I'm a tech creator
【Computer Technology Corner】Focus on sharing computer knowledge, cutting-edge information, operating systems, all kinds of practical software, program source code and other high-quality good things recommended. Welcome to follow me!
Body:
The past 2023 is the first year of the outbreak of AI, and since OpenAI released ChatGPT, various artificial intelligence models at home and abroad have sprung up, forming a situation where you catch up with me.
Recently, OpenAI's Wensheng model Sora has continued to attract attention with its one-minute duration, coupled with the high realism and high quality of the demo. SORA is a GPT-4-based multimodal model that can generate corresponding outputs based on text inputs, thereby enabling text-to-text conversion. SORA's technology is based on GPT-4 and DALL-E and can provide high-resolution and high-quality generation.
Today, I would like to recommend a domestic rising star to youkimi chat, a magical platform for you to chat, learn, and create with AI.
It was founded by a group of AI enthusiasts and experts from Tsinghua University, Carnegie Mellon University and other prestigious universities, who have rich experience in AI research and development, have participated in many well-known AI projects and competitions at home and abroad, and won many AI awards and honors. Their goal is to make AI no longer a distant concept, but an intimate companion, a partner who can accompany you, help you, and grow you.
The biggest advantage of Kimi Chat is its breakthrough in long text technology. Long text technology refers to enabling large models to process longer text inputs, thereby improving their effectiveness and coverage.
Currently, most of the world's large language models have input lengths of less than 2K, and Kimi Chat is able to support it200,000 Chinese charactersis the longest context input length that can be supported in a large model service that can be used productively in the global market.
Through innovative network structure and engineering optimization, it realizes a lossless long-range attention mechanism under 100 billion parameters, and does not rely on "shortcut" schemes that have great performance damage such as sliding windows, downsampling, and small models.
You can just throw it a document, an e-book, or a ** and let it analyze the main content for you.
kimi chatIt also lets you create your own AI persona, such as:
You can choose your AI character's name, gender, age, occupation, hobbies, personality, etc., to make it more in line with your preferences and needs.
You can teach your AI character all kinds of knowledge, skills, habits, rules, and more to make it smarter, more useful, and more interesting.
You can co-create all kinds of content with your AI character, such as **, comics, **, etc., to make it more creative and talented.
Kimi Chat is a revolutionary platform that aims to empower everyone to have their own AI buddy to communicate, learn, create, and enjoy the infinite possibilities that AI brings.
If you're interested in Kimi Chat, you can visit its official websiteto sign up for an account and start your AI journey.
In addition to text and images, AI can also process other types of data, such as sounds and images. This involves the study of multimodal models, that is, the use of large-scale neural network models to process multiple different types of data at the same time, so as to achieve more comprehensive and accurate analysis and generation.
Multimodal models are a cutting-edge and popular direction in the field of AI, and it has a wide range of application scenarios, such as Q&A, subtitles, generation, image description, image retrieval, etc.
Behind Kimi Chat, there is a company calledMoonshot AI, an AI company, which focuses on the field of general artificial intelligence, is committed to the development and application of large model technology. The founding team of the dark side of the moon is a group of AI researchers and engineers with rich experience and outstanding achievements in the field of large models, who have participated in the research and development of many well-known large models at home and abroad, such as Google Bard, Gemini, Einstein, Pangu, Wudao, etc., and invented AI including Transformer-XL, XLNet, Rope, Detectron2, Group Normalization, etc Milestones in the field.
Founded in April 2023, the Dark Side of the Moon has received nearly 2 billion yuan of financing from well-known investment institutions such as Sequoia Capital, Today Capital, and Lisi Capital in less than a year. In February 2024, the dark side of the moon completed a new round of financing of more than $1 billion, with investors including Sequoia China, Xiaohongshu, Meituan, Alibaba, etc., with a post-investment valuation of about $2.5 billion. This is the largest single round of financing in the field of large models in China, and it is also the highest valuation among domestic large model companies.
According to multiple sources, the dark side of the moon is also secretly developing a general multimodal model, which is expected to be launched within this year. In other words, the dark side of the moon that has made phased progress in the field of text-based large language modelsIn the future, it will also compete with domestic and foreign counterparts such as OpenAI in the field of multimodal models based on images and **.
The multimodal model of the dark side of the moon will be based on its self-developed BEIT-3 model, a GPT-4-based visual and visual-linguistic model that can process both image and text data to enable the analysis and generation of images and text. BEIT-3's technology is based on GPT-4 and Masked Image Modeling, which can provide efficient and powerful visual and visual-linguistic pre-training.
In the future, we will continue to pay attention to whether the dark side of the moon can occupy a place in the field of "multimodal models". The vision of the dark side of the moon is to enable everyone to have their own AI companions, communicate, learn, create, and enjoy the infinite possibilities brought by AI.