Alibaba Cloud, the digitally intelligent lifeblood of Alibaba Group, today threw the curtains open to its latest marvel in AI image generation—Tongyi Wanxiang (where ‘Wanxiang’ stands for ’tens of thousands of images’)—at the 2023 World Artificial Intelligence Conference. This ground-breaking generative AI model is all set for beta testing by enterprise clientele in China. 📸🧪
In addition, the trailblazing cloud behemoth announced the debut of ModelScopeGPT—a multipurpose framework tailored to aid users in navigating intricate and specialized AI tasks across language, vision, and speech domains. This is achieved by leveraging a range of AI models on ModelScope—an open-source Model-as-a-Service (MaaS) platform rolled out by Alibaba Cloud last year, flaunting over 900 AI models. 🚀
👩💻 Tongyi Wanxiang: A New Era in Image Generation! 🎨
“Tongyi Wanxiang symbolizes yet another noteworthy stride in our quest for top-tier generative AI models. We remain committed to pioneering transformational technologies that empower businesses and communities to harness greater creativity and productivity,” voiced Jingren Zhou, the CTO of Alibaba Cloud Intelligence. 🏅🎯
“With Tongyi Wanxiang entering the scene, high-quality generative AI imagery will become more approachable, propelling the evolution of innovative AI art and creative expressions for businesses spanning myriad sectors, such as e-commerce, gaming, design, and advertising,” he added. 🎨💼
The AI model is a whiz at a broad spectrum of tasks, answering text prompts in Chinese and English to produce detailed images in numerous styles—ranging from watercolours, oil and Chinese paintings to animations, sketches, flat illustrations, and 3D cartoons. The model also possesses the talent to transfigure any image into a new one bearing a similar style and apply styles onto images via style transfer, a method that retains the content of the original image while implementing the visual style of another picture. 🖼️🔄
By harnessing Alibaba Cloud’s pioneering technologies in knowledge arrangement, visual AI, and natural language processing (NLP), Tongyi Wanxiang utilizes multilingual materials to bolster its training. It boasts an impressive semantic comprehension capability, yielding more accurate and contextually fitting image generation. 📚💪
And that’s not all. By optimizing the high-resolution diffusion process based on the signal-to-noise ratio, the model can strike a balance between composition accuracy and detail sharpness. This enhances its ability to generate high-contrast, visually striking images with clean backgrounds. 🌟
Curious? Watch the Tongyi Wanxiang demo video here! 🎥 Enterprise customers in China can apply for beta testing of Tongyi Wanxiang at: https://wanxiang.aliyun.com/
🧩 ModelScopeGPT: A Powerhouse for Advanced AI Tasks! 🚀
Alibaba Cloud didn’t stop at Tongyi Wanxiang. They also introduced ModelScopeGPT, a robust framework designed to harness the power of Large Language Models (LLMs) available on the platform. ModelScopeGPT employs LLMs as a controller to connect a comprehensive array of domain-specific expert models in the ModelScope open-source community. 🤖
Built within the rich Model-as-a-Service ecosystem, ModelScopeGPT utilizes the assorted AI capabilities offered on Alibaba Cloud. Enterprises and developers can access and execute the best-suited models for executing sophisticated AI tasks, such as developing multilingual videos, at no cost! 🆓
Since launching its own LLM—Tongyi Qianwen—in April, Alibaba plans to integrate the model across its various businesses to enhance user experiences. Moreover, the company’s customers and developers will be able to access the model to build custom AI features cost-effectively. 🧩
🏆 AI Hackathon Competition: Inspiring Innovation! 💡
ModelScope also hosted its inaugural AI Hackathon in China to promote industrial applications of AI models. With cash prizes and funding opportunities from leading venture capital firms, over 300 teams participated, and 56 made it to the final round. 🏁
Participants battled for the grand prize on two tracks: innovating a large language model to address a real-world problem, and using existing pretrained models to complete an assigned task, such as text-to-image generation or building an LLM-powered autonomous agent. 🥇
“By hosting competitions and other community events, we aspire to engage more with developers and entrepreneurs. We want to motivate them to bring their ideas to fruition, boost productivity, and craft versatile AI tools that shape the future of our industries,” said Jingren Zhou. 🌏🤝
