Forget LLMs, make yourself a mentat
When I was at university, I worked part time tutoring high school maths. We’d regularly get kids who struggled with algebra, things like solving 5x + 7x — 24 = 0.
Bizarrely, taking a step back and encouraging these students to rote-learn basic facts (5 + 7 = 12, 12 x 2 = 24) would help.
Turns out that if these facts are obscure, then you see algebra but through a Casio, darkly. Instead of instantly summing the 5 and the 7 in your head, moving the 24 to the right-hand side, and then dividing, each step in the calculation requires pause, deliberation, and flicking to a external source.
A while ago I visited a Tibetan monastery in South India. The most gifted monks memorise, by heart, a 15th century text called Essence of Eloquence. Being 234 pages of dense philosophy, the endeavour takes one or two years, with many hours every day devoted to recitation of the text, early in the morning for new pages and late at night for revision of the old.
The few who make it receive a prize from the monastery, and knowing the text by heart is said to impart a special blessing that enables these monks to achieve one of their spiritual goals — understanding the nature of reality. This would normally take many lifetimes, but because of the reverence for the author, the saint Tsong Khapa, the achievement is expedited.
The premise for LLM-mania is that we can increase our efficiency by offloading menial tasks to our AI companions. The problem is that efficiency in these menial tasks is the key to our intellectual success.
The best engineers I have met are like the maths students who already know 5 + 7 = 12 when they approach 5x + 7x — 24 = 0. They’re like the mentats from Frank Herbert’s Dune — human calculators with a vast web of knowledge already at their fingertips, immediate and stark without recourse to search engines, books, or other lookup.
One might imagine that this knowledge is a result of deliberate study. Perhaps. But such proficiency is often from solving the same menial problems again and again until solving them is second nature.
Don’t disparage these menial tasks by offloading them to LLMs, because intense familiarity with these is the gateway to lofty achievements of the human intellect. I asked one of the mentat-monks how the blessing worked. They replied, “I have no special insight into the nature of reality. But Tsong Khapa’s voice is so familiar to me now that even if an important passage is obscure for others, it’s not so hard for me.”
