The Daily Index — Monday, August 14, 2023 · TLDR Index

HEADLINES & LAUNCHES

TLDR AI

TLDR AI

ANDREW NG ON LLM WORLD MODELS

Recent studies suggest LLMs like Othello-GPT build world models, implying that they understand the world to an extent. The network discerned the underlying 8x8 board structure and game rules by training on game moves without explicit game rules. This challenges traditional notions, like John Searle's Chinese Room Argument, about machine understanding.

deeplearning.ai · 3 min

TLDR AI

Untitled entry

Serve Robotics, the autonomous sidewalk delivery robot startup that spun out of Uber’s acquisition of Postmates, is going public via a reverse merger with a blank-check company.

techcrunch.com · 3 min

TLDR AI

SDXL CONTROLNET

One of the best ways to control image generation is to use edges to condition the final output. The best open source model, SDXL, has a controlnet variation trained and ready to use.

huggingface.co · 7 min

TLDR AI

SECURE SYSTEM FOR EVALUATING CODE MODELS

One hard challenge when working with language models that write code is evaluating the quality of the generated code. This is difficult because running arbitrary code is dangerous for a system. The code could do anything, including deleting everything. MosaicML offers a secure environment for evaluating generated code.

mosaicml.com · 12 min

TLDR AI

FOLLOW ANYTHING: A ROBOT THAT CAN TRACK AND FOLLOW ANYTHING

A robotic system called "Follow Anything" can spot, trace, and follow any object in real time, even if it has never seen it before. This smart robot can recognize objects from just a text description, a picture, or even a simple click - all powered by a regular laptop!

arxiv.org · 12 min

TLDR AI

DO MACHINE LEARNING MODELS MEMORIZE OR GENERALIZE?

This article examines the training dynamics of a tiny model and reverse engineers the solution it finds. It provides an illustration of the exciting emerging field of mechanistic interpretability.

pair.withgoogle.com · 17 min

TLDR AI

ZOOMING IN ON IMAGES BEFORE CLASSIFYING

If you zoom in on salient pieces of an image before running classification, you can get 98% on ImageNet. Fascinating and fairly easy implementation for real world computer vision systems.

zoom.taesiri.ai · 8 min

TLDR AI

WIZARD MATH

The WizardLM group from Microsoft has released the world’s best open source math model. It’s really good at reasoning through multi-step problems. The model hallucinates a bit, which is really fun for math, and offers explanations. I think this may be one of the best models for education for those interested in learning mathematics.

github.com · 6 min

TLDR AI

NVIDIA LANGUAGE, VISION ASSISTANT

Nvidia makes it easy to train your own vision-language assistant based on its Nemo model.

catalog.ngc.nvidia.com · 3 min

TLDR AI

ORCA AND PLATYPUS LLM COMBINED

There are lots of models based on Llama 2 and trained on GPT-4 outputs. These are somewhat in that camp, but as non-commercial license models go, they are quite powerful.

platypus-llm.github.io · 4 min