3 ms·

>It was deliberately flagged by Google because it was their decision to use an algorithm for flagging to increase profits by having to pay less humans, so it ma

>It was deliberately flagged by Google because it was their decision to use an algorithm for flagging to increase profits by having to pay less humans, so it makes no difference if the actual flag=true operation was performed automatically. I'm quite sure it does from an anti-trust perspective. >I cannot build a bridge and let AI decide how strong it should be and if it crashes, be like „welp sorry it was the AIs fault, we didnt spent on actual engineers to save money, I promise the next bridge will be stronger“. Under some circumstances this will be fine. Using AI doesn't change negligence rules into strict liability rules, it just changes the criteria through which negligence will be determined. It seems likely that most jurors would not be as cantankerous in their outlook towards AI as you are, though.

jlkuester7 5y ago
I am genuinely interested in how this would play out in court. On one hand, sure, intent is important in a lot of situations. It could be challenging to "prove" that Google had "bad" intentions just because there is a disconnect between their documented policy and what their AI actually enforces. That being said, in an anti-trust case it seems like the actual outcome for the consumer is what is most important. Anti-trust laws come into play when monopoly's business practices are not illegal, per se, but ultimately harm the marketplace because of their scale and eventual negative consumer outcomes. Anyway, INAL, so maybe (certainly) I don't have the full picture here....
ineedasername 5y ago
>I'm quite sure it does from an anti-trust perspective. You are incorrect. mistakes, or incompetence, or lack of intent do not excuse behavior that results in harm to consumers or prevents competition, which is generally a significant test for a determination on this sort of thing: In the US in particular, anti-trust law requires courts to evaluate a situation on the basis of 3 benchmarks to determine if one of them is met for an anti-competitive situation: 1) The Per-Se Rule. 2) The Rule of Reason. 3) The Quick Look. Number 3 in particular makes it easy to view this situation as anti competitive because here the single most important question under this for of review is whether or not there is injury in the market. The formal test is will a person "with even a rudimentary understanding of economics could conclude that the arrangements in question have an anticompetitive effect on customers and markets" In this case, a customer cannot use their product of choice and a competitor is prevented from competing. Of course an actual court case could go a variety of different ways, but you say you're quite sure that the anti-trust perspective makes a distinction as the GP comment indicated, you are incorrect.
- yanderekko 5y ago
  In general, I'm quite sure that a broadly-defensible ML model or security rule generating a false positive is not going to support an anti-trust case.

jlkuester7 5y ago

I am genuinely interested in how this would play out in court. On one hand, sure, intent is important in a lot of situations. It could be challenging to "prove" that Google had "bad" intentions just because there is a disconnect between their documented policy and what their AI actually enforces. That being said, in an anti-trust case it seems like the actual outcome for the consumer is what is most important. Anti-trust laws come into play when monopoly's business practices are not illegal, per se, but ultimately harm the marketplace because of their scale and eventual negative consumer outcomes. Anyway, INAL, so maybe (certainly) I don't have the full picture here....

ineedasername 5y ago

>I'm quite sure it does from an anti-trust perspective. You are incorrect. mistakes, or incompetence, or lack of intent do not excuse behavior that results in harm to consumers or prevents competition, which is generally a significant test for a determination on this sort of thing: In the US in particular, anti-trust law requires courts to evaluate a situation on the basis of 3 benchmarks to determine if one of them is met for an anti-competitive situation: 1) The Per-Se Rule. 2) The Rule of Reason. 3) The Quick Look. Number 3 in particular makes it easy to view this situation as anti competitive because here the single most important question under this for of review is whether or not there is injury in the market. The formal test is will a person "with even a rudimentary understanding of economics could conclude that the arrangements in question have an anticompetitive effect on customers and markets" In this case, a customer cannot use their product of choice and a competitor is prevented from competing. Of course an actual court case could go a variety of different ways, but you say you're quite sure that the anti-trust perspective makes a distinction as the GP comment indicated, you are incorrect.

yanderekko 5y ago
In general, I'm quite sure that a broadly-defensible ML model or security rule generating a false positive is not going to support an anti-trust case.