For most companies, hiring more people is strictly better. However, this is often not true in AI research. AI research is often bottlenecked by compute, and when this is the case, hiring more researchers can be counter-productive. I remember back at Google Brain, my manager once
MACHINE LEARNING
-
Moondream1 Vision Model Outperforms Larger Models Locally
By
–
This vision model, Moondream1 is so good that I had to make a @streamlit app for it.
— Pietro Schirano (@skirano) 23 janvier 2024
Despite being just 1.6B parameters, it outperforms much larger models I've tested.
And it runs completely locally and free on your CPU.
Video is at real speed.
Link to the repo 👇 https://t.co/v1gxvwSp4R pic.twitter.com/PPqiOgdCXNThis vision model, Moondream1 is so good that I had to make a @streamlit app for it. Despite being just 1.6B parameters, it outperforms much larger models I've tested. And it runs completely locally and free on your CPU. is at real speed. Link to the repo
-
Stanford research reveals AI disparities in U.S. government IRS algorithms
By
–
A research team supported by @StanfordHAI
’s Hoffman-Yee grant showed how AI-driven learning can benefit U.S. government agencies, and uncovered disparities in IRS algorithms. https://
stanford.io/3vU43q5 -
Exphormer: Scaling Graph Transformers with Sparse Attention
By
–
Graph transformers are important for ML on graph-structured data, but scaling them up for large graphs is difficult. Here you can learn about Exphormer, a sparse attention framework that uses expander graphs to improve scalability for graph transformers. https://t.co/JxQqOq4yNn pic.twitter.com/85sLNnFkiD
— Google AI (@GoogleAI) 23 janvier 2024Graph transformers are important for ML on graph-structured data, but scaling them up for large graphs is difficult. Here you can learn about Exphormer, a sparse attention framework that uses expander graphs to improve scalability for graph transformers. https://
goo.gle/48JVvAW -
Predibase Reduces Deep Learning Model Development Time from Months to Minutes
By
–
"The time it takes to build production #deeplearning models on top of Snowflake has been reduced from months to minutes and at a fraction of the cost with Predibase" –
@tradeparadigm Check out our latest blog to learn more about this cool project: https://
pbase.ai/4b0haWW -
The Next Leap In AI: From Large Language Models To Large World Models
By
–
Check out the latest article in my newsletter: The Next Leap In AI: From Large Language Models To Large World Models? https://
linkedin.com/pulse/next-lea
p-ai-from-large-language-models-world-fabio-moioli-i5byf
… via @LinkedIn -
DiConStruct: Causal Concept Explanations via Black-Box Distillation
By
–
DiConStruct: Causal Concept-based Explanations through Black-Box Distillation Moreira et al.: https://
arxiv.org/abs/2401.08534 H / T : Pedro Saleiro #ArtificialIntelligence #DeepLearning #MachineLearning -
Berkeley Humanoid vs Optimus: Autonomous Learning vs Teleoperation
By
–
Optimus is teleoperated (aka remote controlled videos). Berkeley humanoid is an actual learned Transformer walking autonomously on unseen environments. There’s no comparison.
-
Optimizing HuggingFace Dataset Loading Performance
By
–
wrote some faster dataloading code for HuggingFace datasets – sped up datasets.load_from_disk() from 17 minutes to around 30s the problem I was running into is that our virtual disk at school is really slow. HF loads datasets in a single thread and doesn't always memory-map.