« first day (951 days earlier)      last day (2172 days later) » 

DukeZhou
DukeZhou
00:02
@nbro also a good Chinese Room link at Stanford: https://plato.stanford.edu/entries/chinese-room/

Still, I think your instinct to ask is correct. We need to build this Stack, and the one thing we have going for us is the high search value of SE in general. A question with a simple title, designed for search, and good answers, breaking down the links, would surely benefit SE:AI.

Ideally one of the answers would be suitable for the general audience, avoiding technical language, similar to how shows like Nova do it.
DukeZhou
DukeZhou
Looks like algorithmic error may be the cause of the 737 crashes: forbes.com/sites/taylorarmerding/2018/11/20/…
 
12 hours later…
DuttaA
DuttaA
11:41
@DukeZhou sorry to say this but the article seemed strangely incomplete to draw any proper conclusion. Was the sensors at fault or was the software at fault? What malfunctioned?
The article doesn't seem rigorous.
I mean what type of software cannot be rigorously tested?
AFAIK, checking boundary conditions should be enough. What do you think?
 
8 hours later…
DukeZhou
DukeZhou
19:33
@DuttaA I suspect we won't have full details for quite some time, but it does appear that the autopilot (here an assistance function to keep the nose from elevating too much because of the engine placement) drove the planes into the ground.

Not sure about boundary conditions.
nbro
nbro
19:50
psychologytoday.com/us/blog/the-future-brain/201903/…
DuttaA
DuttaA
20:38
@DukeZhou by boundary conditions I meant the conditions we Generally check like 0, -ve, overflow, underflow, etc. I think this should be enough for software, although for VLSI chips all input sequences should be tested which is impossible so they use advanced methods to check the working all electronic components.
DukeZhou
DukeZhou
20:59
@DuttaA yeah, I'm wondering if "common sense" might have been called for, like a pilot knowing there's no way they have sufficient altitude to drive the plane down. I'll definitely be following this story closely.
 
3 hours later…
nbro
nbro
23:32
A nice article that explains MCTS applied to the RL context: https://towardsdatascience.com/monte-carlo-tree-search-in-reinforcement-learnin‌​g-b97d3e743d0f.

« first day (951 days earlier)      last day (2172 days later) »