139 - Paradox of Expertise

An expert with 20+ years of experience in ML and neural networks is called a "historian".

Just as a Proctologist doesn't specialize in ways to treat your headache, shy of "Cranial-rectal inversion", the 20+ year veterans of narrow AI have no inherent expertise when it comes to creating "general" intelligence, or human-like systems. If anything, they have far more cognitive bias baked into their thinking, oriented toward building the wrong type of systems, with all of the decades of heavy-handed Anthropomorphism and delusions that have been woven into even the core terminology, like "neural network" and "artificial intelligence".

Both the Hype Cycle curve and the Dunning-Kruger effect more broadly play a role in this when observed at scale. The Hype cycle represents a narrative's story arc iterating across society, and anything that runs against that narrative faces backlash. The Dunning-Kruger effect more broadly is the backbone of that hype cycle, which the narrative utilizes as a vehicle.

One common form of backlash from introducing a hype-free or anti-hype element into this narrative flow is a kind of Substitution Bias. For example, a narrow AI expert may be substituted for an expert on "general" or "human-like" intelligence. Although the two have very little in common, the substituted expert will often act out their assigned role in the narrative, mentally placing themselves at the expert level on the Dunning-Kruger effect curve, when in fact they're at the opposite end.

I've watched this play out time and time again, as people with 20+ years of experience in distinctly narrow systems statistically have far greater difficulty grasping new concepts as they relate to general and human-like intelligence than those who understand what they do not know. Many critical breakthroughs in understanding human intelligence are less than 20 years old, and most of those breakthroughs have gone unnoticed by those focused on narrow AI systems.

Unfortunately for society, many of the current tech billionaires turned investors fall into this category, consistently failing to grasp their own substitution of one type of expertise for another. Instead, they consistently fall for the adversarially optimized narratives of those selling snake oil, as the snake oil seller has a strategic advantage, in that they can optimize anything, absent any facts to tie them down.

How much of what you do not know is visible to you, and how do you substitute expertise?