AI News: ChatGPT 4 Coming Next Week!; The first open-source ChatGPT alternative got released; Microsoft just introduced Visual ChatGPT; Fine-tuning 20B LLMs with RLHF on a 24GB consumer GPU...

Microsoft will launch ChatGPT 4 with AI videos next week: The news was revealed by Andreas Braun, Chief Technology Officer at Microsoft Germany, at a recent event titled “AI in Focus — Digital Kickoff” (via Heise). According to Braun, “We will introduce GPT-4 next week … we will have multimodal models that will offer completely different possibilities — for example videos.”

The first open-source ChatGPT alternative got released: TogetherCompute has released a 20B chat-GPT model called OpenChatKit under the Apache-2.0 license, which is available for free on Hugging Face. A demo and announcement are available online. The researchers collaborated with LAION and Ontocord to create the training dataset.

MathPrompter: A technique that improves LLM performance on mathematical reasoning problems. It uses zero-shot chain-of-thought prompting and verification to ensure generated answers are accurate. It improves over state-of-the-art on the MultiArith dataset (78.7% → 92.5%) evaluated using 175B parameter GPT-based LLM.

Microsoft just introduced Visual ChatGPT: It connects ChatGPT and different visual foundation models to enable users to interact with ChatGPT beyond language format. It enables you to send, receive, edit images during chatting.

GigaGAN: A Large-scale Modified GAN Architecture for Text-to-Image Synthesis. Better FID Score than Stable Diffusion v1.5, DALL·E 2, and Parti-750M. Generates 512px outputs at 0.13s. Native Prompt mixing, Prompt Interpolation and Style Mixing.

ODISE: Stable Diffusion but for Open-Vocabulary Segmentation and Detection. Stable Diffusion generates beautiful images, but can it be used for open-world recognition? This CVPR2023 paper shows that the pre-trained diffusion model indeed is a good image parser, and allows for open-vocabulary segmentation and detection.

Fine-tuning 20B LLMs with RLHF on a 24GB consumer GPU: Researchers officially release the integration of trl with peft to make Large Language Model (LLM) fine-tuning with Reinforcement Learning more accessible to anyone. peft is a general tool that can be applied to many ML use-cases but it’s particularly interesting for RLHF as this method is especially memory-hungry!

Do You Know Marktechpost has a community of 1.5 Million+ AI Professionals and Engineers?