Today I decided to subscribe to X just to support Elon; they offer a benefit of their beta AI software, Grok.

I asked Grok how to find road conditions in northern New Hampshire for day of eclipse. Grok directed me to NewEngland511.org and I signed up. It's a very good site and should help a lot.

Actually, I tested Grok earlier with questions like "Who is Abigail Shrier". It gave me pluses and minuses; with references.
Also asked about evidence base for safety of statin drugs: it again directed me to--yes, mainstream--sources, but also highlighted controversy.

If Musk is able to keep Grok neutral and not bat-shit crazy like Google's AI, it may be a first go-to for complicated searches without ending up with three pages of ads. We'll see.

Update: I asked "Is Tik-Tok addictive for young people"and it gave a good response IMO. Then it showed a number of related X posts with different sides of the issue. Of course X has a conflict of interest in this question, but I thought it was pretty good. I look forward to trying more things.

Link?

No link because I think you need to be subscribed to X to get access.

How does Grok compare with AI Chat (formerly known as ChatGPT)?

Went to try it but says you need X Premium+ which is a paid subscription. To bad I like to try it.

True story from a few weeks ago.
Student:
"Dr. Redstone, can you write me a recommendation letter for an internship with the NSA this summer?"
"Sure, send me the information and I will get to it this weekend."
"It is due this afternoon."

crap.

I plugged the Kids name, age, skills, and our relationship (student in networking and IT courses as well as paid employee for a summer camp last summer). Then I plugged in the name of the internship. (which requested you address six main principles etc.)

It wrote one of the best recommendation letters ever. I had to go through and change the "[insert chosen pronouns] here" and other bits of information but 90% of it was written.
What amazed me was it looked up the application and addressed all six of the main requests.

By far one of the best recommendation letters I never wrote.

quote:

"It is due this afternoon."

"Good student, shitty time management and planning skills" in there?

quote:

Originally posted by redstone:
True story from a few weeks ago.
<snip>

Presumably Grok, since that’s what this thread’s about?

quote:

Originally posted by redstone:
True story from a few weeks ago.
Student:
"Dr. Redstone, can you write me a recommendation letter for an internship with the NSA this summer?"
"Sure, send me the information and I will get to it this weekend."
"It is due this afternoon."

crap.

I plugged the Kids name, age, skills, and our relationship (student in networking and IT courses as well as paid employee for a summer camp last summer). Then I plugged in the name of the internship. (which requested you address six main principles etc.)

It wrote one of the best recommendation letters ever. I had to go through and change the "[insert chosen pronouns] here" and other bits of information but 90% of it was written.
What amazed me was it looked up the application and addressed all six of the main requests.

By far one of the best recommendation letters I never wrote.

If you have Grammarly try running that letter through it's plagiarism detector, I'd be intrested to see the results.

Yes, I have a X account subscription and used Grok. It was my first time using Grok in particular. Should I not have commented?

We have a plagiarism checker that we use for our Learning Management system. It passed the checker (it is an original work) but It's AI detection tool flagged it as having been written 40% by AI.
ZeroGPT (another AI detection tool) returned it as 32% having been written by AI.

Since you mention Grammarly, Grammarly is an AI now, and most of its work will return as having been written by AI.

quote:

Originally posted by redstone:
Yes, I have a X account subscription and used Grok. It was my first time using Grok in particular. Should I not have commented?

We have a plagiarism checker that we use for our Learning Management system. It passed the checker (it is an original work) but It's AI detection tool flagged it as having been written 40% by AI.
ZeroGPT (another AI detection tool) returned it as 32% having been written by AI.

Since you mention Grammarly, Grammarly is an AI now, and most of its work will return as having been written by AI.

I hope I didn't accidentally insult you.

I found your comment informative and was looking for the stats you provided in your reply. Specifically, that Grammarly determined that the AI's work was original. The AI detection tools results were very interesting as well. Is that 32% realistic based on the amount of effort you put into the letter?

I've got to see if USA jobs uses an AI detector.

Thank you for sharing your results.

quote:

Originally posted by redstone:


We have a plagiarism checker that we use for our Learning Management system. It passed the checker (it is an original work) but It's AI detection tool flagged it as having been written 40% by AI.
ZeroGPT (another AI detection tool) returned it as 32% having been written by AI.

That's itnteresting that it's flagged something as having been written as some percentage by AI>

I've been using AI recently for checking for readability, accuracy, and clarity. I start with my own text but I've found that AI gives back some useful rewrites. Sometimes though it changes the original meaning. I can even make it review a section for conciseness and it's done that admirably.

Often, I would do several iterations as I'd incorporate some of the edits but not yet satisfied overall. You also have to stay on top of it as far as facts as it's not infallible.

I tried inputting back some of the results just now and it's come back as "the text you provided appears to be original and does not seem to be plagiarized. There's no indication of the use of AI-generated content."

I was afraid it was keeping records of its outputs and then claiming copy right.

We will regret this.

I had a colleague who is still working on her dissertation. While we were talking she ran her current paper through zeroGPT and it came back 100% AI. She was near hysterics. It was all because she depends so heavily on Grammarly. Most do not realize that it is an AI.

She asked, "how do I write without Grammarly?" and I mercilessly suggested she write it herself, and then hire an editor (and then followed it up with my Mandolorian voice, "This is the Way"). She did not think I was very funny.

In Academia, reputation is everything, and what happens when someone writes an article claiming "all of your research, papers, dissertation, presentations etc. were all written with AI?" What happens then? That is why all my formal writing, I avoid the 'assistance' tools.

BUT the reason universities cannot use these tools (AI detectors) is because of how unreliable they are. One of my adult students would get flagged as AI every time but it was because he writes for the Army. He said writing for the Army makes him write very clinically. Also, some historical texts have been flagged as AI.

I see a near future, where assessment returns to a in person format, oral defense, etc. Already, I have companies complaining they are hiring employees that 'on paper' can code. But in reality they cannot. So I believe we will soon have to do things like, "write me a simple fractal script in Python that takes user input and can be adjusted dynamically. The computer you will use is not connected to the Internet."

quote:

Originally posted by redstone:


She asked, "how do I write without Grammarly?" and I mercilessly suggested she write it herself, and then hire an editor

But I have to ask: what's the difference between hiring an editor and using an AI in the function of an editor?

WHy would I need to go to the library to look up facts when I can use Google?

I use the charting capabilities of Excel instead of drawing a chart using pencil and ruler.

I think there is a difference between "write me a paper on such and such" and "review the following for grammatical errors and clarity:"

I look at the rewrites I got from AI. Yes, I've edited my own work before but it takes 4 x the amount to edit paragraphs than the amount it took to write the original material in the first place.

That is the core of the debate. When does it stop being you re-writing it and becomes the work of a machine?

When you edit it, it is your work, when someone else edits it, they do not re-write, only highlight poor grammar, logical errors etc. You still have to complete the re-writes yourself. I have never had an editor actually change what I had written, only point out redundancies, poor grammar, et al.

AI tools actually re-write the paragraph etc. So the debate is now is that ok? Is this the new normal?
Academia as a whole is now facing an existential crises. I clearly fall in the "do the work yourself." category. A programmer in the field emailed me that he was spending more time fixing his staffs programming errors than actually doing his own work. They were letting AI write the code wrongly, did not know how to fix it or were not spending the time to trouble shoot it. So in some ways it is wasting valuable time and resources and not proving helpful.

When I was taking a gunsmithing class, Dremmels were forbidden. The instructor said more incredible firearms had been ruined by that stupid tool.

quote:

Originally posted by redstone:

Academia as a whole is now facing an existential crises. I clearly fall in the "do the work yourself." category. A programmer in the field emailed me that he was spending more time fixing his staffs programming errors than actually doing his own work. They were letting AI write the code wrongly, did not know how to fix it or were not spending the time to trouble shoot it. So in some ways it is wasting valuable time and resources and not proving helpful.

I saw something similar in one of my classes. A classmate used chatgpt to write a script. Most of it worked but not all of it. The interesting part was that among the final lines of code was one that would display "script run successfully." It would display this even if the scripted failed. Luckily he caught it when he reviewed the AI's work.