Vous avez raté #MesFlashsdelaSemaine ? C’est le moment de rattraper votre retard ! Avoir foi en l’humanité ou en l’IA, telle est la question, à mesure que l’intelligence artificielle prend de plus en plus de place dans nos vies.
Open AI construit un nouveau modèle pour doter
RESEARCH
-
Faith in Humanity or AI: OpenAI’s New Model Development
By
–
-
Long-Context Transformers: Training Llama3 with 60k Token Sequences
By
–
Introducing long-context transformer using mini sequences. It is a simple and effective method for highly efficient and accurate LLM training with extremely long sequences. Our research demonstrates that the Llama3-8B model can be trained with context lengths up to 60k tokens on
-
PaliGemma Fine-tuned Model for Visual Question Answering
By
–
https://
huggingface.co/abhishek/autot
rain-paligemma-finetuned-vqa
… -
Llama 3.1 405B Launch: Open Weights, GPT-4o Comparable Performance
By
–
Llama 3.1 405B launches • Comparable to GPT-4o and Claude 3.5 Sonnet, according to the benchmarks
• The weights are publicly available
• 128k context even on the smaller models Discussion on r/ML -
GPT-4 Turbo Outperforms GPT-4o on Key Evaluation Dimensions
By
–
indeed. we haven’t finished our 4o mini eval, but our evals indicate that 4 turbo is still better than 4o on many important dimensions
-
Offline vs Online Distillation: Implementation Discrepancy Analysis
By
–
It feels like they're talking about "offline" distillation here, but actually implemented online distillation as described in the paper
-
Life’s Origin Question Explored at ICML2024 Conference
By
–
What a stimulating day at #ICML2024! No better way to end it than by having dinner with a dear old friend @j_foerst
, enjoying local food, and pondering on: “The transition from non-life to life has never been observed experimentally”. That’s why AI/ML conferences are so amazing -
Aleksander shifts focus to new AI safety research project
By
–
this is wrong. aleksander is working on a new and v important research project, and joaquin and lilian are taking over the preparedness team as part of unifying our safety work. aleksander will continue to support preparedness work in various ways.
-
Karpathy Plans Multiple Llama 3.1 Fine-tuning Projects
By
–
Yep I’d like to do many Llama 3.1 finetunes, coming up.
-
Visa and travel barriers exclude African researchers from AI conferences
By
–
I think it is more nuanced. It is true that visas exclude people, and sadly even among African countries visas are an issue. There is also bias in the sense that these meetings mostly occur in Europe and north America. It is costly and demanding to travel to these conferences