Alibaba launches AI model that can understand images and have more complex conversations

8 months ago 81

An Alibaba Group motion is seen astatine the World Artificial Intelligence Conference successful Shanghai, July 6, 2023.

Aly Song | Reuters

Alibaba connected Friday launched a caller artificial quality exemplary that the institution says tin recognize images and transportation retired much analyzable conversations than the company's erstwhile products, arsenic the planetary contention for enactment successful the exertion heats up.

The Chinese exertion elephantine said that its 2 caller models, Qwen-VL and Qwen-VL-Chat, volition beryllium unfastened root — meaning that researchers, academics and companies worldwide tin usage them to make their ain AI apps without needing to bid their ain systems, truthful redeeming clip and expense.

Alibaba said that Qwen-VL tin respond to open-ended queries related to antithetic images and make representation captions.

Qwen-VL-Chat meantime caters to much "complex interaction," according to Alibaba, specified arsenic comparing aggregate representation inputs and answering respective rounds of questions. Some tasks that Alibaba says Qwen-VL-Chat tin execute see penning stories and creating images based connected photos that a idiosyncratic inputs, arsenic good arsenic solving mathematical equations shown successful a picture.

One illustration Alibaba gave is of an input featuring a infirmary motion successful the Chinese language. The AI tin reply questions astir the locations of definite infirmary departments by interpreting the representation of the sign.

So far, overmuch of generative AI — wherever the exertion generates responses based connected quality inputs — has focused connected responding to text. The latest mentation of OpenAI's ChatGPT besides has the quality to recognize images and respond successful text, overmuch similar Qwen-VL-Chat.

Alibaba's 2 latest models are built upon the company's ample connection exemplary called Tongyi Qianwen, released earlier this year. An LLM is an AI exemplary trained connected immense amounts of information and underpins chatbot applications.

The Hangzhou-headquartered institution this period open sourced 2 different AI models. While not earning Alibaba immoderate licensing fees, the open-source organisation volition assistance the institution get much users for its AI exemplary — astatine a clip erstwhile the firm's unreality part is looking to reignite growth, arsenic it prepares to spell public.

Read Entire Article