3 ms·
I am tempted to paraphrase the Elephant man, "I am not a sigmoid function, I am a human being!" in response to that headline :-) That said, I think it is great
by ChuckMcM 8y ago
I am tempted to paraphrase the Elephant man, "I am not a sigmoid function, I am a human being!" in response to that headline :-)
That said, I think it is great that they are trying to characterize this sort of thing. The second law of institutions is that every process must require more people to execute than can conspire without being detected[1]. But relying so much on humans in one way or another is going to gum it up. My mom for example, a certified human, would always flag anything that offended her belief system as 'fake news.' Fact checking is fine but it is important to understand that conspiracy folks actually believe what they are saying is true, regardless of any facts presented that contradict their world view. I have always attributed that to an internally rooted desire to "need" it to be true, either for their own self image or their own conscience.
[1] No, I just made that up. But the kernel is true, if you have a system that is being gamed you have to build a defense that cannot itself be gamed.
- move-on-by 8y ago> if you have a system that is being gamed you have to build a defense that cannot itself be gamed. I guess, except without transparency and accountability, this will give more power to those gaming the system
- panarky 8y ago> every process must require more people to execute than can conspire without being detected Step 1 - let users report posts that are false Step 2 - human fact checkers investigate user reports Step 3 - rely more on users who report posts as false that really are false, and less on users who report posts as false that really are true Step 4 - repeat step 3 but for the fact checkers themselves Step 5 - train an ML model to automate more of the fact checking work
- degenerate 8y agoStep 6 - abandon the project when an edge case creates a PR nightmare
- antidesitter 8y agoThis reminds me of a generative adversarial network.
- drewmol 8y agoStep x - sell ability to target users based on gullibly and trustworthness scores
- threatworking 8y agothe second law of institutions you say? can we get a source
- trobertson 8y ago> The second law of institutions is that every process must require more people to execute than can conspire without being detected[1] > [1] No, I just made that up. You didn't, actually. This is the Byzantine Generals problems as described by Leslie Lamport back in 1982. It is huge in fault tolerance for distributed systems. https://en.wikipedia.org/wiki/Byzantine_fault_tolerance#Byzantine_Generals'_Problem https://en.wikipedia.org/wiki/Byzantine_fault_tolerance#Byza...
- ChuckMcM 8y agoYou are correct, the origin of that principal is in Lamport's paper (which is one of my favorites by the way). Where I was coming from was that there aren't, as far as I know, a codified set of rules for creating a stable institution. I was thinking that someone would say "there is no such thing" and ignore the advice so in my footnote I pointed out the the basic idea was in fact quite sound. I probably should have referenced the paper as well but assumed most people would already know about it.