4 ms·

> Machines aren’t going away, but they are learning from us - both our lies and our truths. Respectfully, I disagree. Programmers are responsible. Machines on

> Machines aren’t going away, but they are learning from us - both our lies and our truths. Respectfully, I disagree. Programmers are responsible. Machines only follow the instructions humans tell it to follow. Everyone who codes should really look at themselves in the mirror now and then.

notacoward 7y ago
> Machines only follow the instructions humans tell it to follow. I suggest that you read something about AI and ML. The "instructions" that go into these pipelines are many levels removed from the outputs. It's like saying that cells only follow the DNA instructions given to them - ignoring evolution, mutagens, exogenetics, etc. The original instructions don't tell anywhere near the whole story. Is the opacity of these algorithms a problem? Yes, it surely is, precisely because the machines are not just following instructions in any meaningful sense.
- jazzyjackson 7y ago
  Just because the instructions are convoluted and implemented by people who don't code the math themselves doesn't mean the outcomes are removed from the inputs. It's still a machine. Heuristics use random number generators and produce random (non-deterministic) results, is that not following instructions written by programmers?
  - JohnFen 7y ago
    > Heuristics use random number generators and produce random (non-deterministic) results, is that not following instructions written by programmers? Yes, on one level. However, the programmers are not determining what those numbers are, so on another level, no.
  - notacoward 7y ago
    > is that not following instructions written by programmers Not in any useful sense. Everything's deterministic with perfect knowledge, but nobody has perfect knowledge. The universe is still a machine too, and yet somehow manages to surprise us anyway. As soon as you inject non-replicable random numbers into the process, "following instructions" isn't entirely true any more.
- AnimalMuppet 7y ago
  Even if what you say is true, it seems irrelevant to the topic at hand. AI and ML are not (yet) creating false reviews on Amazon.
  - notacoward 7y ago
    How do you know that? Also, even if AI/ML are not being used to generate fake reviews, they sure as heck are being used to determine the search ranking of products stuffed with fake reviews. Amazon has created a game for less-scrupulous sellers to play, and even they don't know the rules.
    - AnimalMuppet 7y ago
      > How do you know that? I admit that I don't. I am guessing that such things are at present beyond AI/ML. An AI/ML-generated review I would expect to read like... well, like obviously-machine-generated text. > Also, even if AI/ML are not being used to generate fake reviews, they sure as heck are being used to determine the search ranking of products stuffed with fake reviews. How do you know that? They have a ranking algorithm, sure, but why do you think it's AI/ML (by normal definitions of AI/ML)?
- salawat 7y ago
  Humans are still the prime movers who cultivate the parameters and models on which the model is trained, and more importantly decide whether or not the results are credible. Thinking like yours is called Mathwashing; and it's a serious problem if we as a society are willing to accept it as a valid justification for dismissing culpability for our choices/actions.
  - notacoward 7y ago
    Before flinging labels, you should have read the part where I said the opacity of the algorithms is a problem. The opacity/non-determinism definitely does not wash away responsibility, and in no way did I imply that it does. The problem is that now people end up being responsible for something they don't fully control or understand, and that's a sure recipe for disaster. Promoting further misunderstanding by putting your own words in others' mouths is not the solution.

notacoward 7y ago

> Machines only follow the instructions humans tell it to follow. I suggest that you read something about AI and ML. The "instructions" that go into these pipelines are many levels removed from the outputs. It's like saying that cells only follow the DNA instructions given to them - ignoring evolution, mutagens, exogenetics, etc. The original instructions don't tell anywhere near the whole story. Is the opacity of these algorithms a problem? Yes, it surely is, precisely because the machines are not just following instructions in any meaningful sense.

jazzyjackson 7y ago
Just because the instructions are convoluted and implemented by people who don't code the math themselves doesn't mean the outcomes are removed from the inputs. It's still a machine. Heuristics use random number generators and produce random (non-deterministic) results, is that not following instructions written by programmers?
- JohnFen 7y ago
  > Heuristics use random number generators and produce random (non-deterministic) results, is that not following instructions written by programmers? Yes, on one level. However, the programmers are not determining what those numbers are, so on another level, no.
- notacoward 7y ago
  > is that not following instructions written by programmers Not in any useful sense. Everything's deterministic with perfect knowledge, but nobody has perfect knowledge. The universe is still a machine too, and yet somehow manages to surprise us anyway. As soon as you inject non-replicable random numbers into the process, "following instructions" isn't entirely true any more.
AnimalMuppet 7y ago
Even if what you say is true, it seems irrelevant to the topic at hand. AI and ML are not (yet) creating false reviews on Amazon.
- notacoward 7y ago
  How do you know that? Also, even if AI/ML are not being used to generate fake reviews, they sure as heck are being used to determine the search ranking of products stuffed with fake reviews. Amazon has created a game for less-scrupulous sellers to play, and even they don't know the rules.
  - AnimalMuppet 7y ago
    > How do you know that? I admit that I don't. I am guessing that such things are at present beyond AI/ML. An AI/ML-generated review I would expect to read like... well, like obviously-machine-generated text. > Also, even if AI/ML are not being used to generate fake reviews, they sure as heck are being used to determine the search ranking of products stuffed with fake reviews. How do you know that? They have a ranking algorithm, sure, but why do you think it's AI/ML (by normal definitions of AI/ML)?
salawat 7y ago
Humans are still the prime movers who cultivate the parameters and models on which the model is trained, and more importantly decide whether or not the results are credible. Thinking like yours is called Mathwashing; and it's a serious problem if we as a society are willing to accept it as a valid justification for dismissing culpability for our choices/actions.
- notacoward 7y ago
  Before flinging labels, you should have read the part where I said the opacity of the algorithms is a problem. The opacity/non-determinism definitely does not wash away responsibility, and in no way did I imply that it does. The problem is that now people end up being responsible for something they don't fully control or understand, and that's a sure recipe for disaster. Promoting further misunderstanding by putting your own words in others' mouths is not the solution.