Language is primarily a tool for communication rather than thought https://
nature.com/articles/s4158
6-024-07522-w
… “Language is a defining characteristic of our species, but the function, or functions, that it serves has been debated for centuries. Here we bring recent evidence from neuroscience and
@hardmaru
-
Language as Communication Tool Rather Than Thought
By
–
-
Base Models Produce More Creative Outputs Than Fine-Tuned Versions
By
–
Just let LLMs be LLMs I notice the same sort of thing in image generation models too. Usually the base model, while less aesthetically pleasing, produce more interesting and creative images. And over time, this ‘artificial aesthetic’ loses its charm as outputs become boring.
-
RLHF Alignment Reduces Language Models Creativity and Diversity
By
–
Creativity Has Left the Chat: The Price of Debiasing Language Models https://
arxiv.org/abs/2406.05587 “While RLHF has proven effective in reducing biases and toxicity in LLMs, this alignment process may inadvertently lead to a reduction in the models’ creativity and output diversity.” -
LLMs Discover Better Algorithms for AI Alignment Training
By
–
New Paper and Blog!https://t.co/zzAGMC2tRO
— hardmaru (@hardmaru) 13 juin 2024
As LLMs become better at generating hypotheses and code, a fascinating possibility emerges: using AI to advance AI itself! As a first step, we got LLMs to discover better algorithms for training LLMs that align with human preferences. https://t.co/80Ldqgxy3TNew Paper and Blog! https://
sakana.ai/llm-squared/ As LLMs become better at generating hypotheses and code, a fascinating possibility emerges: using AI to advance AI itself! As a first step, we got LLMs to discover better algorithms for training LLMs that align with human preferences. -
ARC Prize: $1M+ Competition for Open AGI Progress
By
–
ARC PRIZE: A $1M+ competition towards open AGI progress. Exciting initiative hosted by @mikeknoop @fchollet https://
arcprize.org/blog/launch -
AI Models Should Be Capable of Diverse Thinking
By
–
I want my AI models to be able to think different.
-
Apple Should Lead On-Device Foundation Models Development
By
–
Apple should be leading the development of on-device foundation models, embracing a future powered by energy-efficient, privacy-first, distributed AI systems.
-
LLM Fine-tuning and Model Merging for Cost-Effective Adaptation
By
–
“Training LLMs from scratch costs $ millions in compute. But LLMs can cheaply be adapted to new tasks via fine-tuning, leading to a proliferation of models that suit specific use cases. Fine-tuned models can be rapidly merged to combine capabilities and generalize to new skills.”
-
NeurIPS 2021 Research Paper Presentation
By
–
This was presented at NeurIPS2021:https://t.co/N6yD9VTFfw https://t.co/TvgpsR2whn
— hardmaru (@hardmaru) 1 juin 2024This was presented at NeurIPS2021: https://
openreview.net/forum?id=lM297
1LAwV
… -
Questioning if Large Language Models Truly Represent Intelligence
By
–
Makes you wonder if a giant LLM trained on swaths of data is the best abstraction to capture the nature and characteristics of intelligence.