Live data from Hacker News

A.I. and the Future of Cheating

onezero.medium.com

41–50 of 50 posts

Re: A.I. and the Future of Cheating

#41

Imagine this use-case: 1) Copy some text to be plagiarized. 2) Feed text to AI so it's rephrased. 3) Present result as own work.

Interestingly enough, I fed those lines into an AI[1] and it spit out:

1) Start copying some text from another source.

2) Start rephrasing text.

3) Present result as own work.

You are now a master plagiarist.

-----

So I think you're onto something. The AI is a far faster rewriter of text than human beings.

[1] https://talktotransformer.com/

Re: A.I. and the Future of Cheating

#42
post #34

> What happens when universities can’t tell whether an essay is written by a human or an algorithm? Then in that scenario, that means that essays are worthless at evaluating students, and if AI essays really are as good as human essays, then teaching the writing of essays at the college level is a waste of time. However, looking at the generated essay, it seems to me that the real problem is that universities have re…

Computers have been able to do arithmetic faster, better, and cheaper than humans for decades. But we still teach children to do arithmetic, and grade them on how well they do it (almost never as well as a computer). Why? Because doing more useful and creative math is a lot easier if you can do a little arithmetic. The purpose of college essays has never been the quality of the product. The purpose is to have student…

> The purpose is to have students practice so that, when they really do have to convey something in writing, they can convey it well.

Then have a human grade it. Have personalized feedback instead of a cookie-cutter grading that's just another data collection scheme + some basic keyword matching to catch stupid mistakes.

Re: A.I. and the Future of Cheating

#43
post #34

Earlier quoted context omitted.

Computers have been able to do arithmetic faster, better, and cheaper than humans for decades. But we still teach children to do arithmetic, and grade them on how well they do it (almost never as well as a computer). Why? Because doing more useful and creative math is a lot easier if you can do a little arithmetic. The purpose of college essays has never been the quality of the product. The purpose is to have student…

Speaking from experience, the person that writes best code (in terms of both raw quality and readability) also seem to write the best docs. Maybe teaching essays are sideways to efficient communication?

Is it possible you have cause and effect reversed? I received this as "learning to code well can also teach good communication skills"

Re: A.I. and the Future of Cheating

#44
post #34

Earlier quoted context omitted.

Computers have been able to do arithmetic faster, better, and cheaper than humans for decades. But we still teach children to do arithmetic, and grade them on how well they do it (almost never as well as a computer). Why? Because doing more useful and creative math is a lot easier if you can do a little arithmetic. The purpose of college essays has never been the quality of the product. The purpose is to have student…

Speaking from experience, the person that writes best code (in terms of both raw quality and readability) also seem to write the best docs. Maybe teaching essays are sideways to efficient communication?

I think you're right - an essay should communicate a key thought, broken down in logical steps, coherently tied together. This is important for a whole host of skills.

Re: A.I. and the Future of Cheating

#45
post #36

Earlier quoted context omitted.

>>> Unfortunately, a lot of essays in university lecture courses are graded by underpaid TAs who are skimming quickly, and who might be inclined to give an AI-written essay a D or a C- for effort rather than recognizing it as fraud. This raises an interesting point, which is that there seems to be a relationship between essay-writing and essay-reading. The students are able to cheat on essay-writing because the schoo…

>This raises an interesting point, which is that there seems to be a relationship between essay-writing and essay-reading. The students are able to cheat on essay-writing because the school is already cheating on essay-reading. When they pay so little why would a person spend hours reading and analyzing one essay. The bigger issue if the lack of resources devoted to grading and not that the people on the ground are c…

Indeed, I would say the system is cheating.

Re: A.I. and the Future of Cheating

#46
post #16

Highly recommend testing this new cheating tool out by entering a custom prompt at https://talktotransformer.com/ (which uses GPT-2) and reading what the AI generates for your custom essay snippet. A single sentence with a euphemism should suffice... "I" was able to create some hilarious franken-stories. Anyone care to share their greatest hits?

At first glance I was Wow. But then it went downhill very fast.

Prompt Is Capitalism going to win at the end? Answer I don't think so. I think that, in the end, the people have won this battle, the people have won this fight. The people have done everything they could to take back their lives. They are tired, but they still want to have a future. The fact that capitalism has not won is because the people are still fighting.

You can read more on the struggle and the victories here: www.sabr.org

Prompt Is Communism going to win at the end? Answer So, if you've listened to this podcast for a while, you'll know that I'm a huge fan of the book "Against Our Better Judgement". It's a very well-known book which argues that the current financial crisis is due to bad investment decisions made in the last decades by Western nations. These investments were largely based on faulty economic reasoning which has resulted in high levels of unemployment and under-investment.

In this podcast, I'm going to take a look at the book from a slightly different angle, which is why it was written. What I will argue is that it is also very relevant in today's world because the same flawed reasoning is still in place, which is causing the current situation in the first place.

My arguments are going to be based

Prompt Is gravity going to win at the end? Answer We won't know the answer for sure, but we do have a few clues about how the final battle might unfold.

Advertisement

Here's what we know:

Gravity will win at the end: The most obvious thing to do is just to say gravity is going to win. But what does that really mean?

Advertisement

If we assume the universe is a static two-dimensional space and all matter is a singularity-in-a-bottle, then the amount of mass in the universe is a constant.

What will it take for the universe to collapse? One way to put this is that the amount of mass in the universe will double every second, so the universe will expand by about 1.5 billion kilometers in the time it takes you

------- Notice some really weird text like Advertisement, random web site in the reproduced text etc. This makes it look like copying data from various web sites.

Re: A.I. and the Future of Cheating

#47

Imagine this use-case: 1) Copy some text to be plagiarized. 2) Feed text to AI so it's rephrased. 3) Present result as own work.

Interestingly enough, I fed those lines into an AI[1] and it spit out: 1) Start copying some text from another source. 2) Start rephrasing text. 3) Present result as own work. You are now a master plagiarist. ----- So I think you're onto something. The AI is a far faster rewriter of text than human beings. [1] https://talktotransformer.com/

I can see that item 3 is the same. Still very impressive.

Re: A.I. and the Future of Cheating

#48

> What happens when universities can’t tell whether an essay is written by a human or an algorithm? Then in that scenario, that means that essays are worthless at evaluating students, and if AI essays really are as good as human essays, then teaching the writing of essays at the college level is a waste of time. However, looking at the generated essay, it seems to me that the real problem is that universities have re…

We'll have to develop AIs that are trained to detect if the essay was written by a human or by an AI. This could turn into a pretty interesting arms race with surprising privacy considerations. For example, the AI might require writing samples from the student under consideration from K-12 in order to get sufficient accuracy..

In a GANN, they are already trained together.

Re: A.I. and the Future of Cheating

#49

> What happens when universities can’t tell whether an essay is written by a human or an algorithm? Then in that scenario, that means that essays are worthless at evaluating students, and if AI essays really are as good as human essays, then teaching the writing of essays at the college level is a waste of time. However, looking at the generated essay, it seems to me that the real problem is that universities have re…

Then t
Post reply on HN