2 ms·

After skimming through the paper I can’t immediately pick out the data that says how much more certainty there is for a given text to detect a watermark, and th

by awongh 2y ago

After skimming through the paper I can’t immediately pick out the data that says how much more certainty there is for a given text to detect a watermark, and the graph of that certainty as the text size grows. (They seem to assert that the certainty grows as token count goes up, but it’s not clear by how much.) I worry (and have already read worrying things) about “cheating detection” tools that have been deployed in schools. My intuition would be that there’s just too much entropy between something like an essay prompt and the essay itself. I guess it also depends on how specific the teacher’s essay prompt is as well.

eitland 2y ago
> I worry (and have already read worrying things) about “cheating detection” tools that have been deployed in schools. This is my worry as well. Punishment for cheating can be can easily set back a student a year or more. This is fair if the student has been cheating, but really harsh. So while this isn't criminal court, I think schools should apply the same principles here: innocent until proven guilty. And in my view, secret probability distributions isn't exactly good proof. Furthermore, to make it even worse: if someone is actually innocent it will be next to impossible to argue their innocence since everyone will trust the system and as far as I can see the system cannot be actually verified by a board without disclosing the weights. And that is assuming they would care to try to help a student prove their innocence in the first place. AFAIK this is a topic that has been explored to some depth in science fiction, but more importantly, we have case like the mail service in UK where multiple people lost their jobs because nobody could belive the system they had built or paid for could make such crazy mistakes. Back to students: For a less privileged student I guess it can easily ruin their studies. TBH as someone who struggled a lot in school I am not sure I'd finished if I had gotten my studies delayed by a year. Which would have been sad, given how well I have managed once I didn't have to juggle full time studies and part time work. Recently (last year and this) we (Norway) have had some debates that seemed to be way overdue regarding what can ve considered cheating (with some ridiculous examples of students getting punished for "self-plagiarism" for the most absurd things, including not specifying a failed previous exam written by themselves as a source). This could easily have gotten nowhere except for the fact that: 1. the person in charge of the board of appeals was caught for something else 2. Somebody took the effort to dig out the master thesis from two ministers, including the then sitting Minister of Education and proved that they had clearly been "cheating" according to the rules that they were judging students by.

by awongh 2y ago

eitland 2y ago

> I worry (and have already read worrying things) about “cheating detection” tools that have been deployed in schools. This is my worry as well. Punishment for cheating can be can easily set back a student a year or more. This is fair if the student has been cheating, but really harsh. So while this isn't criminal court, I think schools should apply the same principles here: innocent until proven guilty. And in my view, secret probability distributions isn't exactly good proof. Furthermore, to make it even worse: if someone is actually innocent it will be next to impossible to argue their innocence since everyone will trust the system and as far as I can see the system cannot be actually verified by a board without disclosing the weights. And that is assuming they would care to try to help a student prove their innocence in the first place. AFAIK this is a topic that has been explored to some depth in science fiction, but more importantly, we have case like the mail service in UK where multiple people lost their jobs because nobody could belive the system they had built or paid for could make such crazy mistakes. Back to students: For a less privileged student I guess it can easily ruin their studies. TBH as someone who struggled a lot in school I am not sure I'd finished if I had gotten my studies delayed by a year. Which would have been sad, given how well I have managed once I didn't have to juggle full time studies and part time work. Recently (last year and this) we (Norway) have had some debates that seemed to be way overdue regarding what can ve considered cheating (with some ridiculous examples of students getting punished for "self-plagiarism" for the most absurd things, including not specifying a failed previous exam written by themselves as a source). This could easily have gotten nowhere except for the fact that: 1. the person in charge of the board of appeals was caught for something else 2. Somebody took the effort to dig out the master thesis from two ministers, including the then sitting Minister of Education and proved that they had clearly been "cheating" according to the rules that they were judging students by.