACOC’s principal goal is to meet all pertinent human benchmarks.

By way of explanation: the important thing in machine learning is iterations.  We use computers with vast data-crunching abilities.  Such computers are fed real-life situations over and over and over.  Machine learning occurs as the computers resolve optimal behaviors based on very large data sets. At some point naturally the machines begin performing better than fallible humans.   Afterwards, left to their own resources, the machines reach perfection.  But to assemble the inaugural data sets, humans are needed. 

Initially humans must paddle downriver accompanied by computers gathering data. Afterwards, when the machines are taking their turns, humans must escort them, occasionally modelling corrective behaviors.  The project therefore involved recruiting “trainers” and “test pilots” - canoeists and kayakers skilled in the Class III to V rapids. In and around North Carolina plenty of these existed. Thinking themselves adventurers, they called themselves "hair boaters" who "chased rain".

Some of the first experimental challenges were (i) data-gathering obstacles and (ii) person/equipment losses.  It was found that the “test pilots” were “rowdy” type humans - unruly because without societal role or function. Before becoming involved with the experiments, the test pilots had been living on junk food, dried fruit and granola, rolling out sleeping bags next to their dirty stickered vehicles along forested backroads, paddling over waterfalls, abusing substances, and failing to maintain steady jobs. The consequence of employing these “dirtbags” for our research, summarized, was:

• They corrupted datasets by supplying all kinds of false and made-up terminology, though NOCMACOC made progress iteratively in recognizing vulgarisms. Some phrases were kept as machine-learned terminology, e.g., “bites the big one”. (AI frontier: 100% facility with vulgarisms.)


• As the “test pilots” came to grasp the import of the experiments and their possible effect on the pilots’ lifestyle, etc., NOCMACOC began experiencing equipment losses in increasing frequency. Phenomena captured on videos indicated ‘chucking junk overboard’.


• For obvious reasons, “test pilots” needed to wear headsets and also electrodes fastened to their scalps. A challenging human protoplasm-electronics interface was encountered.  Inundated, the electrodes tended to “short out” across the shortest distances, e.g., between ear-holes. Such “test pilots” would be said to “amputate their mahoganies” or “go off their chumps”. Many pieces of electronic equipment were lost too in this iterative process. “R.I.P.”

It shouldn’t go unsaid that on the banks of the pertinent rivers were oftentimes human observers from (i) grants organizations, (ii) tourism, outfitter and sporting goods industries, and (iii) extremely rural persons.  All were measurably observed to be “rooting for the humans”. The rural persons’ efforts to sidetrack AI data gathering (self-categorized as “just farting around”) could not be dissuaded without strong prospect of legal authorities’ punitive measures.  (AI frontier: determining probability of action by human peace officers.)

The “test pilots” experienced reductions-in-forces. When research resumed, virtually all non-NOCMACOC human observers were banned.  In lieu of hair boaters, NOCMACOC sought out non-whitewater-savvy humans to make shoulder/arm motions in boats with electronically-implemented paddles mounted on the boats’ gunwales.  Gratuitous monitoring of heart beat, respiration, hydration, electrolyte level, bone health and blood platelets continued as to such “partial humans”.  

Partial humans proved subject to some of the test pilots’ frailties.  Thus NOCMACOC initiated replacement of humans with colloidal suspension bags. Such bags were shaped like humans with appropriate appendages, but were filled with colloids the consistency of tapioca or frozen blue slush used in Yeti™ mobile refrigerant products. Unfortunately, colloidal suspension bags were not cost-effective compared with local humans.  

At the very end of the research, human persons were reinstituted (i) as a means of addressing politically sensitive North Carolina unemployment/substance abuse issues, and (ii) because of the need of human involvement under terms of certain federal research grants. Partial humans were also utilized, as in the military, to guide flying drones stationed over rapids.

NOCMACOC had “grudging respect” for whitewater dirtbags and local partial humans who participated in our research efforts and bit the big one. “R.I.P.” and “Don’t let the door hit you on the way out.”

NOCMACOC ANNOUNCEMENT OF EXPERIMENTAL RESULTS: 100% synchronization of the optical guidance and machine propulsion of the watercraft was achieved by a “machine learning” iterative process, augmented with flying “drone” technology.  

Some other NOCMACOC research determinations, put into “plain English”:

• Path is sketched for human loss-of-life to be eliminated by the implementation of AI in canoe/kayak watercraft.  Likewise human injury. 
• AI fearless but not stupid. 
• AI can roll kayaks/canoes 100% success rate.
• After only one hundred iterations: AI surpasses instincts of novice person, 98%. 
• After ten million iterations, in relation to a “whitewater professional”: 
• AI assembles a more totalized view of the physics, fluid dynamics, meteorological anomalies/concerns, water flow rate and terrestrial drainage rates, and percentage probability of predatory organisms/microorganisms.
• AI-guided boats travel 120% faster than human-guided boats during periods of polychromatic sunset and other “pretty” obstacles.
• AI unable to determine whether any given day will be “a good day on the river”. 

Technical paper on this subject is linked here. 

---

David Portz