076 - Parallels

What do typical AI technologies and migraines have in common?

Though the question may sound like the premise for a joke, the parallel struck me as I awoke this morning. In terms of my personal experience, the two share a virtually identical pattern.

At age 6 I was diagnosed with chronic migraines, and over all of the school-age years that followed I was reliably home from school with a migraine around half of the time. I saw doctors who were allegedly the best neurologists and neurosurgeons in the world, or at least in the US. Beyond very briefly effective and grossly overpriced treatment methods, they were utterly useless at addressing the problem.

For the 20 years following that diagnosis at age 6, the medical industry wasted my time, at great expense to my family. I too once held doctors in some esteem, that default expectation of cognitive bias that says they must know something of merit that I do not. Once I shed that naïve belief the problem was trivial to overcome.

Having done my own research in the neuroscience domain, learning as an autodidact, I purchased 20 different laboratory purity compounds and began experiments to adjust my neurochemistry to prevent migraines. Within 3 months of applying the scientific method, I'd discovered that 300 mg of anhydrous caffeine paired with 300 mg of L-theanine reliably reduced my migraine frequency by roughly 90%. The effects of anhydrous caffeine were also very distinctly different than caffeine from traditional sources like coffee and energy drinks, lasting up to 3 times longer and with no "crash".

Those doctored, regarded as the best, were utterly worthless. The problem was outside of their depth, as in truth they had no expertise for it, and the solution used methods outside of the scope of their consideration. I haven't seen a doctor in 8 years now, and I further reduced my migraine frequency to about 1 or 2 per year, another 90% reduction on top of the first.

Those two pure powders, which I've used every morning for more than 10 years now, are dirt cheap, costing me roughly $60 per year. That couldn't even get you 10 Venti beverages from Starbucks. In contrast, over those 20 years, doctors easily wasted over $100,000 of family funds, not even counting what insurance covered. That is conservatively a 50-fold increase in cost, with no meaningful benefit.

That pattern is also precisely what we see in AI today. We're overflowing with "AI Experts" who are utterly worthless in practice, as they lack useful expertise, and the solutions fall outside of the scope they work within. The predictable result is that they produce systems that fail to offer any meaningful benefit, and easily cost 50x or more to run relative to competently designed systems.

Today's typical AI is enough to give anyone a headache, but effective solutions can be achieved at a relatively trivial cost, and within short periods of time.