Human-in-the-loop computing is the future | Hacker News Reader

Human-in-the-loop computing is the future

computerworld.com

48 pointsyzmtf200823 comments

1950 - work is 9-5, but your free time will soon be liberated by machines!

2015 - work is 8-6, but machine processing time will soon be liberated by humans!

2016 - we cached your workflows and did enough analysis on it to extrapolate functionality, but best of luck in convincing your sociopolitical handlers of your relevancy

Automate an employment agency, the Human-Machine Interaction equivalent of dividing by zero.

I dunno, plenty of tech recruiters can be automated:

https://news.ycombinator.com/item?id=10399680

I don't think that is an accurate characterization. A time use study by the Federal Reserve Bank of Boston[1] found that the time a male worker spends at the workplace decreased from 51 hours in 1965 to 40 hours in 2003. The average male in 2003 spent most of that extra time on leisure activities. That's only one study, of course, but it squares with what I have consistently heard. Another study, which I can't find right now unfortunately, looked at leisure time enjoyed at the workplace. That would be a contradictory concept in 1965, but today, many people spend an hour per day or more reading, playing games or visiting social media sites.

[1] http://www.bostonfed.org/economic/wp/wp2006/wp0602.pdf

You can come to all kinds of satisfying conclusions, when you ignore the fact that a lot of the productivity gains of the developed world comes of the backs of dirt cheap labor in the under developed world.

It's a safe bet that, for the most part, the human in the loop computing is going to be built on the cheapest humans available.

These ideas are not quite "emerging", but now is certainly a great time to implement them! Consider this quote:

"Man-computer symbiosis is a subclass of man-machine systems. There are many man-machine systems. At present, however, there are no man-computer symbioses. The purposes of this paper are to present the concept and, hopefully, to foster the development of man-computer symbiosis by analyzing some problems of interaction between men and computing machines, calling attention to applicable principles of man-machine engineering, and pointing out a few questions to which research answers are needed. The hope is that, in not too many years, human brains and computing machines will be coupled together very tightly, and that the resulting partnership will think as no human brain has ever thought and process data in a way not approached by the information-handling machines we know today."

Licklider, J.C.R., "Man-Computer Symbiosis", IRE Transactions on Human Factors in Electronics, vol. HFE-1, 4-11, Mar 1960. http://groups.csail.mit.edu/medg/people/psz/Licklider.html

" As the CEO of CrowdFlower, I’ve..."

It's supervised learning with the supervision coming from something like Mechanical Turk.

I see it as it always "human-in-the-loop" since the beginning of computing. Human using spreadsheet to do accounting is a human-in-the-loop operation with human deciding what to add, and computer doing the adding. The key difference with AI is that it's now more difficult for the human to understand what the "smart" computer is doing and how to cooperate with it. For e.g., how do the self-driving take over control from and hand over control to human/ How to visualise and give directions to complex algorithms / how gain business insight from data.

This is how I imagine how Google can bring it's self driving Uber-like taxi service to market. They can have a bunch of "remote drivers" that can take over the cars if the machine is not able to decide what to do. Anything 100% automatic is pretty much impossible, there is always an unsolvable or never seen before problem on the street.

No mention of the IP value/licensing of these human signals, sometimes unwittingly donated?

Shush now. Intellectual property law is for investors, not workers.

We are a social animal. We care what others think about us.

But we don't care what computers think.

So there is also a role for humans in mentoring systems. Software can remind us and track whether we've met our objectives. It can do the boring, repetitive work.

But the interactions really only mean something if your mentor is in the loop. He or she might be asking you a question. Or maybe its the software, with the mentor having only approved the question and seeing the response.

This blended system could greatly reduce the effort involved in mentoring someone. And if everybody knows its the software asking repetitive questions (Is this done yet?), the mentor doesn't come off as a nag either.

I am not convinced we don't care what computers think, or will care in the near future. I don't see any innate ability for humans to distinguish a decent AI-like opinion from human opinion.

So when a computer tells you that you made a mistake, would you get all embarrassed and apologize to it?

I prefer to think of it as intelligently precomputing options for a person, or setting an automatic helper to solve most of the cases.

I read the title as "Congratulations, you're now a x87."

> While the concept of chess being “solved” is still considered rather remote (there are at least 1043 board positions to account for)

What does this 1043 refer to?

Typo for 10^43, a rough estimate of the number of possible chess positions, originally due to Shannon [1].

[1] http://www.pi.infn.it/~carosi/chess/shannon.txt

This reminds me.

"Luke, you switched off your targeting computer! What's wrong?"

"Nothing. I'm all right."

Which reminds me of:

Loke Groundrunner: I'm going to trust my feelings and use the power of the thumb.

Voice of Oobedoob Benubi: Use the instrument panel, Loke.

Loke Groundrunner: What?

Voice of Oobedoob Benubi: The instrument panel. That's what it's there for. Advanced weaponry designed to hit tiny targets.

Loke Groundrunner: Ok, ok.

Which reminds me of (emphasis mine):

Gunnery Chief: This, recruits, is a 20-kilo ferrous slug. Feel the weight. Every five seconds, the main gun of an Everest-class dreadnought accelerates one to 1.3 percent of light speed. It impacts with the force of a 38-kilotomb bomb. That is three times the yield of the city buster dropped on Hiroshima back on Earth. That means Sir Isaac Newton is the deadliest son-of-a-bitch in space. Now! Serviceman Burnside! What is Newton's First Law?

First Recruit: Sir! A object in motion stays in motion, sir!

Gunnery Chief: No credit for partial answers, maggot!

First Recruit: Sir! Unless acted on by an outside force, sir!

Gunnery Chief: Damn straight! I dare to assume you ignorant jackasses know that space is empty. Once you fire this husk of metal, it keeps going till it hits something. That can be a ship, or the planet behind that ship. It might go off into deep space and hit somebody else in ten thousand years. If you pull the trigger on this, you're ruining someone's day, somewhere and sometime. That is why you check your damn targets! That is why you wait for the computer to give you a damn firing solution! That is why, Serviceman Chung, we do not "eyeball it!" This is a weapon of mass destruction. You are not a cowboy shooting from the hip!

Second Recruit: Sir, yes sir!

https://www.youtube.com/watch?v=hLpgxry542M

But the physics of that is entirely wrong. See Olbers' paradox. Eyeballing it is fine; if you miss you will almost certainly never hit anything - it was considered fine to eyeball AA in WWII, and the odds of hitting a friendly fighter there were worse; also there are already plenty of big rocks going fast in space.

If you want to be charitable towards it you can assume it was meant to scare those recruits into submissions so they won't waste precious ammo in deep space battles. Also, the explanation holds in orbital fights when you're shooting downwards (wrt. closest planetary body).

Reply on news.ycombinator.com