mynameisvinn124

i understand why we'd like to have near-perfect classification but do we need 100% accuracy in practice?

for example, the paper identifies an instance where a stop sign was misclassified as a speed limit under various conditions.

however, wouldnt applying bayesian priors ("how often do i see speed limits at intersections?") presumably negate misclassifications?

rather than chase the elusive ~100% accuracy from a single model, why not take a layered, ensemble approach?

darpa recently launched a new program to do just that - force ML models to "explain" their predictions. it's called "explainable artificial intelligence" and the goal is to increase trust of autonomous systems.

[1] https://www.darpa.mil/program/explainable-artificial-intelli...

"- You also escape having your income determined by the market for labor, and instead have it determined by the market for your product."

one of the most insightful comment on the concept of value ive read in my 34+ years.

"Many breach announcements this year pointed to a “sophisticated attacker” as a narrative of their issue. This usually is followed up by criticism when an initial means of their compromise is revealed.

Most breaches begin with spear phishing, commodity exploits, a leaked key, or some other obvious or preventable detail. However, this is almost never the “sophisticated” aspect of a breach worth talking about. It’s easy to point at an embarrassing vector and dismiss the rest of an attack. Thus, do not judge an adversary by the vector they’ve chosen.

An adversary may show you what “sophistication” means after advancing from their beachhead."

* https://medium.com/starting-up-security/learning-from-a-year...