183 - Credibility Paradox

Empirical evidence offers value that no abundance of theories can deliver on, and the evidence is at least as likely to prove counterintuitive, as the common theories are to prove BS. One striking example of this occurred to me this morning, looking back over a collection of personal experiences living in a US tech hub, and considering what has changed in the years since.

While in Seattle I had a collection of experiences orbiting a central theme:

  • An AWS Director chased me to the door and kept me there for ~10 minutes following me asking a single question that caught his interest, with him arranging an interview promptly thereafter.

  • A 25-year recruiting veteran informed me that he was sure I'd become a millionaire within 5 minutes of meeting me.

  • Former NSA engineers stalked me for some time after I gave a presentation for a novel line of technology.

  • ...and various other extremely low-probability reactions and events.

This illustrates the past, and begs two questions: "What has changed in the past 5+ years?" and "What is the current reaction following such changes?".

Answering the first question:

  • The most advanced and genuinely "general" AI system by far was publicly demonstrated for several years.

  • The hardest version of the Alignment Problem was solved.

  • The first cognitive bias detection system exceeded human performance.

  • Numerous research papers were published, some in high-tier journals.

  • I became one of only two people to have familiarity with the codebase of our systems and have worked with them extensively in practice.

  • ...and more similar points.

The naïve expectation would be that many positive and noteworthy changes should produce a positive response relative to the historical responses. However, this has observably not been the case:

  • Attempts by other parties to recruit me dropped to zero as those positive and noteworthy changes occurred. Strong inverse correlation.

  • At the same time, how most people value and follow others has dramatically changed over the past few years as well, with the most obvious frauds becoming the most highly valued and influential by naïve public opinion. Extremely strong inverse correlations with credibility.

Any variety of theories may be proposed to explain such divergence, but the empirical evidence remains regardless. With the average recruiter now observably holding a value at or below zero, and the average value estimator or layman now favoring disinformation, most of the mechanisms by which a terminal crash might be avoided have been disabled.

The inmates are running the Asylum, and their theories are every bit as delusional as one might expect.