https://www.nature.com/articles/d41586-025-02936-6

AI tool detects LLM-generated text in research papers and peer reviews
Authors and peer reviewers are failing to disclose the use of LLMs despite journal policies limiting their use.

Testing the AI-detection tool on manuscripts before ChatGPT was released in November 2022, it flagged only seven abstracts and no methods or peer-review reports as containing potentially AI-generated text. “From there on, the detections just increased linearly and at what we would think is a very high rate,” says Evanko.

The tool can also distinguish between different LLMs, including ChatGPT models, DeepSeek, LLaMa and Claude. “We’re only able to do this because we’ve generated our entire training set ourselves, so we know the exact provenance, we know what model the training data came from,” explains Spero.

The current model of Pangram cannot distinguish between passages that are fully generated by AI and those that are written by humans but edited using AI.

 

https://www.nature.com/articles/d41586-025-02936-6

AI tool detects LLM-generated text in research papers and peer reviews
Authors and peer reviewers are failing to disclose the use of LLMs despite journal policies limiting their use.

Testing the AI-detection tool on manuscripts before ChatGPT was released in November 2022, it flagged only seven abstracts and no methods or peer-review reports as containing potentially AI-generated text. “From there on, the detections just increased linearly and at what we would think is a very high rate,” says Evanko.

The tool can also distinguish between different LLMs, including ChatGPT models, DeepSeek, LLaMa and Claude. “We’re only able to do this because we’ve generated our entire training set ourselves, so we know the exact provenance, we know what model the training data came from,” explains Spero.

The current model of Pangram cannot distinguish between passages that are fully generated by AI and those that are written by humans but edited using AI.

So to prove someone has used AI we must rely on AI.

 

So to prove someone has used AI we must rely on AI.

This is actually one of the tasks that things like neural nets are pretty good at -- and for reasons that are well understood. It's a classification problem in which the system is trained on lots of different examples, each time being told what the answer should be. Then, when presented with something it hasn't seen, it merely determines which of the things it has been trained on most resembles it.

 

The devil knows the demon.

 

The devil knows the demon.

I don't think that has anything to do with it in this case at all. It doesn't matter that the things being classified in this instance are things produced by AIs (and classification AIs are very different critters than generative AIs, on top of that). The same would be true in trying to have a system classify medieval works of art, ancient manuscripts, or prehistoric cave paintings.

 

I don't think that has anything to do with it in this case at all. It doesn't matter that the things being classified in this instance are things produced by AIs (and classification AIs are very different critters than generative AIs, on top of that). The same would be true in trying to have a system classify medieval works of art, ancient manuscripts, or prehistoric cave paintings.

I know, just making fun of the AI reference to something that's been in use for decades before modern LLM systems.
https://forum.allaboutcircuits.com/threads/chatgpt.191318/post-1932033

It's really about classifiers (convolutional neural networks), not LLM systems like ChatCGT.
https://www.mlopsaudits.com/blog/llms-vs-traditional-ml-algorithms-comparison

September 22, 2017
https://www.topbots.com/chihuahua-muffin-searching-best-computer-vision-api/
Chihuahua OR Muffin? Searching For The Best Computer Vision API

Binary classification has been possible ever since the perceptron algorithm was invented in 1957. If you think AI is hyped now, the New York Times reported in 1958 that the invention was the beginning of a computer that would “be able to walk, talk, see, write, reproduce itself and be conscious of its existence.” But while perceptron machines like the Mark 1 were designed for image recognition, in reality they could only discern patterns that are linearly separable, preventing them from learning the complex patterns that underlie most visual media.

https://forum.allaboutcircuits.com/threads/theory-of-everything.168244/post-1492322
https://zetlab.com/en/types-of-sonar-systems-lofar-and-demon-analysis/

There are two basic methods used for passive sonar signal processing: LOFAR and DEMON analysis (LOFAR – Low-Frequency Analysis and Recording; DEMON – Demodulation of Envelope Modulation On Noise).

 

https://www.404media.co/librarians-are-being-asked-to-find-ai-hallucinated-books/
Librarians Are Being Asked to Find AI-Hallucinated Books

“Librarians are reporting this overall atmosphere of confusion and lack of trust they’re experiencing from their patrons,” Macrina told 404. “They’re seeing patrons having seemingly diminished critical thinking and curiosity. They’re definitely running into some of these psychosis and other mental health issues, and certainly seeing the people who are more widely adopting it also being those who have less digital literacy about it and a general sort of loss of retention.”
...
“We are trying to teach how to construct useful, exact searching,” she said. “But really [these products’ intent] is to make that not happen. The problem with that in a university library is we’re trying to teach those skills but we have tools that negate that necessity. And because those tools don’t work well, you’ve not learned the skill and you’re still getting crap results, so you’re never going to get better results because you didn’t learn the skill.”

https://ezgranet.github.io/thiotimoline/
The Endochronic Properties of Resublimated Thiotimoline

 

https://cacm.acm.org/blogcacm/feeling-cranky-about-ai-and-cs-education/

So here we are today, in the throes of the next hand-wringing “what will we do” moment with AI. Here are my issues and concerns:

1. First, let’s be clear about nomenclature. Everyone says AI, but what they really mean is generative AI, which is based on machine learning. There’s no discussion of the myriad other subfields and techniques that have historically been part of AI and still help us solve all sorts of interesting problems.

2. We’re jumping through hoops without stopping first to question the run-away train. In much discussion about CS education:
a.) There’s little interest in interrogating the downsides of generative AI, such as the environmental impact, the data theft impact, the treatment and exploitation of data workers.
b.) There’s little interest in considering the extent to which, by incorporating generative AI into our teaching, we end up supporting a handful of companies that are burning billions in a vain attempt to each achieve performance that is a scintilla better than everyone else’s.
c.) There’s little interest in thinking about what’s going to happen when the LLM companies decide that they have plateaued, that there’s no more money to burn spend, and a bunch of them fold—but we’ve perturbed education to such an extent that our students can no longer function without their AI helpers.

 

https://cacm.acm.org/blogcacm/feeling-cranky-about-ai-and-cs-education/

So here we are today, in the throes of the next hand-wringing “what will we do” moment with AI. Here are my issues and concerns:

1. First, let’s be clear about nomenclature. Everyone says AI, but what they really mean is generative AI, which is based on machine learning. There’s no discussion of the myriad other subfields and techniques that have historically been part of AI and still help us solve all sorts of interesting problems.

2. We’re jumping through hoops without stopping first to question the run-away train. In much discussion about CS education:
a.) There’s little interest in interrogating the downsides of generative AI, such as the environmental impact, the data theft impact, the treatment and exploitation of data workers.
b.) There’s little interest in considering the extent to which, by incorporating generative AI into our teaching, we end up supporting a handful of companies that are burning billions in a vain attempt to each achieve performance that is a scintilla better than everyone else’s.
c.) There’s little interest in thinking about what’s going to happen when the LLM companies decide that they have plateaued, that there’s no more money to burn spend, and a bunch of them fold—but we’ve perturbed education to such an extent that our students can no longer function without their AI helpers.

I agree with her on just about all of the points, and she brings up some points that are almost always completely overlooked, but still relevant in the long run.

Her observations that CS programs love to jump on board and go whole hog with the latest craze and fad is true in spades. I've seen it time and time again, including all of the ones she mentioned and then a few. It's actually a point that some people put forth, somewhat facetiously, that the field of "computer science" isn't a "science" at all. It's too quick to trade emphasis on fundamentals in order to incorporate those latest fads. It's those weakening fundamentals, which have been weakening for a long time and only amped up by generative-AI tools, that is my primary concern.

 

I agree with her on just about all of the points, and she brings up some points that are almost always completely overlooked, but still relevant in the long run.

Her observations that CS programs love to jump on board and go whole hog with the latest craze and fad is true in spades. I've seen it time and time again, including all of the ones she mentioned and then a few. It's actually a point that some people put forth, somewhat facetiously, that the field of "computer science" isn't a "science" at all. It's too quick to trade emphasis on fundamentals in order to incorporate those latest fads. It's those weakening fundamentals, which have been weakening for a long time and only amped up by generative-AI tools, that is my primary concern.

It's the boring skills (CS major and minor in Physics) my daughter is being taught in her third and final year before she completes the PSU Bachelor of Science degree.
So far I've been happy with their course materials, direction and she's been hired (her first 'real' paying job) by PSU as a remote learning assistant for a few extra bucks this year.
I've got lots of my old fundamental books (OS, networking database, data structures, etc ...) that I've shown to her.

She's reading it, not dumping it.

 

It's the boring skills (CS major and minor in Physics) my daughter is being taught in her third and final year before she completes the PSU Bachelor of Science degree.
So far I've been happy with their course materials, direction and she's been hired (her first 'real' paying job) by PSU as a remote learning assistant for a few extra bucks this year.
I've got lots of my old fundamental books (OS, networking database, data structures, etc ...) that I've shown to her.
View attachment 356248
She's reading it, not dumping it.

I'm sure that you're being of tremendous help to her. And that she's beyond grateful to have a dad like you.

Life's been good to you, dude ... and it ain't fair! ...

 

I'm sure that you're being of tremendous help to her. And that she's beyond grateful to have a dad like you.

Life's been good to you, dude ... and it ain't fair! ...

Yes, I was very lucky.

 

https://www4.courts.ca.gov/opinions/documents/B331918.PDF

What sets this appeal apart—and the reason we have elected to publish this opinion—is that nearly all of the legal quotations in plaintiff’s opening brief, and many of the quotations in plaintiff’s reply brief, are fabricated. That is, the quotes plaintiff attributes to published cases do not appear in those cases or anywhere else. Further, many of the cases plaintiff cites do not discuss the topics for which they are cited, and a few of the cases do not exist at all. These fabricated legal authorities were created by generative artificial intelligence (AI) tools that plaintiff’s counsel used to draft his appellate briefs. The AI tools created fake legal authority—sometimes referred to as AI “hallucinations”—that were undetected by plaintiff’s counsel because he did not read the cases the AI tools cited.
Although the generation of fake legal authority by AI sources has been widely commented on by federal and out-ofstate courts and reported by many media sources, no California court has addressed this issue. We therefore publish this opinion as a warning. Simply stated, no brief, pleading, motion, or any other paper filed in any court should contain any citations— whether provided by generative AI or any other source—that the attorney responsible for submitting the pleading has not personally read and verified. Because plaintiff’s counsel’s conduct in this case violated a basic duty counsel owed to his client and the court, we impose a monetary sanction on counsel, direct him to serve a copy of this opinion on his client, and direct the clerk of the court to serve a copy of this opinion on the State Bar.
...
We conclude by noting that “hallucination” is a particularly apt word to describe the darker consequences of AI. AI hallucinates facts and law to an attorney, who takes them as real and repeats them to a court. This court detected (and rejected) these particular hallucinations. But there are many instances— hopefully not in a judicial setting—where hallucinations are circulated, believed, and become “fact” and “law” in some minds. We all must guard against those instances. As a federal district court recently noted: “There is no room in our court system for the submission of fake, hallucinated case citations, facts, or law. And it is entirely preventable by competent counsel who do their jobs properly and competently.”
...
DISPOSITION The judgment is affirmed. Attorney Amir Mostafavi is directed to pay $10,000 in sanctions, payable to the clerk of this court, no later than 30 days after the remittitur is filed. The clerk is directed to deposit this sum into the court’s general fund.

 

I agree with her on just about all of the points, and she brings up some points that are almost always completely overlooked, but still relevant in the long run.

Her observations that CS programs love to jump on board and go whole hog with the latest craze and fad is true in spades. I've seen it time and time again, including all of the ones she mentioned and then a few. It's actually a point that some people put forth, somewhat facetiously, that the field of "computer science" isn't a "science" at all. It's too quick to trade emphasis on fundamentals in order to incorporate those latest fads. It's those weakening fundamentals, which have been weakening for a long time and only amped up by generative-AI tools, that is my primary concern.

Hi there,

I've been saying that Computer Science is not really a Science for years now. That's because I noticed that it is mixed in with manufacturing bias. This is not just a little bias but it's big.
Also, it is not usual for a scientist to make the same mistakes over and over and over again, just because they did the same experiment more than once, and even if they corrected the first version already.
With Microsoft for example, they make the same mistakes over and over again, then either have to correct them AGAIN or just ignore them. I should create a list. So it's not even recursive. It's really annoying too when we get an update and it includes errors that were corrected LONG ago. In fact, wayyyy long ago. It makes them look stupid too. It tells me that their internal software documentation method is not very good at all; they can't keep track of what they have done in the past even if it's just a year ago.

 

https://www4.courts.ca.gov/opinions/documents/B331918.PDF

What sets this appeal apart—and the reason we have elected to publish this opinion—is that nearly all of the legal quotations in plaintiff’s opening brief, and many of the quotations in plaintiff’s reply brief, are fabricated. That is, the quotes plaintiff attributes to published cases do not appear in those cases or anywhere else. Further, many of the cases plaintiff cites do not discuss the topics for which they are cited, and a few of the cases do not exist at all. These fabricated legal authorities were created by generative artificial intelligence (AI) tools that plaintiff’s counsel used to draft his appellate briefs. The AI tools created fake legal authority—sometimes referred to as AI “hallucinations”—that were undetected by plaintiff’s counsel because he did not read the cases the AI tools cited.
Although the generation of fake legal authority by AI sources has been widely commented on by federal and out-ofstate courts and reported by many media sources, no California court has addressed this issue. We therefore publish this opinion as a warning. Simply stated, no brief, pleading, motion, or any other paper filed in any court should contain any citations— whether provided by generative AI or any other source—that the attorney responsible for submitting the pleading has not personally read and verified. Because plaintiff’s counsel’s conduct in this case violated a basic duty counsel owed to his client and the court, we impose a monetary sanction on counsel, direct him to serve a copy of this opinion on his client, and direct the clerk of the court to serve a copy of this opinion on the State Bar.
...
We conclude by noting that “hallucination” is a particularly apt word to describe the darker consequences of AI. AI hallucinates facts and law to an attorney, who takes them as real and repeats them to a court. This court detected (and rejected) these particular hallucinations. But there are many instances— hopefully not in a judicial setting—where hallucinations are circulated, believed, and become “fact” and “law” in some minds. We all must guard against those instances. As a federal district court recently noted: “There is no room in our court system for the submission of fake, hallucinated case citations, facts, or law. And it is entirely preventable by competent counsel who do their jobs properly and competently.”
...
DISPOSITION The judgment is affirmed. Attorney Amir Mostafavi is directed to pay $10,000 in sanctions, payable to the clerk of this court, no later than 30 days after the remittitur is filed. The clerk is directed to deposit this sum into the court’s general fund.

I would say that they got off way too lightly. How much deterrence is a $10k fine for a law firm? How much would they have expected to earn, total, from the case had they prevailed? I think the penalty should have been some multiple of that amount. In Colorado (and I think this is pretty common), if I write a bad check, I'm liable for the larger of $100 or three times the amount of the check. The intent is pretty clear -- people that hope to profit by writing bad checks only have to get caught a fraction of the time before they are on the losing end. They essentially perpetrated a fraud on the court, at least in essence. I think the defendants should also be able to sue the law firm for either abuse of process or misrepresentation. But, this is almost certainly not possible since lawyers have managed to exempt themselves from the very laws that they demand everyone else toe the line on. Hopefully, the California Bar will come down on them hard -- but I'm not holding my breath.

 

https://hbr.org/2025/09/ai-generated-workslop-is-destroying-productivity

AI-Generated “Workslop” Is Destroying Productivity

A confusing contradiction is unfolding in companies embracing generative AI tools: while workers are largely following mandates to embrace the technology, few are seeing it create real value. Consider, for instance, that the number of companies with fully AI-led processes nearly doubled last year, while AI use has likewise doubled at work since 2023. Yet a recent report from the MIT Media Lab found that 95% of organizations see no measurable return on their investment in these technologies. So much activity, so much enthusiasm, so little return. Why?

In collaboration with Stanford Social Media Lab, our research team at BetterUp Labs has identified one possible reason: Employees are using AI tools to create low-effort, passable looking work that ends up creating more work for their coworkers. On social media, which is increasingly clogged with low-quality AI-generated posts, this content is often referred to as “AI slop.” In the context of work, we refer to this phenomenon as “workslop.” We define workslop as AI generated work content that masquerades as good work, but lacks the substance to meaningfully advance a given task.

 

https://hbr.org/2025/09/ai-generated-workslop-is-destroying-productivity

AI-Generated “Workslop” Is Destroying Productivity

This arguably belongs in the "Captain Obvious Headlines" thread.

Just more evidence of the one-level-deep thinking that so many "leaders" exhibit, jumping on whatever the latest craze is without even attempting to consider things rationally. Also, further evidence that if a technology can be abused and misused, it will be.

 

It's the boring skills (CS major and minor in Physics) my daughter is being taught in her third and final year before she completes the PSU Bachelor of Science degree.
So far I've been happy with their course materials, direction and she's been hired (her first 'real' paying job) by PSU as a remote learning assistant for a few extra bucks this year.
I've got lots of my old fundamental books (OS, networking database, data structures, etc ...) that I've shown to her.
View attachment 356248
She's reading it, not dumping it.

I would say that she's fortunate to have you backfilling her with traditional materials and it's good that she's interested enough to be paying attention to them.

I took a look at PSU's program and it is going down the same path as most of the ones I'm aware off -- getting rid of what used to be key upper division requirements, such as Computer Architecture (as a second, follow-on course to Computer Organization), Automata, Compiler Design, and Computer Networks. But this is actually the first program I've seen that has eliminated all four as required courses, most of only eliminated two of them (so far). But, then, they had to get rid of another two courses from the traditional program in order to make room for the university's new requirement for taking two courses on racism and the evils of European and American colonialism and imperialism. I was surprised to see that it still requires three semesters of calculus, though I'm a bit puzzled as to why, since none of the required courses even has Calc 1 as a requirement. This has become pretty common, even though a decade or two ago, the first programming course often had Calc 1 as a requirement, though admittedly, it wasn't really needed to actually understand the programming concepts in the course, but it provided the basis for a lot of programming problems that were germane to real-world applications while remaining very simple to understand and solve. Unfortunately, the fundamental math skills have been so eroded that programs have gone out of their way to excise anything that requires calculus, even from courses where it is really central to understanding some of the core concepts. Instead, those concepts are now presented as equations that the students have to memorize instead of ones that they derived from foundational principles.

 

She's also in the honors program with a physics minor. If she later wants to go down that path, most of the math required is done.

 

She's also in the honors program with a physics minor. If she later wants to go down that path, most of the math required is done.

The fact that she's doing physics, even as a minor, is a good sign. People that choose an emphasis in physics are generally people that have a thirst for a deeper understanding of whatever they are studying. As a result, they are often able to rise and get a good education even in programs or at institutions that have gone downhill (or where never up the hill in the first place). That's a general observation, BTW, and is completely divorced from PSU or any specific institution and program. I've always maintained that a truly motivated student can find a way to get an MIT-level education at even a low-ranked state college, and that a very poorly motivated MIT student can similarly find a way to get a low-ranked state college level education there. It just requires a lot of effort on both parties' parts.