Lava Alpaca LLaVA is here: Like GPT-4, you can view pictures and chat, no invitation code is required, and you can play online
Link: https://news.miracleplus.com/share_link/7263
The model demonstrates some near-picture-text understanding capabilities of multimodal GPT-4: it achieves a relative score of 85.1% relative to GPT-4. The synergy of LLaVA and GPT-4 achieves a new SOTA of 92.53% accuracy when fine-tuned on Science QA.
120 top technical experts use GPT-4 to create a brain hole invention award
Link: https://news.miracleplus.com/share_link/7264
Last weekend, the ChatGPT Plugin Hackathon hosted by Cerebral Valley and sponsored by companies and developers such as OpenAl, StabilityAl and Replit was held in Los Angeles. Based on the exclusive access to the GPT-4 plugin provided by OpenAl, 120 top technical talents launched a fierce programming competition within 24 hours.
OpenAl demonstrates the official version of AUtoGPT live! One-stop creation and tweeting of paintings, autonomous calls to external applications to complete tasks
Link: https://news.miracleplus.com/share_link/7265
Just when the AutoGPT project broke 100,000 Stars, OpenAl also released a blockbuster. Co-founder Greg Brockman (Greg Brockman) personally demonstrated the upcoming new features of ChatGFT on the spot.
Ali's latest research: When ChatGPT meets the recommendation system, it encounters Waterloo
Link: https://news.miracleplus.com/share_link/7266
ChatGPT performs poorly on many tasks. The strong generalization ability of ChatGFT seems to be temporarily invalid in recommender systems.
'ChatGFT' of Satellite Photos Already in Military Use
Link: https://news.miracleplus.com/share_link/7267
Using advanced generative artificial intelligence and a massive dataset of Earth imagery, objects can be found almost anywhere in just a few hours.
———————————————
Twitter:
One picture to master chatGPT prompt skills
Link: https://news.miracleplus.com/share_link/7268
A rap song generated using Drake's voice caused a sensation
Link: https://news.miracleplus.com/share_link/7269
Brief description of the key steps of large model training
Link: https://news.miracleplus.com/share_link/7270
Use GPT-4 on the prompt to write formulas for painting
Link: https://news.miracleplus.com/share_link/7271
GPT-4 to describe consciousness and then use its description to generate images, videos and soundtracks based on those images.
Link: https://news.miracleplus.com/share_link/7272
The application of GPT in education
Link: https://news.miracleplus.com/share_link/7273
"I've never seen so many entrepreneurs start up on Replit before"
Link: https://news.miracleplus.com/share_link/7274
———————————————
product:
Noty.ai
Link: https://news.miracleplus.com/share_link/7275
Noty.ai is a meeting assistant powered by ChatGPT that converts meetings into transcriptions, action items, summaries and follow-ups in seconds.
———————————————
study:
This article summarizes the classic models necessary for super-resolution analysis
Link: https://news.miracleplus.com/share_link/7278
Super-resolution analysis can be divided into video super-resolution VSR (Video Super Resolution) and single-frame super-resolution SISR (Single Image Super Resolution). This article focuses on single-frame super-resolution analysis methods. This article focuses on deep learning technology, reviewing the necessary TOP model in SISR,
ChatGPT architect, PPO algorithm author, Berkeley alumnus John Schulman's artificial intelligence journey
Link: https://news.miracleplus.com/share_link/7279
In December 2015, shortly before finishing his Ph.D. in electrical engineering and computer science at UC Berkeley, John Schulman co-founded the ambitious software company OpenAl. At openAl, he led the reinforcement learning team that developed chatbots based on the company's generative pretrained language model (GPT)
"Scaling large language models unleashes their emergent power"
Link: https://news.miracleplus.com/share_link/7279
Jason Wei, a senior research scientist at Google (now OpenAl), gave a speech at the Institute of Information Science at the University of Southern California in February: "Scaling large language models unleashes their emerging capabilities"
Microsoft President: China will be the main opponent of ChatGPT; Moss large model is open source; Twitter is restoring blue V certification for millions of fan accounts
Link: https://news.miracleplus.com/share_link/7281
Microsoft integrates LLMs into SDK; Google uses Al to generate ads; Atlassian Iintelligence releases
Link: https://news.miracleplus.com/share_link/7282