π§΅7/7 Image Credits: Curiosamente alucinante
06.02.2025 14:16 β π 0 π 0 π¬ 0 π 0
π§΅ 6/7
What other analogies can you draw between the book and LLMs? Think about learned embeddings, attention mechanisms, etc. Let's discuss! π #Analogy #TechTalk
06.02.2025 14:16 β π 0 π 0 π¬ 1 π 0
π§΅ 5/7
Now for the techy bit: The core principle at play is compositionality. The book combines lines to create poems, LLMs combine words, sub-word units, and learned representations to generate text. This combined with massive datasets and complex architectures, enables diverse and novel outputs.
06.02.2025 14:16 β π 0 π 0 π¬ 1 π 0
π§΅4/7
So next time you're amazed by an LLM's writing skills, remember Cent mille milliards de poΓ¨mes. It's a beautiful illustration of how a little creativity and a LOT of combinations can go a LONG way. πβ‘οΈπ» #RaymondQueneau #CentMilleMilliardsDePoemes
06.02.2025 14:16 β π 0 π 0 π¬ 1 π 0
π§΅3/7
Just like the book can create an almost infinite number of poems, LLMs can generate a mind-boggling variety of text. π€― From poems to code, emails to essays, it's all about combining the building blocks in new and creative ways. β¨ #ChatGPT #deepseek #gemini #claude
06.02.2025 14:16 β π 0 π 0 π¬ 1 π 0
π§΅ 2/7
Think of it like this: each line in the book is like a word in an LLM's vocabulary. π€ The book's structure? That's like the LLM's grammar rules. βοΈ And every new poem you create? That's like the LLM generating a unique sentence or story! βοΈ #Creativity #Innovation
06.02.2025 14:16 β π 0 π 0 π¬ 1 π 0
π€― One book 10 pages. 100 TRILLION poems. And how it helps us understand LLMs a π§΅ π€―
Cent mille milliards de poΓ¨mes, is a real book by Raymond Queneau! It lets you create 100,000,000,000,000 different sonnets just by mixing and matching lines! π€― #LLM #AI #Poetry #Books #MindBlown
π§΅ 1/7
06.02.2025 14:16 β π 6 π 1 π¬ 1 π 0
Sotirios Tsaftaris
Professor Sotirios (Sotos) Tsaftaris, is Chair in Machine Learning and Computer Vision. He also holds the Canon Medical/Royal Academy of Engineering Research Chair in Healthcare AI and is an ELLIS Fel...
π Excited to see our Director, @stsaftaris.bsky.social, featured in @edinnovations.bsky.social "Inspirational Innovators" series!
A fascinating glimpse into the experiences and inspirations shaping his journey in AI and innovation.
π‘ Read more: eil.ac/iisotiriosts...
#AILeadership #Innovation #EI
15.01.2025 14:26 β π 5 π 1 π¬ 0 π 0
Scotlandβs second largest health board investing Β£1.6B a year to protect & improve health in Lothians
Professor of Surgery & Data Science, Edinburgh.
AI / Machine learning / Data Science / Global Surgery
Co-Head Centre for Medical Informatics
Liver and pancreas surgeon (HPB)
R for Health Data Science author
http://finalfit.org
@GlobalSurg
Professor in Law and Technology @University of Sussex
https://profiles.sussex.ac.uk/p305243-phoebe-li/about
Prof in Philosophy and History of Science & Technology, Technical University of Munich | scientific director Ethical Data Initiative & Public Science Lab | http://opensciencestudies.eu | President ISHPSSB | loves almost all forms of life
History/philosophy/social studies of life sciences.
Research Fellow, Causality in Healthcare AI Hub, Egenis, University of Exeter - www.chai.ac.uk
Co-author, 'A History of Genomics across Species, Communities and Projects' - https://tinyurl.com/histgen
The Causality in Healthcare AI Hub (CHAI). Where next-gen AI is developed addressing societal needs on real data.
San Diego Dec 2-7, 25 and Mexico City Nov 30-Dec 5, 25. Comments to this account are not monitored. Please send feedback to townhall@neurips.cc.
#MICCAI #medical_imaging #computer_aided_interventions #women_in_science
Professor a NYU; Chief AI Scientist at Meta.
Researcher in AI, Machine Learning, Robotics, etc.
ACM Turing Award Laureate.
http://yann.lecun.com
Prof at TUM and Imperial College London. Doing research on AI in healthcare and medicine
2xDad, DL practitioner, real-time computing fanatic, Prof. at AIBE, FAU Erlangen, Reader at DoC, Imperial College London, advisor at ThinkSono, former chef
Professor in Machine Learning for Imaging | Imperial College London
From the editors of Radiology: Artificial Intelligence, the leading journal on AI in radiology
#Radiology #RadSky #MedSky #AI #MachineLearning #DeepLearning
Founder of the worldβs most read daily AI newsletter:
www.therundown.ai. Sharing the latest developments in the world of artificial intelligence.
official Bluesky account (check usernameπ)
Bugs, feature requests, feedback: support@bsky.app