This morning, I opened a PDF to find Adobe Acrobat giving me an AI-generated summary of it. That led to a rabbit hole of finding all the places Acrobat’s AI Assistant has to be turned off. Here they are: www.wordsbywes.ink/a...
#LLMs #AmEditing
#LLMs #AmEditing
Turning Off Adobe Acrobat's AI Assistant - Words by Wes
This morning, I opened a PDF to find Acrobat giving me an AI-generated summary of it. Since I frequently deal with unpublished manuscripts and other confidential documents for clients, I can’t have them processed by generative AI tools (and I don’t want my PDFs being read by LLMs either). So…
www.wordsbywes.ink
April 30, 2024 at 7:51 PM
This morning, I opened a PDF to find Adobe Acrobat giving me an AI-generated summary of it. That led to a rabbit hole of finding all the places Acrobat’s AI Assistant has to be turned off. Here they are: www.wordsbywes.ink/a...
#LLMs #AmEditing
#LLMs #AmEditing
She’s right. Billions of computations to output just a single token. Then the whole process is repeated to produce another token, and so on. Biological systems, on the other hand, employ continuous processes that operate in real time using a tiny fraction of the energy. #LLMs #GenAI
September 21, 2026 at 4:53 PM
OpenAI’s so-called “reasoning” models generate more falsehoods than their predecessors. Since #LLMs are model language usage and not factual information, it’s built-into their design that they will output false or misleading information. #PhilAI techcrunch.com/2025/04/18/o...
April 19, 2025 at 9:52 AM
OpenAI’s so-called “reasoning” models generate more falsehoods than their predecessors. Since #LLMs are model language usage and not factual information, it’s built-into their design that they will output false or misleading information. #PhilAI techcrunch.com/2025/04/18/o...
#LLMs feel *exactly* like crypto did in 2017, with nearly daily articles about how it can't possibly work, and a die hard community earnestly pleading "but you just don't UNDERSTAND!"
The main difference is that there *are* reasonable use cases. They're just far smaller than people want to admit.
The main difference is that there *are* reasonable use cases. They're just far smaller than people want to admit.
February 24, 2025 at 2:36 PM
#LLMs feel *exactly* like crypto did in 2017, with nearly daily articles about how it can't possibly work, and a die hard community earnestly pleading "but you just don't UNDERSTAND!"
The main difference is that there *are* reasonable use cases. They're just far smaller than people want to admit.
The main difference is that there *are* reasonable use cases. They're just far smaller than people want to admit.
January 19, 2025 at 4:10 PM
Anthropic & other AI companies are spending billions of dollars to slice the spines off millions of books, including very rare ones.
Books that survived wars, fires, and centuries of handling are being shredded so an AI can learn to write a better marketing email.
#AI #LLMs #Books #Culture #Tech
Books that survived wars, fires, and centuries of handling are being shredded so an AI can learn to write a better marketing email.
#AI #LLMs #Books #Culture #Tech
July 29, 2026 at 3:45 PM
A review of the proceedings from four major computer-science conferences showed that none from 2021, and all from 2025, had fake citations.
arxiv.org/abs/2602.058...
#AI #LLMs #Hallucinations #Misconduct #ScholComm
arxiv.org/abs/2602.058...
#AI #LLMs #Hallucinations #Misconduct #ScholComm
The Case of the Mysterious Citations
Mysterious citations are routinely appearing in peer-reviewed publications throughout the scientific community. In this paper, we developed an automated pipeline and examine the proceedings of four ma...
arxiv.org
February 11, 2026 at 2:22 PM
A review of the proceedings from four major computer-science conferences showed that none from 2021, and all from 2025, had fake citations.
arxiv.org/abs/2602.058...
#AI #LLMs #Hallucinations #Misconduct #ScholComm
arxiv.org/abs/2602.058...
#AI #LLMs #Hallucinations #Misconduct #ScholComm
"To invoke #language when talking about #LLMs is to misunderstand the nature of language and miss its fundamentally lived and embodied character ... Where #AI promises magic, the most pressing need is for basic research driven by communities"
A must-read: www.thedial.world/articles/new... #langsky
A must-read: www.thedial.world/articles/new... #langsky
AI Won’t Protect Endangered Languages — The Dial
Combinations of characters on a screen mean nothing without agency and intention.
www.thedial.world
January 3, 2025 at 6:32 PM
"To invoke #language when talking about #LLMs is to misunderstand the nature of language and miss its fundamentally lived and embodied character ... Where #AI promises magic, the most pressing need is for basic research driven by communities"
A must-read: www.thedial.world/articles/new... #langsky
A must-read: www.thedial.world/articles/new... #langsky
Ok, here is an incomplete list of things we don't entirely know about #llms:
+ what their training data is
+ which tokens the attention mechanism pays attention to
+ how exactly quantization and compression impact them
+ why slight changes in prompts impact the results so much
....
???
+ what their training data is
+ which tokens the attention mechanism pays attention to
+ how exactly quantization and compression impact them
+ why slight changes in prompts impact the results so much
....
???
October 19, 2024 at 12:06 AM
Ok, here is an incomplete list of things we don't entirely know about #llms:
+ what their training data is
+ which tokens the attention mechanism pays attention to
+ how exactly quantization and compression impact them
+ why slight changes in prompts impact the results so much
....
???
+ what their training data is
+ which tokens the attention mechanism pays attention to
+ how exactly quantization and compression impact them
+ why slight changes in prompts impact the results so much
....
???
#AI #CriticalThinking #AIHallucinations #LLMs
Things to keep in mind, when you use AI...
1. Watch @charalanahzard.bsky.social's video (and please watch it all the way to the end).
2. Then, if you wanna dig deeper, go read @richardcarrier.info's article.
Things to keep in mind, when you use AI...
1. Watch @charalanahzard.bsky.social's video (and please watch it all the way to the end).
2. Then, if you wanna dig deeper, go read @richardcarrier.info's article.
December 23, 2025 at 12:46 AM
#AI #CriticalThinking #AIHallucinations #LLMs
Things to keep in mind, when you use AI...
1. Watch @charalanahzard.bsky.social's video (and please watch it all the way to the end).
2. Then, if you wanna dig deeper, go read @richardcarrier.info's article.
Things to keep in mind, when you use AI...
1. Watch @charalanahzard.bsky.social's video (and please watch it all the way to the end).
2. Then, if you wanna dig deeper, go read @richardcarrier.info's article.
What Abeba says! A report by academics from Memorial University, Newfoundland, on, amongst other things, the ethical use of #AI in education, contained fake references and seems likely to have been created using #LLMs ...
www.cbc.ca/news/canada/...
www.cbc.ca/news/canada/...
September 12, 2025 at 11:30 PM
What Abeba says! A report by academics from Memorial University, Newfoundland, on, amongst other things, the ethical use of #AI in education, contained fake references and seems likely to have been created using #LLMs ...
www.cbc.ca/news/canada/...
www.cbc.ca/news/canada/...
ICYMI: "AI generates covertly racist decisions about
people based on their dialect" Hofman, et al., 2024. Nature.
www.nature.com/articles/s41...
#linguistics #sociolinguistics #computational #AI #accentism #racism #prejudice #stereotypes #bias #LLMs #science
people based on their dialect" Hofman, et al., 2024. Nature.
www.nature.com/articles/s41...
#linguistics #sociolinguistics #computational #AI #accentism #racism #prejudice #stereotypes #bias #LLMs #science
AI generates covertly racist decisions about people based on their dialect - Nature
Despite efforts to remove overt racial prejudice, language models using artificial intelligence still show covert racism against speakers of African American English that is triggered by features...
www.nature.com
October 6, 2024 at 5:09 PM
ICYMI: "AI generates covertly racist decisions about
people based on their dialect" Hofman, et al., 2024. Nature.
www.nature.com/articles/s41...
#linguistics #sociolinguistics #computational #AI #accentism #racism #prejudice #stereotypes #bias #LLMs #science
people based on their dialect" Hofman, et al., 2024. Nature.
www.nature.com/articles/s41...
#linguistics #sociolinguistics #computational #AI #accentism #racism #prejudice #stereotypes #bias #LLMs #science
Large language models are less effective at clinical prediction tasks than locally trained machine learning models academic.oup.com/jamia/advanc... #llms #machinelearning #MLSky (not surprising but important to document)
Large language models are less effective at clinical prediction tasks than locally trained machine learning models
AbstractObjectives. To determine the extent to which current large language models (LLMs) can serve as substitutes for traditional machine learning (ML) as
academic.oup.com
April 8, 2025 at 6:17 PM
Large language models are less effective at clinical prediction tasks than locally trained machine learning models academic.oup.com/jamia/advanc... #llms #machinelearning #MLSky (not surprising but important to document)
My commentary on the do's and don'ts of cognitive evaluations in LLMs is now out in Nature Human Behavior:
doi.org/10.1038/s415...
posting here with a figure that didn't make it into the final draft and is now instead a boring table :P
#CogSci #LLMs #AI
doi.org/10.1038/s415...
posting here with a figure that didn't make it into the final draft and is now instead a boring table :P
#CogSci #LLMs #AI
January 16, 2025 at 10:26 PM
My commentary on the do's and don'ts of cognitive evaluations in LLMs is now out in Nature Human Behavior:
doi.org/10.1038/s415...
posting here with a figure that didn't make it into the final draft and is now instead a boring table :P
#CogSci #LLMs #AI
doi.org/10.1038/s415...
posting here with a figure that didn't make it into the final draft and is now instead a boring table :P
#CogSci #LLMs #AI
I had the pleasure to teach a 3-day crash course on #LLMs for PhD students.
We covered:
1️⃣ Text Representation and Embeddings
2️⃣ Machine Learning & Transformer Architecture
3️⃣ Generative Models for Social Sciences
The 6 slide sets and 10 notebooks are available on github: github.com/nicolaiberk/...
We covered:
1️⃣ Text Representation and Embeddings
2️⃣ Machine Learning & Transformer Architecture
3️⃣ Generative Models for Social Sciences
The 6 slide sets and 10 notebooks are available on github: github.com/nicolaiberk/...
GitHub - nicolaiberk/llm_ws: Materials for my Workshop on LLMs
Materials for my Workshop on LLMs. Contribute to nicolaiberk/llm_ws development by creating an account on GitHub.
github.com
September 29, 2025 at 1:38 PM
I had the pleasure to teach a 3-day crash course on #LLMs for PhD students.
We covered:
1️⃣ Text Representation and Embeddings
2️⃣ Machine Learning & Transformer Architecture
3️⃣ Generative Models for Social Sciences
The 6 slide sets and 10 notebooks are available on github: github.com/nicolaiberk/...
We covered:
1️⃣ Text Representation and Embeddings
2️⃣ Machine Learning & Transformer Architecture
3️⃣ Generative Models for Social Sciences
The 6 slide sets and 10 notebooks are available on github: github.com/nicolaiberk/...
The difference between #AI errors and confabulation in humans is that in producing false or misleading information, the LLM is functioning precisely as intended. Nothing has gone wrong because nothing in the mechanism regulates alignment with the truth. We might as well say #LLMs always confabulate…
"It would be more appropriate to say that [LLMs] ‘confabulate’. In humans, confabulation involves the unintentional – and usually linguistic – fabrication of content: making stuff up without realising that one is doing so. It is primarily about *doing*, rather than *experiencing*."
Conscious artificial intelligence and biological naturalism | Behavioral and Brain Sciences | Cambridge Core
Conscious artificial intelligence and biological naturalism
www.cambridge.org
April 23, 2025 at 3:17 AM
The difference between #AI errors and confabulation in humans is that in producing false or misleading information, the LLM is functioning precisely as intended. Nothing has gone wrong because nothing in the mechanism regulates alignment with the truth. We might as well say #LLMs always confabulate…
#LLMs are reshaping how people #code and #software. But what are the #security trade-offs? Get a clear-eyed assessment from this #openSUSE Conference 2026. #Linux #AI youtu.be/33jpOZ0-aJM?...
openSUSE Conference 2026 - LLMs: The good, the bad and the ugly from a security POV
YouTube video by openSUSE
youtu.be
September 23, 2026 at 12:02 PM
Quantifying memory recall is hard! Luckily, natural language processing (incl. #LLMs) offers new, automated, and scalable ways to do that!
Great new review by Fenerci & @signysheldon.bsky.social in @cp-trendscognsci.bsky.social!
www.cell.com/trends/cogni...
Great new review by Fenerci & @signysheldon.bsky.social in @cp-trendscognsci.bsky.social!
www.cell.com/trends/cogni...
Studying memory narratives with natural language processing
Cognitive neuroscience research has begun to use natural language processing (NLP)
to examine memory narratives with the hopes of gaining a nuanced understanding of
the mechanisms underlying differenc...
www.cell.com
August 28, 2025 at 1:35 PM
Quantifying memory recall is hard! Luckily, natural language processing (incl. #LLMs) offers new, automated, and scalable ways to do that!
Great new review by Fenerci & @signysheldon.bsky.social in @cp-trendscognsci.bsky.social!
www.cell.com/trends/cogni...
Great new review by Fenerci & @signysheldon.bsky.social in @cp-trendscognsci.bsky.social!
www.cell.com/trends/cogni...
Welcome Bluesky followers!
A 🧵 about my JMP: Using language to generate hypotheses
This paper explores how language shapes behavior. Our contribution, however, is not in testing specific hypotheses- its in generating them (using #LLMs + #ML + #BehSci)
But how exactly?
www.rafaelmbatista.com/jmp/
A 🧵 about my JMP: Using language to generate hypotheses
This paper explores how language shapes behavior. Our contribution, however, is not in testing specific hypotheses- its in generating them (using #LLMs + #ML + #BehSci)
But how exactly?
www.rafaelmbatista.com/jmp/
November 15, 2024 at 4:50 PM
Welcome Bluesky followers!
A 🧵 about my JMP: Using language to generate hypotheses
This paper explores how language shapes behavior. Our contribution, however, is not in testing specific hypotheses- its in generating them (using #LLMs + #ML + #BehSci)
But how exactly?
www.rafaelmbatista.com/jmp/
A 🧵 about my JMP: Using language to generate hypotheses
This paper explores how language shapes behavior. Our contribution, however, is not in testing specific hypotheses- its in generating them (using #LLMs + #ML + #BehSci)
But how exactly?
www.rafaelmbatista.com/jmp/
#LLMs do not "reason" or "think"... do not buy the marketing / VC-funds-seeking bullshit. #AI www.theguardian.com/technology/2...
OpenAI says the latest ChatGPT can ‘think’ – and I have thoughts
The AI company says its ‘o1’ model is capable of reason, a key blocker in the way of truly gamechanging artificial intelligence – but is it that simple?
www.theguardian.com
September 17, 2024 at 11:12 AM
#LLMs do not "reason" or "think"... do not buy the marketing / VC-funds-seeking bullshit. #AI www.theguardian.com/technology/2...
What happens if we tokenize cat as [ca, t] rather than [cat]?
LLMs are trained on just one tokenization per word, but they still understand alternative tokenizations. We show that this can be exploited to bypass safety filters without changing the text itself.
#AI #LLMs #tokenization #alignment
LLMs are trained on just one tokenization per word, but they still understand alternative tokenizations. We show that this can be exploited to bypass safety filters without changing the text itself.
#AI #LLMs #tokenization #alignment
March 11, 2025 at 11:13 PM
What happens if we tokenize cat as [ca, t] rather than [cat]?
LLMs are trained on just one tokenization per word, but they still understand alternative tokenizations. We show that this can be exploited to bypass safety filters without changing the text itself.
#AI #LLMs #tokenization #alignment
LLMs are trained on just one tokenization per word, but they still understand alternative tokenizations. We show that this can be exploited to bypass safety filters without changing the text itself.
#AI #LLMs #tokenization #alignment
Research shows scaling up makes #LLMs both worse and more unsupervisable. But you can't "shift" them "towards reliability" as the authors recommend. We need to counter the hype and control these bullshit machines: they are poisoning information ecosystems and accelerating the #climatecrisis. #AI
Scaling up and shaping up large language models increased their tendency to provide sensible yet incorrect answers at difficulty levels humans cannot supervise, highlighting the need for a shift in AI design towards reliability, according to a paper in Nature. go.nature.com/4eCAnis 🧪
October 1, 2024 at 9:47 AM
Research shows scaling up makes #LLMs both worse and more unsupervisable. But you can't "shift" them "towards reliability" as the authors recommend. We need to counter the hype and control these bullshit machines: they are poisoning information ecosystems and accelerating the #climatecrisis. #AI
📣 Super excited to organize the first workshop on ✨NLP for Democracy✨ at COLM @colmweb.org!!
Check out our website: sites.google.com/andrew.cmu.e...
Call for submissions (extended abstracts) due June 19, 11:59pm AoE
#COLM2025 #LLMs #NLP #NLProc #ComputationalSocialScience
Check out our website: sites.google.com/andrew.cmu.e...
Call for submissions (extended abstracts) due June 19, 11:59pm AoE
#COLM2025 #LLMs #NLP #NLProc #ComputationalSocialScience
NLP 4 Democracy - COLM 2025
sites.google.com
May 21, 2025 at 4:39 PM
📣 Super excited to organize the first workshop on ✨NLP for Democracy✨ at COLM @colmweb.org!!
Check out our website: sites.google.com/andrew.cmu.e...
Call for submissions (extended abstracts) due June 19, 11:59pm AoE
#COLM2025 #LLMs #NLP #NLProc #ComputationalSocialScience
Check out our website: sites.google.com/andrew.cmu.e...
Call for submissions (extended abstracts) due June 19, 11:59pm AoE
#COLM2025 #LLMs #NLP #NLProc #ComputationalSocialScience
RATE IT! #keadon #postmalone #music #jvke #juicewrld #trending #viralshorts #timmins #NowPlaying #art #MusicSky #photography
#Hot21Radio
#HipHop
#RnB
#WeTheBestHipHopandRnB
#Disability
#MCM
#vtuber
#Europe
#LLMs
#6music
#booksky
#Canada
#nightflight
#MusicTherapy
#bluesky
#Bandcamp
#music
#Hot21Radio
#HipHop
#RnB
#WeTheBestHipHopandRnB
#Disability
#MCM
#vtuber
#Europe
#LLMs
#6music
#booksky
#Canada
#nightflight
#MusicTherapy
#bluesky
#Bandcamp
#music
RATE IT! #keadon #postmalone #music #jvke #juicewrld #trending #viralshorts #timmins
YouTube video by Keadon
www.youtube.com
September 29, 2025 at 8:19 AM