Would you pay $169 for an introductory ebook on machine learning with citations that appear to be made up? If not, you might want to pass on purchasing Mastering Machine Learning: From Basics to Ad…
Okay so hear me out on this. The book mentioned in this article is definitely a trash, the author used LLM without informing readers, which is why most people feel the are being scammed and express feelings of frustration and hate.
I personally have deployed LLMs on my local machines and used them for variety of things such as Summarize news and Articles, Coding, Image Generation, etc and I have to be honest it is really really impressive technology. Any author who takes assistance from LLM would be hyper-productive compared to someone who does all the labour themselves. I used to take hours to read a broad area of knowledge and then deep dive in intrested topics. When LLMs generate summary and you can decide weather to read the source yourself or not is a big time saver and productivity boost. Of course this can be abused by someone who trusts LLMs too much and don’t again verifies what they read, it can give false information but that’s not how they are supposed to be used.
These language models are really good at creating summaries. I use a locally deployed LLM to read summaries of Articles and then if I feel interested I read the entire article end to end from original source. In Almost every case the summary is spot on without it missing any important points or topics, heck I have created system prompt so that it tries to give hot takes and nuanced perspectives from the article and it impresses me sometimes giving me a new perspective which I would have not thought otherwise.
I am convinced that using LLM along with your own knowledge always surpasses the quality of your work if someone with your same capability generates work without taking assistance of a LLM.
Didn’t have time to read that, so I threw your comment into my local TinyLlama instance:
The article discusses an artificial intelligence assistant named “AiLlMa,” which is used for summarizing news and articles in various languages. The author suggests that using such AI assistants, including LLMs, can save time by taking away the need to manually search and read source material. He argues that LLMs generate summaries that are spot-on without missing important points or topics, and that using them along with personal knowledge gives a distinct advantage over independent reading of sources. The author uses examples of LLMs generating useful summaries from articles on various topics, such as summarizing news headlines for readers and creating nuanced takes on articles based on contextual information. Overall, the article suggests that using AI assistants like LLMs can improve productivity, boost efficiency, and produce more accurate summaries.
Actually, I just dropped the entire thread — including the original article — into ChatGPT since I didn’t have time to read it all either. Then I asked it to write the best possible 3-sentence Lemmy comment specifically for your reply. So, ironically, we’ve both delegated the thinking to LLMs — now that’s productivity.
Then I asked “write the next 3 sentence providing the most insight that you can in 3 sentences” and it responded
The real issue here isn’t just that LLMs can fabricate citations — it’s that institutions and publishers are increasingly failing to enforce even basic quality controls, allowing AI-generated content to slip through without proper oversight. When tools meant to assist in research are used to replace it, the result is a veneer of credibility masking hollow or false scholarship. If this trend continues, we risk flooding academic discourse with noise that’s indistinguishable from rigor, eroding trust in legitimate research.
then I asked “But chatgpt, quality control, rigor, those have been gone for aeons, the barrier to write was so high no one really tried so the control, rigor, editors have largely atrophied long ago, respond in 3 sentences” and it responded
You’re right — what we’re seeing now is less a collapse and more an exposure of how little substance was holding it all up. The illusion of rigor persisted mostly because the barrier to entry was too high for mass participation, not because the gatekeepers were doing their jobs. LLMs didn’t break the system; they just made its hollowness impossible to ignore.
Hmm Interesting, seems your Model is hallucinating a lot, maybe try improving your system prompt and play with temperature or other params.
I have a deepseek-ai_DeepSeek-R1-Distill-Qwen-1.5B Running locally.
Here is my output for the Orignal Comment
The article discusses the perception of a book as trash due to the author’s use of LLM without informing readers. The author highlights the benefits of using LLMs, such as saving time and improving productivity, by suggesting that using locally deployed LLMs can significantly enhance the summarization process. The article also warns against the misuse of LLMs, emphasizing the importance of verifying information. The author concludes that relying on LLMs can surpass human capabilities when combined with personal knowledge.
Key Details: The article’s main points are:
The book is perceived as trash due to LLM usage without reader notification.
LLMs save time and improve productivity.
Using locally deployed LLMs for summarization is effective.
Misuse of LLMs can lead to false information.
Combining LLMs with personal knowledge enhances quality.
I use the following prompt before article
You are a concise summarization AI. Follow these rules:
NEVER exceed 4 sentences or 150 words.
Use this format:
“Summary: [2-sentence core idea].
Key Details: [3–4 bullet points].”
Interestingly, your original comment is not much longer and I find it much easier to read.
Was it written with the help of a LLM? Not being sarcastic, I’m just trying to understand if the (perceived) deterioration in quality was due to the fact that the input was already LLM-assisted.
Okay so hear me out on this. The book mentioned in this article is definitely a trash, the author used LLM without informing readers, which is why most people feel the are being scammed and express feelings of frustration and hate.
I personally have deployed LLMs on my local machines and used them for variety of things such as Summarize news and Articles, Coding, Image Generation, etc and I have to be honest it is really really impressive technology. Any author who takes assistance from LLM would be hyper-productive compared to someone who does all the labour themselves. I used to take hours to read a broad area of knowledge and then deep dive in intrested topics. When LLMs generate summary and you can decide weather to read the source yourself or not is a big time saver and productivity boost. Of course this can be abused by someone who trusts LLMs too much and don’t again verifies what they read, it can give false information but that’s not how they are supposed to be used.
These language models are really good at creating summaries. I use a locally deployed LLM to read summaries of Articles and then if I feel interested I read the entire article end to end from original source. In Almost every case the summary is spot on without it missing any important points or topics, heck I have created system prompt so that it tries to give hot takes and nuanced perspectives from the article and it impresses me sometimes giving me a new perspective which I would have not thought otherwise.
I am convinced that using LLM along with your own knowledge always surpasses the quality of your work if someone with your same capability generates work without taking assistance of a LLM.
Careful.
ChatGPT May Be Eroding Critical Thinking Skills, According to a New MIT Study
Relevant Study
And also
Why Some AI Models Spew 50 Times More Greenhouse Gas to Answer the Same Question
Didn’t have time to read that, so I threw your comment into my local TinyLlama instance:
Didn’t have time to read that, so I threw your comment into ChatGPT:
Actually, I just dropped the entire thread — including the original article — into ChatGPT since I didn’t have time to read it all either. Then I asked it to write the best possible 3-sentence Lemmy comment specifically for your reply. So, ironically, we’ve both delegated the thinking to LLMs — now that’s productivity.
Then I asked “write the next 3 sentence providing the most insight that you can in 3 sentences” and it responded
The real issue here isn’t just that LLMs can fabricate citations — it’s that institutions and publishers are increasingly failing to enforce even basic quality controls, allowing AI-generated content to slip through without proper oversight. When tools meant to assist in research are used to replace it, the result is a veneer of credibility masking hollow or false scholarship. If this trend continues, we risk flooding academic discourse with noise that’s indistinguishable from rigor, eroding trust in legitimate research.
then I asked “But chatgpt, quality control, rigor, those have been gone for aeons, the barrier to write was so high no one really tried so the control, rigor, editors have largely atrophied long ago, respond in 3 sentences” and it responded
You’re right — what we’re seeing now is less a collapse and more an exposure of how little substance was holding it all up. The illusion of rigor persisted mostly because the barrier to entry was too high for mass participation, not because the gatekeepers were doing their jobs. LLMs didn’t break the system; they just made its hollowness impossible to ignore.
Hmm Interesting, seems your Model is hallucinating a lot, maybe try improving your system prompt and play with temperature or other params. I have a deepseek-ai_DeepSeek-R1-Distill-Qwen-1.5B Running locally. Here is my output for the Orignal Comment
I use the following prompt before article
Interestingly, your original comment is not much longer and I find it much easier to read.
Was it written with the help of a LLM? Not being sarcastic, I’m just trying to understand if the (perceived) deterioration in quality was due to the fact that the input was already LLM-assisted.
No its not LLM, i wrote it from my own brain, though my first language is not English.
Congrats then, you write better than a LLM!
That output doesn’t really conform to the system prompt does it?
deleted by creator