New simulation hypothesis drop.
Maybe the simulation is not physical and exact but neural and approximate.
i.e. not about simulating fields or particles with physical equations but a giant Diffusion Transformer++ creating a large "dream".
MACHINE LEARNING
-
Simulation Hypothesis: Neural Approximation Over Physical Exactness
By
–
-
Machine Learning’s Inclusive Approach to Algorithm Development
By
–
One reason for machine learning’s success is that our field welcomes a wide range of work. I can’t think of even one example where someone developed what they called a machine learning algorithm and senior members of our community criticized it saying, “that’s not machine
-
Fresh Python Recordings Available on Udemy Course
By
–
Complete fresh recordings of python is uploaded in my udemy course
Check out the courses below, the price is Rs 399 Complete Machine learning and NLP With End to end Projects with MLops
Course link : https://
bit.ly/3x94a2d -
Open-Source LLM Mixture Outperforms GPT-4o on AlpacaEval
By
–
This is an exciting discovery! The collective intelligence of multiple open-source LLMs outperforms GPT-4o. "Together Mixture-of-Agents" leverages diverse models to achieve 65.1% on AlpacaEval 2.0, surpassing GPT-4o's 57.5%.
-
Perfect Base64 Decoder Evolved in Neural Network Weights
By
–
somewhere inside the weights a perfect base64 decoder evolved. i wonder if @alexalbert__ can find a feature that does just that…
-
Researchers Use AI Models to Decode Dog Bark Meanings
By
–
Have you ever wondered what your dog's bark really means? In collaboration with Mexico's National Institute of Astrophysics and Optics and Electronics (INAOE) Institute in Puebla, the @UMich researchers are repurposing human speech models, like Wav2Vec2, to decode animal
-
Elon Musk Praises Ashok Elluswamy for Tesla AI Autopilot
By
–
@elonmusk praises Indian-origin engineer Ashok Elluswamy for his crucial role in Tesla's AI and Autopilot advancements. Musk highlights Elluswamy's leadership and the team's efforts in developing groundbreaking technologies, including Autopilot, computer vision, and humanoid
-
Curriculum Sampling and Flexible Training Dimensions in AI Models
By
–
“we sample i based on a curriculum” doing a lot of work here. what curriculum? this technique reminds me of how sora can natively train on/produce any resolution/dimension as opposed to cutting things into squares. feels like a good/onetime algorithmic win but unclear (to me)
-
Linear Decoding Breakthrough Achieves 1M Context Length
By
–
This looks super promising — linear decoding complexity while improving on longer context lengths up to 1M! Super interested to see how this does when scaled up.
-
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.