TLDR AI 2023-10-19

ADEPT RELEASES MULTIMODAL LLM

Adept has released an extremely simple (no image encoder) and very performant (outperforms 13B models) multimodal model. This article outlines how Adept evaluated the model’s performance and some details about the model.

MUJOCO 3

OPENAI FAILS TO DEVELOP MORE EFFICIENT AI MODEL

USE SMALL MODELS TO STUDY LARGE MODELS

Instabilities in training at large scales are hard to reproduce for most researchers. This paper shows how to reproduce instabilities in small models and verifies that many commonly reported fixes work just as well on these small models.

A MORE EFFICIENT LANGUAGE MODEL ARCHITECTURE

Researchers have developed BitNet, a new design for large language models that uses much less energy and memory.

ZIPFORMER: A NEW TRANSFORMER FOR SPEECH RECOGNITION

The study introduces Zipformer, a revamped transformer for automatic speech recognition (ASR) that's faster, uses less memory, and outperforms the widely-used Conformer.

LOCAL CHATBOT WITH MEMORY

MemGPT is a local chatbot that can read and write to memory. This allows you to have a local assistant that remembers important context about previous conversations.

CHATGPT SYSTEM PROMPTS

System prompts allow you to change the behavior of ChatGPT-based language models. This repository contains some very well constructed system prompts that could improve usability based on your preferences.

A NEW APPROACH TO 3D OBJECT DETECTION FROM SINGLE IMAGES

This repository introduces MonoSKD, a groundbreaking method for detecting 3D objects using only one image that effectively bridges the knowledge gap between LiDAR and regular RGB-based models.

THE WORLD’S BIGGEST AI MODELS AREN’T VERY TRANSPARENT

FOXCONN AND NVIDIA ARE BUILDING AI FACTORIES

CHATGPT WEB BROWSING EXITS BETA, DALL-E 3 ENTERS BETA

OPENAI IS BUILDING AN IMAGE DETECTOR WITH 99% ACCURACY