Create accountLogin

Recent searches

No recent searches

Search options

Not available on lingo.lol.
lingo.lol is one of the many independent Mastodon servers you can use to participate in the fediverse.
A place for linguists, philologists, and other lovers of languages.

Administered by:

Server stats:

63
active users

lingo.lol: AboutProfiles directoryPrivacy policy

Mastodon: AboutGet the appKeyboard shortcutsView source codev4.3.8

#LLMs

28 posts24 participants0 posts today
Public
Miguel Afonso Caetano

Jason Koebler (404 Media) "We’re not burying our heads in the sand. We use AI tools every day to understand how they work, their limitations, and, most importantly, how they are being leveraged to become the dominant type of content on the internet. Because we do this, we have some of the leading coverage of generative AI on the internet. AI isn’t going away, and I could imagine using it in the future if it becomes more trustworthy and perhaps if the companies pushing it find more ethical business models. I have experimented with using AI to write complex Freedom of Information Act requests and FOIA appeals and to parse large documents, though I haven’t been impressed with the results. I passively use forms of AI to help transcribe interviews, get the gist of YouTube videos in foreign languages, and edit short-form videos and podcasts. Language translation and transcription feel like true game changers, while other AI tools feel like spam machines. I’ll use AI to help find new information, but not to write my words."

cjr.org/feature-2/how-were-usi

Columbia Journalism ReviewHow We’re Using AIThe rapid development of AI is already changing how journalists operate. Reporters, editors, executives, and others across the news industry share their advice on how to engage—and where to draw the line.
#AI#GenerativeAI#Journalism
Continued thread
Public
Brian Gettler

@HalifaxExaminer As a professor, I particularly appreciate this point, which extends to universities around the world:

"The recent push for teched in schools and universities is further threatening the very core of education, while deskilling children and young people, who are losing faith in their own abilities to read, write, and think critically."

It's deeply disturbing how many of our political leaders seem to have drunk the Kool-Aid.

#HigherEd#LLMs#academia
Public
Bibliolater 📚 📜 🖋

**Comparison of Large Language Model with Aphasia**

“_Large language models (LLMs) respond fluently but often inaccurately, which resembles aphasia in humans. Does this behavioral similarity indicate any resemblance in internal information processing between LLMs and aphasic humans?_”

T. Watanabe, K. Inoue, Y. Kuniyoshi, K. Nakajima, K. Aihara, Comparison of Large Language Model with Aphasia. Adv. Sci. 2025, 2414016. doi.org/10.1002/advs.202414016.

#OpenAccess #OA #Research #Article #DOI #AI #ArtificialIntelligence #LLMS #Technology #Tech #Aphasia #Academia #Academic @ai

Public
Mastokarl 🇺🇦

Unexpectedly, science suggests that using ChatGPT for learning doesn't make us all stupid. Quite the opposite.

nature.com/articles/s41599-025

"The results indicate that ChatGPT has a large positive impact on improving learning performance (g = 0.867) and a moderately positive impact on enhancing learning perception (g = 0.456) and fostering higher-order thinking (g = 0.457). "

NatureThe effect of ChatGPT on students’ learning performance, learning perception, and higher-order thinking: insights from a meta-analysis - Humanities and Social Sciences CommunicationsAs a new type of artificial intelligence, ChatGPT is becoming widely used in learning. However, academic consensus regarding its efficacy remains elusive. This study aimed to assess the effectiveness of ChatGPT in improving students’ learning performance, learning perception, and higher-order thinking through a meta-analysis of 51 research studies published between November 2022 and February 2025. The results indicate that ChatGPT has a large positive impact on improving learning performance (g = 0.867) and a moderately positive impact on enhancing learning perception (g = 0.456) and fostering higher-order thinking (g = 0.457). The impact of ChatGPT on learning performance was moderated by type of course (QB = 64.249, P < 0.001), learning model (QB = 76.220, P < 0.001), and duration (QB = 55.998, P < 0.001); its effect on learning perception was moderated by duration (QB = 19.839, P < 0.001); and its influence on the development of higher-order thinking was moderated by type of course (QB = 7.811, P < 0.05) and the role played by ChatGPT (QB = 4.872, P < 0.05). This study suggests that: (1) appropriate learning scaffolds or educational frameworks (e.g., Bloom’s taxonomy) should be provided when using ChatGPT to develop students’ higher-order thinking; (2) the broad use of ChatGPT at various grade levels and in different types of courses should be encouraged to support diverse learning needs; (3) ChatGPT should be actively integrated into different learning modes to enhance student learning, especially in problem-based learning; (4) continuous use of ChatGPT should be ensured to support student learning, with a recommended duration of 4–8 weeks for more stable effects; (5) ChatGPT should be flexibly integrated into teaching as an intelligent tutor, learning partner, and educational tool. Finally, due to the limited sample size for learning perception and higher-order thinking, and the moderately positive effect, future studies with expanded scope should further explore how to use ChatGPT more effectively to cultivate students’ learning perception and higher-order thinking.
#chatgpt#education#llms
Public *
APTiger

[Post modifié et corrigé, la précédente version du post, que je viens de supprimer, renvoyait par erreur d'inattention ou d'étourderie à un mauvais lien web et à un mauvais numéro de DOI] AI hallucinations can’t be stopped — but these techniques can limit their damage |
21 January 2025

Developers have tricks to stop artificial intelligence from making things up, but large language models are still struggling to tell the truth, the whole truth and nothing but the truth.

nature.com/articles/d41586-025

#AI #IA #LLM #LLMs #informatique #Programming #Programmation #Computering #OSINT #ROSO #Renseignement #Intelligence

Je ne vous poste pas l'article complet parce que sinon Springer pas content et Springer pas gentil.

Je vous indique cependant le numéro de DOI pour le récupérer comme vous savez ; mais moi je n'ai pas dit ça, ah non, je n'ai pas dit ça, et après on rigole et on va dire que c'est moi qui l'ai dit… 🤷‍♂️ (comme disait Coluche) :
doi: 10.1038/d41586-025-00068-5

www.nature.comAI hallucinations can’t be stopped — but these techniques can limit their damageDevelopers have tricks to stop artificial intelligence from making things up, but large language models are still struggling to tell the truth, the whole truth and nothing but the truth.
Public
Harald Sack

This week, we were discussing the central question Can we "predict" a word? as the basis for statistical language models in our #ISE2025 lecture. Of course, I wasx trying Shakespeare quotes to motivate the (international) students to complement the quotes with "predicted" missing words ;-)

"All the world's a stage, and all the men and women merely...."

#nlp #llms #languagemodel #Shakespeare #AIart lecture @fiz_karlsruhe @fizise @tabea @enorouzi @sourisnumerique #brushUpYourShakespeare

Slide from the Information Service Engineering 2025 lecture, Natural Language Processing 03, 2.10 Language Models. The Slide shows a graphical portrait of William Shakespeare (created by midjourney AI) as an ink sketch with yellow accents. The text states "Can we "predict" a word?"
Public
Tinker ☀️

Musing: Anti-AI defense

I've been thinking about "hashing" algorithms, specifically around passwords.

The idea is that a hashing algorithm should be slow enough - take a computer a little bit of time to do the computation - to slow down a hash cracking rig but fast enough that it doesn't inconvenience the person entering in their password.

So it might take a half second after typing in the password and having the login screen come up. But that half a second adds up a lot after thousands, millions, and billions of hash cracking attempts.

With that in mind, how can we add artifacts and copy pastes to our web pages / documents / social media posts that would both slow down AI and fill AI with absolute gibberish.

Once we sort out a way to do that, how can we build it in to our technologies so that we aren't manually adding it in each time.

Like... wouldn't it be cool if every Mastodon post was filled with junk comments that weren't shown in the UI but were absolutely part of the source code HTML / ActivityPub. Any AI web scraper would be bogged down and/or filled with junk but the user experience isn't massively changed.

Again - this has to be balanced with user experience. So if the bogdown affects a normal fediverse server, that's bad. (we'll see who read this paragraph)

But if there's a meaningful way of targeting AI specifically and uniquely - that'd be cool.

#AI#artificalIntelligence#LLM
Public
José A. Alonso

Lean Copilot: Large language models as copilots for theorem proving in Lean. ~ Peiyang Song, Kaiyu Yang, Anima Anandkumar. arxiv.org/abs/2404.12534 #LLMs #ITP #LeanProver

arXiv logo
arXiv.orgLean Copilot: Large Language Models as Copilots for Theorem Proving in LeanNeural theorem proving combines large language models (LLMs) with proof assistants such as Lean, where the correctness of formal proofs can be rigorously verified, leaving no room for hallucination. With existing neural theorem provers pretrained on a fixed collection of data and offering valuable suggestions at times, it is challenging for them to continually prove novel theorems in a fully autonomous mode, where human insights may be critical. In this paper, we explore LLMs as copilots that assist humans in proving theorems. We introduce Lean Copilot, a general framework for running LLM inference natively in Lean. It enables programmers to build various LLM-based proof automation tools that integrate seamlessly into the workflow of Lean users. Lean users can use our pretrained models or bring their own ones that run either locally (with or without GPUs) or on the cloud. Using Lean Copilot, we build LLM-based tools that suggest proof steps, complete proof goals, and select relevant premises. Experimental results on the Mathematics in Lean textbook demonstrate the effectiveness of our method compared to existing rule-based proof automation in Lean (aesop). When assisting humans, Lean Copilot requires only 2.08 manually-entered proof steps on average (3.86 required by aesop); when automating the theorem proving process, Lean Copilot automates 74.2% proof steps on average, 85% better than aesop (40.1%). We open source all code and artifacts under a permissive MIT license to facilitate further research.
ExploreLive feeds
About