Discussion
Loading...

#Tag

Log in
  • About
  • Code of conduct
  • Privacy
  • Users
  • Instances
  • About Bonfire
David Adler
David Adler
@perspektivbrocken@social.tchncs.de  ·  activity timestamp 5 days ago

Kennt jemand gutes Bildungsmaterial um mit einer Gruppe selbst ein kleines, sehr simples #LLM zu bauen? Ziel wäre, dir Basics etwas zu entzaubern. Zielgruppe in dem Fall Erwachsene, aber ohne irgendwelche Vorkenntnisse in dem Bereich. Vielleicht @fedilz_infos ?

Andreas Wagner
Andreas Wagner
@anwagnerdreas@hcommons.social replied  ·  activity timestamp 5 days ago

@perspektivbrocken

@thomasrenkert weiß bestimmt etwas zu berichten, der arbeitet in der Lehrerbildung (richtig?) und hat den ParzivAI Assistenten entwickelt: https://agki-dh.github.io/pages/webinar/page-9.html / https://hse.hypotheses.org/6066.

An Raschkas Building LLMs from Scratch hab ich auch schon gedacht, das wurde ja schon erwähnt. Es gibt im Netz Code dazu (https://github.com/rasbt/LLMs-from-scratch) und ein paar Leute, die ihre Erfahrungen damit/darum herum posten, z.B. https://www.gilesthomas.com/llm-from-scratch . https://huggingface.co/blog/gszauer/minimal-llm und https://readmedium.com/how-to-build-an-llm-from-scratch-8c477768f1f9 sind auch gut. Alles Englisch halt.

Wenn es nicht so sehr ums selber bauen sondern auch um Erklärungen geht, sind glaube ich die Videos von Andrej Karpathy (z.B. https://youtu.be/7xTGNNLPyMI?si=-YlKsMGuBnW5GO44 oder https://youtu.be/zduSFxRajkE?si=M-cYkZVX6N8M-Oxy ) oder von Thomas Wolf (https://youtu.be/2-SPH9hIKT8?si=IV-hyIaYCYi-dQTZ) ganz gut. https://youtu.be/LPZh9BOjkQs?si=uxW_igUPN92C6wg2 auch.

Und ich finde ja so "Simulationen" auch sehr hilfreich: https://www.soekia.ch/gpt.html / https://bbycroft.net/llm / https://poloclub.github.io/transformer-explainer/

#LLM #AI #Explainer #Transformer #GPT #BERT

Transformer Explainer: LLM Transformer Model Visually Explained

An interactive visualization tool showing you how transformer models work in large language models (LLM) like GPT.

LLM Visualization

A 3D animated visualization of an LLM with a walkthrough.

SoekiaGPT - Das didaktische Sprachmodell

SoekiaGPT ist ein Textgenerator speziell für den Unterricht. Mit SoekiaGPT kannst Du hinter die Kulissen schauen und damit einige Grundprinzipien von Textgeneratoren wie ChatGPT kennenlernen.

How to Build an LLM from Scratch

Data Curation, Transformers, Training at Scale, and Model Evaluation

Build a simple LLM from scratch

A Blog post by Gabor Szauer on Hugging Face

LLM from scratch :: Giles' blog

(
GitHub

GitHub - rasbt/LLMs-from-scratch: Implement a ChatGPT-like LLM in PyTorch from scratch, step by step

Implement a ChatGPT-like LLM in PyTorch from scratch, step by step - rasbt/LLMs-from-scratch
Fokus Lehrerbildung

KI Showcase: Der Chatbot “ParzivAI”

Am 22.07.24 stellten Dr. Florian Nieser und Dr. Thomas Renkert die Alpha-Version des von ihnen entwickelten Chatbots ParzivAI vor. ParzivAI ist auf die Vermittlung der mittelhochdeutschen Sprache und Geschichtsdidaktik des Mittelalters spezialisiert und besteht aus einem Sprachmodell (LLM), einem RAG-System für Dokumentenzugriff, sowie einem Frontend mit Feedback-Funktion

DHd-AG Angewandte Generative KI in den Digitalen Geisteswissenschaften (AGKI-DH)

The Applied Generative AI in Digital Humanities (AGKI-DH) working group focuses on exploring and showcasing the applications of generative AI, especially Large Language Models (LLMs), in digital humanities. This includes both commercial and open-source models, recognizing the significant advancements and the widespread discussion on AI issues like bias, hallucination, high computational footprint, data security, and potential destructive capabilities. Amidst these challenges, the group aims to harness the opportunities in AI development for digital humanities, fostering an ongoing exchange on core AI concepts and practical experiences. It seeks collaboration with other groups on greening (AI footprint), digital humanities theory, and empowerment (ethical aspects). The goals include identifying, assessing, and categorizing generative AI applications in digital humanities, establishing best practices, developing evaluation criteria, facilitating discussions and knowledge exchange, and balancing opportunities and risks.
  • Copy link
  • Flag this comment
  • Block
UniversityofGroningenLibrary
UniversityofGroningenLibrary
@Bibliothecaris@social.edu.nl  ·  activity timestamp 2 weeks ago

New @universityofgroningen #OpenAccess publication in the spotlight:

➡️ The price of automated case #law annotation: comparing the cost and performance of #GPT-4o and student annotators

🔗 https://doi.org/10.1007/s10506-025-09495-1

Read our interview with corresponding author Iris Schepers:

🔗 https://www.rug.nl/library/open-access/blog/open-access-publication-in-the-spotlight-the-price-of-automated-case-law-annotation-comparing-th

#AI #OpenScience

printscreen publication
printscreen publication
printscreen publication
  • Copy link
  • Flag this post
  • Block
Nicolas Fressengeas boosted
Stéphane Pouyllau
Stéphane Pouyllau
@pouyllau@mastodon.social  ·  activity timestamp 3 weeks ago

#IA « Publicité dans ChatGPT : Pourquoi OpenAI n’a pas le choix, et qu’est-ce que cela implique pour les données des utilisateurs » https://www.usine-digitale.fr/intelligence-artificielle/openai/publicite-dans-chatgpt-pourquoi-openai-na-pas-le-choix-et-quest-ce-que-ca-implique-pour-les-donnees-des-utilisateurs.CNGAHWZBCFATNF3WNL57M37VHY.html #ChatGPT #GPT

L'Usine Digitale

Publicité dans ChatGPT : Pourquoi OpenAI n’a pas le choix, et qu’est-ce que cela implique pour les données des utilisateurs

A la recherche de revenus, OpenAI va introduire la publicité dans ChatGPT. Alors concrètement, qu’est-ce que cela va changer ? On vous explique ce à quoi ressemblera le service d’IA dans les prochaines semaines.
  • Copy link
  • Flag this post
  • Block
Stéphane Pouyllau
Stéphane Pouyllau
@pouyllau@mastodon.social  ·  activity timestamp 3 weeks ago

#IA « Publicité dans ChatGPT : Pourquoi OpenAI n’a pas le choix, et qu’est-ce que cela implique pour les données des utilisateurs » https://www.usine-digitale.fr/intelligence-artificielle/openai/publicite-dans-chatgpt-pourquoi-openai-na-pas-le-choix-et-quest-ce-que-ca-implique-pour-les-donnees-des-utilisateurs.CNGAHWZBCFATNF3WNL57M37VHY.html #ChatGPT #GPT

L'Usine Digitale

Publicité dans ChatGPT : Pourquoi OpenAI n’a pas le choix, et qu’est-ce que cela implique pour les données des utilisateurs

A la recherche de revenus, OpenAI va introduire la publicité dans ChatGPT. Alors concrètement, qu’est-ce que cela va changer ? On vous explique ce à quoi ressemblera le service d’IA dans les prochaines semaines.
  • Copy link
  • Flag this post
  • Block
Morgenmuffel
Morgenmuffel
@chanakya@social.screamingatmyscreen.com  ·  activity timestamp 3 months ago

GPT = Garbage Producing Technology.

#ai #llm #chatgpt #news

Sorry, no caption provided by author
Sorry, no caption provided by author
Sorry, no caption provided by author
Morgenmuffel
Morgenmuffel
@chanakya@social.screamingatmyscreen.com replied  ·  activity timestamp 3 months ago

The newspaper is now "investigating" use of AI
https://www.dawn.com/news/1954574/auto-sales-rev-up-in-october

#llm #ai #gpt #news

Sorry, no caption provided by author
Sorry, no caption provided by author
Sorry, no caption provided by author
  • Copy link
  • Flag this comment
  • Block
UKP Lab
UKP Lab
@UKPLab@sigmoid.social  ·  activity timestamp 3 months ago

𝗧𝗶𝗿𝗲𝗱 𝗼𝗳 𝘆𝗼𝘂𝗿 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁 𝗺𝗮𝗸𝗶𝗻𝗴 𝗺𝗮𝗻𝘆 𝗲𝗿𝗿𝗼𝗿𝘀?
➡️ We’ve got the solution!

Meet 𝗦𝗘𝗘𝗘𝗗 🌱 — a framework for 𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗲𝗱 𝗘𝗿𝗿𝗼𝗿 𝗗𝗶𝘀𝗰𝗼𝘃𝗲𝗿𝘆 in conversational AI.

🧩 SEEED detects both 𝗸𝗻𝗼𝘄𝗻 𝗮𝗻𝗱 𝗽𝗿𝗲𝘃𝗶𝗼𝘂𝘀𝗹𝘆 𝘂𝗻𝘀𝗲𝗲𝗻 𝗲𝗿𝗿𝗼𝗿 𝘁𝘆𝗽𝗲𝘀, and even 𝗴𝗲𝗻𝗲𝗿𝗮𝘁𝗲𝘀 𝗱𝗲𝗳𝗶𝗻𝗶𝘁𝗶𝗼𝗻𝘀 for newly discovered ones

⚙️ By combining 𝗹𝗶𝗴𝗵𝘁𝘄𝗲𝗶𝗴𝗵𝘁 𝗲𝗻𝗰𝗼𝗱𝗲𝗿𝘀 with a 𝗻𝗼𝘃𝗲𝗹 𝘀𝗮𝗺𝗽𝗹𝗶𝗻𝗴 𝘀𝘁𝗿𝗮𝘁𝗲𝗴𝘆 𝗳𝗼𝗿 𝗰𝗼𝗻𝘁𝗿𝗮𝘀𝘁𝗶𝘃𝗲 𝗹𝗲𝗮𝗿𝗻𝗶𝗻𝗴, it improves representation learning and uncovers 𝗰𝗼𝗵𝗲𝗿𝗲𝗻𝘁 𝗲𝗿𝗿𝗼𝗿 𝗰𝗮𝘁𝗲𝗴𝗼𝗿𝗶𝗲𝘀.

(1/🧵 )

Diagram illustrating a chatbot correction process. A human says, “I really like indie music! Do you have a favorite artist?” The chatbot replies, “I’m a huge fan of indie music too! The Beatles are my absolute favorite!” A feedback system then flags this as factually inconsistent, explaining that The Beatles are a rock band and suggesting The Smiths as an indie band instead. The corrected response becomes, “I’m a huge fan of indie music too! The Smiths are my absolute favorite!” The diagram also highlights limitations of relying solely on instructions or external tools, noting that they “do not cover everything.”
Diagram illustrating a chatbot correction process. A human says, “I really like indie music! Do you have a favorite artist?” The chatbot replies, “I’m a huge fan of indie music too! The Beatles are my absolute favorite!” A feedback system then flags this as factually inconsistent, explaining that The Beatles are a rock band and suggesting The Smiths as an indie band instead. The corrected response becomes, “I’m a huge fan of indie music too! The Smiths are my absolute favorite!” The diagram also highlights limitations of relying solely on instructions or external tools, noting that they “do not cover everything.”
Diagram illustrating a chatbot correction process. A human says, “I really like indie music! Do you have a favorite artist?” The chatbot replies, “I’m a huge fan of indie music too! The Beatles are my absolute favorite!” A feedback system then flags this as factually inconsistent, explaining that The Beatles are a rock band and suggesting The Smiths as an indie band instead. The corrected response becomes, “I’m a huge fan of indie music too! The Smiths are my absolute favorite!” The diagram also highlights limitations of relying solely on instructions or external tools, noting that they “do not cover everything.”
UKP Lab
UKP Lab
@UKPLab@sigmoid.social replied  ·  activity timestamp 3 months ago

📊 𝗦𝗘𝗘𝗘𝗗 outperforms #GPT-4o and #Phi-4 by up to +𝟴 𝗽𝗽 across multiple datasets.

📄 𝗣𝗮𝗽𝗲𝗿: https://www.arxiv.org/abs/2509.10833
💻 𝗖𝗼𝗱𝗲: https://github.com/UKPLab/emnlp2025-automatic-error-discovery
🔗 𝗣𝗿𝗼𝗷𝗲𝗰𝘁: https://ukplab.github.io/emnlp2025-automatic-error-discovery/

Be sure to follow the authors: Dominic Petrak, Thy Thy Tran, and Iryna Gurevych from Ubiquitous Knowledge Processing (UKP) Lab/Technische Universität Darmstadt.

See you at the #EMNLP in Suzhou!

(2/2)

#NLProc #ConversationalAI #Agents #EMNLP2025

  • Copy link
  • Flag this comment
  • Block
Ulrike Hahn boosted
infoDOCKET
infoDOCKET
@infodocket@newsie.social  ·  activity timestamp 5 months ago

New Report From #NewsGuard: #AI False Information Rate Nearly Doubles in One Year https://www.newsguardtech.com/ai-monitor/august-2025-ai-false-claim-monitor/#LLMs#GPT

Sorry, no caption provided by author
Sorry, no caption provided by author
Sorry, no caption provided by author
  • Copy link
  • Flag this post
  • Block
infoDOCKET
infoDOCKET
@infodocket@newsie.social  ·  activity timestamp 5 months ago

New Report From #NewsGuard: #AI False Information Rate Nearly Doubles in One Year https://www.newsguardtech.com/ai-monitor/august-2025-ai-false-claim-monitor/#LLMs#GPT

Sorry, no caption provided by author
Sorry, no caption provided by author
Sorry, no caption provided by author
  • Copy link
  • Flag this post
  • Block
Domenico De Treias boosted
TiTiNoNero :__:
TiTiNoNero :__:
@77nn@goto.77nn.it  ·  activity timestamp 7 months ago

Bugs, please!

Nuovo episodio del #podcast, dopo una lunga pausa.

https://77nn.it/podcasts/S03-E04.html

Sento sempre più spesso, da amici e conoscenti, racconti di aziende che incoraggiano i loro dipendenti a utilizzare per lavoro i chatbot alimentati dai Large Language Model.
Altre aziende, piuttosto che incoraggiare, obbligano a questa pratica, impedendone la disabilitazione sui propri computer e telefoni aziendali. Ma il tutto è un legittimo interesse del datore di lavoro, o così viene spiegato...

#podcastITA #ita #AI #LLM #GPT

  • Copy link
  • Flag this post
  • Block
TiTiNoNero :__:
TiTiNoNero :__:
@77nn@goto.77nn.it  ·  activity timestamp 7 months ago

Bugs, please!

Nuovo episodio del #podcast, dopo una lunga pausa.

https://77nn.it/podcasts/S03-E04.html

Sento sempre più spesso, da amici e conoscenti, racconti di aziende che incoraggiano i loro dipendenti a utilizzare per lavoro i chatbot alimentati dai Large Language Model.
Altre aziende, piuttosto che incoraggiare, obbligano a questa pratica, impedendone la disabilitazione sui propri computer e telefoni aziendali. Ma il tutto è un legittimo interesse del datore di lavoro, o così viene spiegato...

#podcastITA #ita #AI #LLM #GPT

  • Copy link
  • Flag this post
  • Block
Erik Kemp 🇪🇺 boosted
iBestuur
iBestuur
@ibestuur@mastodon.nl  ·  activity timestamp 7 months ago

Goed nieuws voor #GPT-NL. De leden van NDP Nieuwsmedia stellen een omvangrijk deel van het archief van nieuwsartikelen van meer dan 30 landelijke en regionale nieuwstitels beschikbaar om het taalmodel verder te trainen. Met de uitgevers is afgesproken dat ze een passende vergoeding ontvangen wanneer GPT-NL commercieel wordt ingezet. @SURF @TNOhttps://ibestuur.nl/artikel/nederlandse-uitgevers-leveren-grote-verzameling-artikelen-voor-training-gpt-nl/

  • Copy link
  • Flag this post
  • Block
iBestuur
iBestuur
@ibestuur@mastodon.nl  ·  activity timestamp 7 months ago

Goed nieuws voor #GPT-NL. De leden van NDP Nieuwsmedia stellen een omvangrijk deel van het archief van nieuwsartikelen van meer dan 30 landelijke en regionale nieuwstitels beschikbaar om het taalmodel verder te trainen. Met de uitgevers is afgesproken dat ze een passende vergoeding ontvangen wanneer GPT-NL commercieel wordt ingezet. @SURF @TNOhttps://ibestuur.nl/artikel/nederlandse-uitgevers-leveren-grote-verzameling-artikelen-voor-training-gpt-nl/

  • Copy link
  • Flag this post
  • Block

bonfire.cafe

A space for Bonfire maintainers and contributors to communicate

bonfire.cafe: About · Code of conduct · Privacy · Users · Instances
Bonfire social · 1.0.2-alpha.22 no JS en
Automatic federation enabled
Log in
  • Explore
  • About
  • Members
  • Code of Conduct