Comment by karmacondon | Hacker News Reader

Comment by karmacondon | Hacker News Reader

Parent Full thread

As someone pointed out in another thread, we're ruled by algorithms now and have been for quite some time. Tinder itself is based on a human processed algorithm:

  IF ((I think that he/she is hot) && (he/she thinks I am hot)) 
  THEN 
  date();

The good thing about computer algorithms is that we can open them up and see how they work. Like, y'know, hackers. This program is doing the same basic task as a person would, "Look at a face and judge hotness", except it can be tuned and tweaked and quantifiably understood. I don't understand what's dystopian about that.

I do understand why people don't like it. Everyone feels that an algorithm could screw them over in a situation where a person wouldn't. It's the driverless car argument. And we all know the response, that humans aren't perfect either and we have better metrics and tools for understanding and fixing machine mistakes than we do human mistakes. The same is true for algorithms that control Tinder, credit applications, health insurance rates, et al. They aren't always right and it does suck when you can't make an appeal to human compassion. But humans can be just as wrong and just as obstinate, if not moreso. At least with an algorithm you know where you stand.

I, for one, welcome our algorithmic overlords. As a hacker I've always been able to figure out how technical systems work. Other humans? Not so much. If anything, this new world will play to my strengths and to the strengths of those like me.

> I, for one, welcome our algorithmic overlords.

Just imagine a world where you'd never have to hear "because I said so" and "becasue fuck you, that's why" ever again. A world where for everything that goes wrong, there'll always be a clear way to fix it and some clearly identifiable someone liable for the mistake. A world where being handsome is no more an advantage than having blue as your favorite color. Etc.

Maybe it's not perfect, but it would most certainly take away power from people that don't deserve it. I can see why the "powers that be" would try to convince us that this is "dystopian".

Until computers are as generally intelligent as humans, the algorithms aren't going to be the overlords -- they're going to have overlords. The algorithm isn't processing your loan application because it runs a bank; it's just doing what the bank's owners paid for it to do, and if they want to say, "because fuck you, that's why", they can happily continue to do so. If anything, better AI may make them better able to leverage their existing advantages over the rest of us.

And when the AI does reach human level (a long time from now), we may be no more capable of understanding its motivations than we are of understanding those of other humans today.

> if they want to say, "because fuck you, that's why", they can happily continue to do so

Not if there are incentives for the algorithms to be subject to public scrutiny.

> And when the AI does reach human level (a long time from now), we may be no more capable of understanding its motivations

(1) You're implying they/it'll have "motivations" in the first place. My whole argument was about a lack of motives.

(2) There's a difference between understanding how a conclusion was reached and why it was reached. Example: You may not understand how your compiler decided that line X in your program is dead code, but you may still be capable of comprehending that it in fact is dead code.

> Not if there are incentives for the algorithms to be subject to public scrutiny.

The problem is that we don't have that transparency now, and aren't particularly likely to in the future. [1] Particularly for software-as-a-service: I can't audit Facebook's algorithms, despite their effects on what my friends and family see and discuss. And I can't easily audit the black box models used to make public policy, despite the value of transparency there, either.

[1] http://www.theatlantic.com/technology/archive/2015/01/the-ca...

Also, it's far from clear that transparency into an AI's algorithms buys you very much. A mathematically inclined college freshman has no trouble understanding how to train a deep neural network today, but it's at the cutting edge of research to understand what function a trained network is actually computing.

Scale that up in complexity a few orders of magnitude, and the thing may as well be a human. That is, you'll likely have broad understanding of its "motives", but we have that for humans today and it doesn't help that much. I understand Facebook wants to sell advertising, but that transparency isn't sufficient for understanding why a particular message is delivered to your followers while another isn't.

Algorithmic middle managers then?

That would be what Uber's dispatcher system is.

A lot of the advanced applications of algorithms are using deep neural networks. Unfortunately, it's notoriously difficult to figure out specifically why a neural network did what it did. You just end up with "because the algorithm said so, that's why."

My experience with algorithmization so far is just the opposite. 20 years ago the bank manager would tell you "no, I'm not willing to offer you that mortgage because the house is overpriced / I called your manager and they said you were flaky / your hair was scruffy when you came in"; of course they could lie or refuse to give you a reason, but most of the time you got some explanation. Nowadays you just get "computer says no".

Looks like the bank need a random excuse generating algorithm.

> Just imagine a world where you'd never have to hear "because I said so" and "becasue fuck you, that's why" ever again

Instead you'll hear "sorry, but it's the algorithm. I understand you need your paycheck, but there's nothing I can do and the programmer won't be available until the middle the of next week".

Gahh, close your if statement man.

END IF;

Don't worry, he was using C and he defined THEN as an empty macro statement.

Maybe you want to look into the 'Psycho Pass' series for some illustration of the risks.

Don't use the brains of criminals in your computer. Problem solved.

That's nowhere near the only problem with Psycho-pass's society.

Then we can read some Heinlein to learn why libertarianism is the obvious correct answer to every problem, as well as the valuable insight that the only thing on the minds of attractive young women is sex with old men.

I learn so much from fiction.

Reply on news.ycombinator.com