[Big Model Daily, December 23-24] Research: Unlimited new clothes can be purchased for zero yuan, Alibaba Outfit Anyone enables real people to change their clothes Industry OpenAI plans a new round of financing, with a valuation of over US$100 billion

Buy unlimited new clothes for zero yuan, Alibaba Outfit Anyone allows real people to change their clothes

https://news.miracleplus.com/share_link/14200

You can try on various clothes without actually trying them on. Virtual fitting technology has upgraded the “QQ Show” into a real-person version, opening the door to a new world for the fashion industry. However, existing virtual fitting methods still have challenges in terms of realism and consistency in details. While diffusion models excel at creating high-quality and photorealistic images, they leave much to be desired in maintaining control and consistency in certain scenarios such as virtual fittings. Outfit Anyone utilizes an innovative two-flow conditional diffusion model to effectively solve these problems and can accurately handle the deformation effect of clothing to achieve a more realistic try-on experience. The biggest feature of Outfit Anyone is its extreme adaptability and wide range of applications. Not only can it be adjusted to suit different postures and body shapes, whether it is an animated image or a real person, it can be changed with one click. Now open for trial play.

Combining LLM with drug discovery, NVIDIA, Mila, and Caltech teams release multimodal molecular structure-text models

https://news.miracleplus.com/share_link/14201

Starting in 2021, the combination of big language and multi-modality has swept the machine learning research community. With the recent application of large models and multi-modality, a natural thought is whether we can also use these technologies for drug discovery? And do these natural language text descriptions bring new perspectives to the challenging problem of drug discovery? The answer is yes and optimism. Recently, research teams from Canada’s Montreal Institute for Learning Algorithms (Mila), NVIDIA Research, the University of Illinois at Urbana-Champaign (UIUC), Princeton University and the California Institute of Technology jointly learned the chemical structure and text description of molecules through comparative learning strategies. , a multimodal molecular structure-text model MoleculeSTM is proposed.

OpenAI plans new round of financing, valuation exceeds 100 billion US dollars

https://news.miracleplus.com/share_link/14202

OpenAI, the leader in large language model technology, is raising funds again without stopping. OpenAI is beginning discussions to raise a new round of financing that would value it at more than $100 billion, Bloomberg reported on Friday. This transaction will add another important step to the financing history of OpenAI, the proposer of ChatGPT. An anonymous source familiar with the matter said that investors who may participate in this round of financing are already involved in preliminary discussions. It said details such as the terms, valuation and timing of this round of financing have not yet been finalized and may continue to change. If the funding round goes as planned, it will make OpenAI, the world’s hottest AI company, the second-largest startup in the United States, behind Elon Musk’s SpaceX, according to CBInsights.

At least $50 million, sources say Apple is buying news gathering rights to train generative AI

https://news.miracleplus.com/share_link/14203

According to the New York Times, Apple has reached agreements with a number of major publishers to collect their relevant news content to train generative AI systems. The report states that Apple aims to reach multi-year cooperation agreements with these major publishers and has already reached out to Condé Nast, NBC News and IAC.