AI essays - holy moly

[Anonymous]

Anonymous wrote:

Anonymous wrote:I don't know much about AI and my impressions and thoughts may be be wrong. However, I do casually follow developments from a birds-eye view. With, this a few points:

A key problem with AI detectors are lots of false positives, i.e., your DC essay etc. may be identified as AI generated even if it wasn't.

I suspect and worry that people, laymen and professors, will also have many false positives when deciding if a piece of writing was AI generated (fully or partly). I haven't seen this point made a lot, but it is an obvious issue to me, and look no further than the first few comments in this thread for evidence.

- "Wow, your kid must be a crap writer if an AI generated essay was 'amazing'"; "Certain you don't know good writing...also suspect there is no "kid."" But, luckily, DCUM posters do know good writing and can beat an AI bot - they will let you know when they see it!

- Professors will beat an AI bot over the course of a semester - the professor agrees and will find cheaters to show it.

The capability of generative AI relevant to essay writing is improving extremely rapidly and has become very, very good. It may be that AI bots don't know what a good essay that "stands out" to an admission office looks like and has limited ability to directly "train" on that. But, then virtually no applicant will know it either. Also, it is not fully understood how AI bots do learn as they use an incomprehensible amount of data and "learn" things that they were not asked to learned. So, they may be and very soon become much better than we feel comfortable with.

- "you don’t know good writing if you think these things write in a way that will make someone stand out in admissions." But HS parents do know what these things don't know - it's a heroic assumption.

Dismissing AI is a mistake IMO.

Many top schools give examples of essays written by students that were accepted. JHU is one that immediately comes to mind. They are essentially providing you with the blueprint for the AI. Also, there are expensive college admissions consultants with websites that will provide an example of essays that kids used for Harvard, Stanford, etc. They could be lying, but I don't think so. Again, you can feed those into the AI to train it.

That's very interesting and I'm not particularly surprised. It supports my points.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:Stop pretending the AI bots can’t produce quality work. Many heads of admission at the most elite colleges say it produces essays in the top 1% if you know how to train the AI.

Funny how everyone on DCUM thinks they are Shakespeare.

I agree. Need to find some examples to show here.

I listened to a podcast the other day where the guest was a professor and someone who studies AI. His prediction is that the college essay totally goes away, in short-order, due to the quality of AI essays. Additionally, he thinks that schools need to be moving a lot faster to change the way they teach to account for AI. It's not going away and it's time to reinvent the classroom experience to account for it.

which podcast and who's the professor?

Would be interested to know

[Anonymous]

Anonymous wrote:Agree detecting AI may be impossible. But detecting and rejecting tedious writing is tried and true. Sorry, blowhards.

What does this mean? Who are the blowhards here?

[Anonymous]

Anonymous wrote:My kid entered a draft essay into an AI bot.
It’s now amazing.
And then we entered the result into an AI “test”…..and it passed and said 100% human generated.
The AI version really was better than the human version…it was a “why” essay with a lot of personalization.


I’m just shocked.

Doesn’t your kid to agree to a statement that they didn’t use AI to write their essay for Common App?

[Anonymous]

Anonymous wrote:My kid entered a draft essay into an AI bot.
It’s now amazing.
And then we entered the result into an AI “test”…..and it passed and said 100% human generated.
The AI version really was better than the human version…it was a “why” essay with a lot of personalization.


I’m just shocked.

You are promoting cheating in the college process? That's great for the rest of the kids who are grinding it out on their own.

[Anonymous]

Anonymous wrote:

Anonymous wrote:I don't know much about AI and my impressions and thoughts may be be wrong. However, I do casually follow developments from a birds-eye view. With, this a few points:

A key problem with AI detectors are lots of false positives, i.e., your DC essay etc. may be identified as AI generated even if it wasn't.

I suspect and worry that people, laymen and professors, will also have many false positives when deciding if a piece of writing was AI generated (fully or partly). I haven't seen this point made a lot, but it is an obvious issue to me, and look no further than the first few comments in this thread for evidence.

- "Wow, your kid must be a crap writer if an AI generated essay was 'amazing'"; "Certain you don't know good writing...also suspect there is no "kid."" But, luckily, DCUM posters do know good writing and can beat an AI bot - they will let you know when they see it!

- Professors will beat an AI bot over the course of a semester - the professor agrees and will find cheaters to show it.

The capability of generative AI relevant to essay writing is improving extremely rapidly and has become very, very good. It may be that AI bots don't know what a good essay that "stands out" to an admission office looks like and has limited ability to directly "train" on that. But, then virtually no applicant will know it either. Also, it is not fully understood how AI bots do learn as they use an incomprehensible amount of data and "learn" things that they were not asked to learned. So, they may be and very soon become much better than we feel comfortable with.

- "you don’t know good writing if you think these things write in a way that will make someone stand out in admissions." But HS parents do know what these things don't know - it's a heroic assumption.

Dismissing AI is a mistake IMO.

Many top schools give examples of essays written by students that were accepted. JHU is one that immediately comes to mind. They are essentially providing you with the blueprint for the AI. Also, there are expensive college admissions consultants with websites that will provide an example of essays that kids used for Harvard, Stanford, etc. They could be lying, but I don't think so. Again, you can feed those into the AI to train it.

So AI spits out an amalgam of personal anecdotes, how does that go anywhere?

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:I don't know much about AI and my impressions and thoughts may be be wrong. However, I do casually follow developments from a birds-eye view. With, this a few points:

A key problem with AI detectors are lots of false positives, i.e., your DC essay etc. may be identified as AI generated even if it wasn't.

I suspect and worry that people, laymen and professors, will also have many false positives when deciding if a piece of writing was AI generated (fully or partly). I haven't seen this point made a lot, but it is an obvious issue to me, and look no further than the first few comments in this thread for evidence.

- "Wow, your kid must be a crap writer if an AI generated essay was 'amazing'"; "Certain you don't know good writing...also suspect there is no "kid."" But, luckily, DCUM posters do know good writing and can beat an AI bot - they will let you know when they see it!

- Professors will beat an AI bot over the course of a semester - the professor agrees and will find cheaters to show it.

The capability of generative AI relevant to essay writing is improving extremely rapidly and has become very, very good. It may be that AI bots don't know what a good essay that "stands out" to an admission office looks like and has limited ability to directly "train" on that. But, then virtually no applicant will know it either. Also, it is not fully understood how AI bots do learn as they use an incomprehensible amount of data and "learn" things that they were not asked to learned. So, they may be and very soon become much better than we feel comfortable with.

- "you don’t know good writing if you think these things write in a way that will make someone stand out in admissions." But HS parents do know what these things don't know - it's a heroic assumption.

Dismissing AI is a mistake IMO.

Many top schools give examples of essays written by students that were accepted. JHU is one that immediately comes to mind. They are essentially providing you with the blueprint for the AI. Also, there are expensive college admissions consultants with websites that will provide an example of essays that kids used for Harvard, Stanford, etc. They could be lying, but I don't think so. Again, you can feed those into the AI to train it.

So AI spits out an amalgam of personal anecdotes, how does that go anywhere?

Here is some advice…if you don’t understand it or how it works…just don’t comment.

It is impossible to respond to your comment. It’s as though someone shows you a computer and your response is “it’s just a plastic box with some circuits inside…how will that go anywhere?”

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:I don't know much about AI and my impressions and thoughts may be be wrong. However, I do casually follow developments from a birds-eye view. With, this a few points:

A key problem with AI detectors are lots of false positives, i.e., your DC essay etc. may be identified as AI generated even if it wasn't.

I suspect and worry that people, laymen and professors, will also have many false positives when deciding if a piece of writing was AI generated (fully or partly). I haven't seen this point made a lot, but it is an obvious issue to me, and look no further than the first few comments in this thread for evidence.

- "Wow, your kid must be a crap writer if an AI generated essay was 'amazing'"; "Certain you don't know good writing...also suspect there is no "kid."" But, luckily, DCUM posters do know good writing and can beat an AI bot - they will let you know when they see it!

- Professors will beat an AI bot over the course of a semester - the professor agrees and will find cheaters to show it.

The capability of generative AI relevant to essay writing is improving extremely rapidly and has become very, very good. It may be that AI bots don't know what a good essay that "stands out" to an admission office looks like and has limited ability to directly "train" on that. But, then virtually no applicant will know it either. Also, it is not fully understood how AI bots do learn as they use an incomprehensible amount of data and "learn" things that they were not asked to learned. So, they may be and very soon become much better than we feel comfortable with.

- "you don’t know good writing if you think these things write in a way that will make someone stand out in admissions." But HS parents do know what these things don't know - it's a heroic assumption.

Dismissing AI is a mistake IMO.

Many top schools give examples of essays written by students that were accepted. JHU is one that immediately comes to mind. They are essentially providing you with the blueprint for the AI. Also, there are expensive college admissions consultants with websites that will provide an example of essays that kids used for Harvard, Stanford, etc. They could be lying, but I don't think so. Again, you can feed those into the AI to train it.

So AI spits out an amalgam of personal anecdotes, how does that go anywhere?

Here is some advice…if you don’t understand it or how it works…just don’t comment.

It is impossible to respond to your comment. It’s as though someone shows you a computer and your response is “it’s just a plastic box with some circuits inside…how will that go anywhere?”

Exactly, what's going on here? So many relying on denial and wishful thinking.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:I don't know much about AI and my impressions and thoughts may be be wrong. However, I do casually follow developments from a birds-eye view. With, this a few points:

A key problem with AI detectors are lots of false positives, i.e., your DC essay etc. may be identified as AI generated even if it wasn't.

I suspect and worry that people, laymen and professors, will also have many false positives when deciding if a piece of writing was AI generated (fully or partly). I haven't seen this point made a lot, but it is an obvious issue to me, and look no further than the first few comments in this thread for evidence.

- "Wow, your kid must be a crap writer if an AI generated essay was 'amazing'"; "Certain you don't know good writing...also suspect there is no "kid."" But, luckily, DCUM posters do know good writing and can beat an AI bot - they will let you know when they see it!

- Professors will beat an AI bot over the course of a semester - the professor agrees and will find cheaters to show it.

The capability of generative AI relevant to essay writing is improving extremely rapidly and has become very, very good. It may be that AI bots don't know what a good essay that "stands out" to an admission office looks like and has limited ability to directly "train" on that. But, then virtually no applicant will know it either. Also, it is not fully understood how AI bots do learn as they use an incomprehensible amount of data and "learn" things that they were not asked to learned. So, they may be and very soon become much better than we feel comfortable with.

- "you don’t know good writing if you think these things write in a way that will make someone stand out in admissions." But HS parents do know what these things don't know - it's a heroic assumption.

Dismissing AI is a mistake IMO.

Many top schools give examples of essays written by students that were accepted. JHU is one that immediately comes to mind. They are essentially providing you with the blueprint for the AI. Also, there are expensive college admissions consultants with websites that will provide an example of essays that kids used for Harvard, Stanford, etc. They could be lying, but I don't think so. Again, you can feed those into the AI to train it.

So AI spits out an amalgam of personal anecdotes, how does that go anywhere?

Here is some advice…if you don’t understand it or how it works…just don’t comment.

It is impossible to respond to your comment. It’s as though someone shows you a computer and your response is “it’s just a plastic box with some circuits inside…how will that go anywhere?”

Exactly, what's going on here? So many relying on denial and wishful thinking.

The number of published “successful” essays is not sufficient to use a training set

[Anonymous]

Soon the college application process will exclude standardized test scores, essays, demographic information, and anything else that might distinguish applicants due to cheating, favoritism, and/or discrimination.

[Anonymous]

Anonymous wrote:Soon the college application process will exclude standardized test scores, essays, demographic information, and anything else that might distinguish applicants due to cheating, favoritism, and/or discrimination.

People have always cheated.

[Anonymous]

I write things at work all the time and AI it and the results are much better than me spending an hour rewriting and having staff edit.

Also the content is not AI generated.

Interesting though I wrote something and the bot noticed there were 3 themes and structured it better to highlight the 3 themes I didn’t notice.

I actually don’t think having a bot edit is cheating.

[Anonymous]

We use it at work now all the time. We’ll have it essentially proof read boring documents, or spit out a first draft of something that we use as discussion points. I have a great staff member whose emails are not well written and often offend people. I’ve told her to send all her emails through ChatGPT to correct the tone. It takes 20 seconds, and prevents days of interpersonal mop up.

My kid does use it for school, I’ve found. I’ve also found he is pretty savvy about what it can and can’t do, and when it is acting as an editor and when it is cheating.

It’s here, people. The question is how we use it, not how we detect it or outlaw it.

[Anonymous]

How do you learn how to use AI well? Are you just asking it to improve something and then inserting your text? I’ve asked questions and the writing isn’t great but it sounds like I don’t know how to use it.

[Anonymous]

Anonymous wrote:I actually don’t think having a bot edit is cheating.

Some colleges seem to think so and ask for an attestation in the app that the kid did not use AI.