Live data from Hacker News

Human in A.I. Loop

sanealytics.com

1–10 of 26 posts

Re: Human in A.I. Loop

#5
post #2

It seems that the name of the concept this guy is looking for is: Mechanical Turk service.

Not sure if MT works when some specific knowledge is needed (I think their employees know or have quick ways to find out the correct matching dress given a photo)

Re: Human in A.I. Loop

#6
"In the loop" is a reference to the OODA loop, which is the decision cycle of observe, orient, decide, and act, developed by military strategist and United States Air Force Colonel John Boyd. "In the loop" has a specific tactical meaning, which is that you're able to do a full OODA loop cycle faster than a comparable AI/man system which gives a strategic advantage and maybe used against oppositional forces in a variety of ways; most interesting is applying it to PsychOps.

"Man in the loop" reference can be found in use by the US AirForce going back to the 80s for weapon systems design; specifically between fighter jet systems and pilots. I've heard that US AirForce used brain implants as early as the 90s in fighter pilots.

Here's an example of the US Army reads soldier's brain waves to speed up image analysis:

http://newatlas.com/us-army-eeg-brainwave-image-analysis/403...

Re: Human in A.I. Loop

#7
Does anyone know any good tools to make that sort of page quickly in Python? I mean, I could cobble together a web app, but I feel like dataset annotation is so common that there should be something out there already.

Re: Human in A.I. Loop

#8
post #2

It seems that the name of the concept this guy is looking for is: Mechanical Turk service.

The concept is actually more commonly called "human computation" (which includes stuff like digitizing books using CAPTCHAs).

It has academic journals/conferences under that name: http://hcjournal.org/ojs/index.php?journal=jhc

Re: Human in A.I. Loop

#9
post #7

Does anyone know any good tools to make that sort of page quickly in Python? I mean, I could cobble together a web app, but I feel like dataset annotation is so common that there should be something out there already.

Why Python? Why not just focus on finding the best open source data annotation framework?

(Love Python, but normally don't bias my selection of a solution by the language it's written in.)

Re: Human in A.I. Loop

#10
"What we need to build is a gamified quick truther thingy that all our 200 employees in main office can log in during their lunch break and rate a few items."

Lunch break? Or you could just, I don't know... go with me here, I know it sounds crazy at first..., pay them for providing this value to your company.

Post reply on HN