Alibaba launches its ChatGPT-like AI model for public use amid loosening restrictions in China

Alibaba introduced that its proprietary massive language mannequin, a synthetic intelligence system referred to as Tongyi Qianwen, will probably be obtainable for public and enterprise entry all through China beginning Sept. 13. 

Tongyi Qianwen is a ChatGPT-like massive language mannequin educated on a corpus of English and Chinese language textual content. Whereas its actual specs aren’t recognized — early rumors indicated it could be educated with as many as 10 trillion parameters, 10 occasions as many as OpenAI’s GPT4, however these stay unsubstantiated — Alibaba previously released two 7 billion-parameter open-source models based mostly on the Tongyi Qianwen structure.

Alibaba Cloud launched two open-source massive imaginative and prescient language fashions, Qwen-VL and Qwen-VL-Chat, that may comprehend photos, texts, and bounding containers in prompts and facilitate multi-round query answering in each English and Chinese language.

Be taught extra: https://t.co/LrVNzNIRVy… pic.twitter.com/0t0Et6j1BU

— Alibaba Cloud Intelligence Group (@cloudintelgrp) August 30, 2023

Beforehand, Tongyi Qianwen had solely been obtainable to a restricted group of customers throughout its beta check part. The general public rollout coincides with a current loosening of restrictions associated to using synthetic intelligence (AI) applied sciences within the Folks’s Republic of China.

A set of pointers revealed by the Chinese language authorities in June dictated that, going ahead, all AI applied sciences launched to the general public would require a particular vetting and certification course of.

The principles went into impact on Aug. 15. As Cointelegraph reported, a number of Chinese companies were given approval to launch fashions, together with Baidu, Tencent, TikTok and ByteDance.

Associated: Tencent unveils ChatGPT rival in China amid continuing US AI chip ban

Among the many provisions included within the up to date restrictions are guidelines barring the era of photos within the likeness of China’s president, Xi Jinping, and mandates indicating organizations will tackle objectionable content material inside a three-month interval. Earlier variations of the laws referred to as for financial fines, however as Cointelegraph reported, those plans were axed.

As China explores a loosening of its rules, the USA has taken solely preliminary steps to manage AI applied sciences. Most just lately, on Sept. 13, Senate Minority Chief Chuck Schumer hosted a litany of prime U.S. tech CEOs and founders within the first of 9 scheduled boards to debate potential coverage concepts.