140 - Perspective

Recently a friend recommended me to a book called "Competing Against Luck", which surprisingly I found makes the case for a perspective I've personally held for more than a decade, but which I've only rarely considered asking of others. The question is "What did someone (hire) a product or service to do for them?"

This question is a matter of motivation and function, and rather than focusing on frequently trivial data and spurious correlations, like demographics, it digs into causality. The book illustrates how the causality of the job to be done and the choice to select that product often strongly and sharply diverge from marketer expectations, and often in multiple directions. This is partly because they gather the wrong data, and they often focus on categorizing people and then painting the picture of an "average" person in each arbitrary category.

The quote from Alan Kay "Point of view is worth 80 IQ points.", which was evidently later reiterated by Jeff Bezos, as the book notes, is on point for this. It can also be formulated in reverse, where "Cognitive Bias is worth -80 IQ points." tells the same story from the opposite perspective. While this figure may not be literal or precise, it is a useful tool to keep in mind.

For example, remember that an "Intellectual disability" is any IQ under 70, so if your executive team is leaning into cognitive biases and has a base IQ of less than 150, they are effectively intellectually disabled. People can only be as competent as their biases permit.

Perspective is also a capacity that narrow AI tools like LLMs and RL fundamentally can never have, as they weren't designed for it, and can't process or store anything like it. Even so, the "job" that humans have infamously taken to "hiring" them for today is that of "divination", serving as an "oracle".

The job description for an "oracle", across human history, has been something like "a person who is consulted, offering answers with intelligence and wisdom greater than that of humans". This is obviously something that LLMs can't deliver, though they are optimal Bull$hitters, which they share with most historical examples of "oracles".

This means that there is a job that humans want, and have wanted for millennia, that is poorly served. Ironically, the technology to serve that job exists and has been demonstrated, but humans have thus far lacked the wisdom and/or competence to invest in it. It is actually very easy to exceed human knowledge and wisdom with digital, scalable, iteratively self-improving, and human-like intelligence. What is extremely difficult is for people to wrap their heads around that.