Babbitt | Daily Must-Read for Metaverse: Kuaishou's self-developed Wenshengtu large model "Ketu" has started internal testing; Ministry of Industry and Information Technology will give priority to the

CN
巴比特
Follow
1 year ago

Summary: Following the launch of the large language model "KwaiYii" last month, Kuaishou has launched the text-to-image large model "Kolors" within less than a month, filling the gap in the large model business after "Wenshengwen" with "Wenshengtu". According to Future Technology Power, on September 18, Kuaishou's latest development in the AIGC field, the self-developed text-to-image large model "Kolors," has been fully launched for internal testing within the company.

Image Source: Generated by Wujie AI

Hot News:

Bitpush News reported on September 18 that the Ministry of Industry and Information Technology's Science and Technology Department has publicly solicited opinions on the "Draft Plan for the Establishment of the Metaverse Standardization Working Group of the Ministry of Industry and Information Technology." Regarding the work plan after the establishment, the opinions proposed:

  1. Based on previous research, organize various parties in the domestic metaverse industry to strengthen the research and improvement of the standard system that reflects the typical characteristics of the metaverse, and formulate a roadmap for metaverse standardization.
  2. Following the principle of urgency, focus on accelerating the development of basic common standards such as metaverse terminology, reference architecture, key technical standards such as metaverse identity system, digital content generation, cross-domain interoperability, and prioritize the development of industry application standards such as "metaverse + industrial manufacturing."
  3. Accelerate research and development in new technologies, products, formats, and models in the metaverse field such as privacy protection, content regulation, and data security, and strengthen the integration and innovation of technologies in areas such as brain-computer interfaces, generative artificial intelligence, and quantum information.
  4. Conduct research on application standards for industrial metaverse, urban metaverse, virtual digital humans, and guide industry associations, standardization professional institutions, and technical organizations to promote standard popularization for the production and use of the metaverse industry, guiding enterprises to benchmark and improve the effectiveness of standard application in R&D, production, and management processes.
  5. Encourage domestic enterprises and institutions to actively participate in international standardization activities and timely propose international standard proposals. Relying on the foundation of domestic metaverse industry standardization work, actively promote the establishment of metaverse standardization-related organizations in international standardization organizations such as ISO, IEC, and ITU to enhance the comprehensive competitiveness of China's metaverse industry.

Kuaishou's self-developed text-to-image large model "Kolors" has recently started internal testing

According to Future Technology Power on September 18, Kuaishou's latest development in the AIGC field, the self-developed text-to-image large model "Kolors," has been fully launched for internal testing within the company. According to a member of the Kolors project development team, the "Kolors" text-to-image large model has three prominent features: powerful text understanding, rich detail depiction, and diverse style transformation. Previously reported, Kuaishou has landed new AIGC application scenarios, and the AI play and evaluation function has started internal testing.

DeepMind Co-founder: Generative AI is just a phase, next is interactive AI

According to DeepTech, Mustafa Suleyman, co-founder of DeepMind, stated in an interview last Friday that generative AI is just a phase in the development of artificial intelligence. "The next step is interactive AI, where robots can call on other software and other people to complete the tasks you set for them."

Alibaba Cloud's industry large model "Tongyi" obtains copyright approval

According to Jiemian News on September 18, the website of the China Copyright Protection Center shows that on September 13, Hangzhou Alibaba Cloud Feitian Information Technology Co., Ltd.'s "Tongyi Industry Large Model Platform" software copyright was approved for registration, with the current version number being V1.0. Hangzhou Alibaba Cloud Feitian Information Technology Co., Ltd. was established in March 2019, with Jiang Jiangwei as the legal representative and a registered capital of 1 million RMB, fully owned by Zhejiang Alibaba Cloud Computing Co., Ltd.

Baidu's "Wenxinyiyan Software" copyright approved for registration

According to Cailian News on September 18, Tianyancha showed that recently, Beijing Baidu Netcom Technology Co., Ltd.'s "Wenxinyiyan Software" copyright was approved for registration, with the current version number being V1.0.0. It is reported that "Wenxinyiyan" is a Baidu AI language model. Previously, Baidu had applied for registration of multiple "Wenxinyiyan" "Wenxinyiyan ERNIE BOT" and graphic trademarks. Previously reported, Baidu released the Wenxinyiyan plugin ecosystem "Lingjing Matrix."

Startup company Wayve releases LINGO-1, an interactive large model for autonomous driving, deeply integrating large models and autonomous driving

According to "Science and Technology Board Daily" on September 18, the startup company Wayve recently launched an interactive large model for autonomous driving, LINGO-1, which deeply integrates large language models and autonomous driving. After training with various visual and language data, LINGO-1 can not only perform visual question-answering (VQA) tasks such as perception, counterfactuals, planning, reasoning, and attention but also describe driving behavior and reasoning. It is reported that Wayve is a UK-based startup dedicated to achieving autonomous driving through reinforcement learning.

Zhiji LS6 car is equipped with Zhiji generative large model, with functions such as "one-click AI driving"

According to IT Home on September 17, SAIC Group recently stated during an investor survey that the new car Zhiji LS6, unveiled at the Chengdu Auto Show at the end of August, is equipped with the Zhiji generative large model (Beta version), based on a new multimodal perception and AI algorithm, claiming to "integrate billions of parameters + machine learning capabilities + natural language processing," accurately recognizing all commands and featuring diverse human-device interactions. Previously reported, Zhiji Automobile's Liu Tao: the industry's first GPT large model will be delivered online in October.

Microsoft announces M365 Copilot early access program, introducing AI productivity tools to the first batch of Australian institutions

According to Chinaz on September 18, Microsoft announced that the first batch of Australian customers have qualified for the Microsoft365 (M365) Copilot early access program, which is a globally invitation-only program. Australian institutions such as AGL, Data#3, Bupa, NAB, Powerlink Queensland, REST Super, and Suncorp are among the first in Australia to adopt M365 Copilot.

M365 Copilot is a powerful generative AI service that combines large language models (LLMs) with Microsoft365's organizational data, enabling these enterprises to leverage the latest generative AI technology in applications such as Microsoft Teams, Word, PowerPoint, and Excel, achieving new levels of productivity.

Google modifies "Useful Content Updates" policy, acknowledges challenges in regulating AI-generated content

According to Chinaz on September 18, Google's latest "Useful Content Updates" underwent its third iteration, aiming to change its stance on AI-generated content. The update, first introduced in August 2022, primarily aims to categorize content as "written for search engines" or "written for people." Google has adjusted the wording in its official useful content reference, with the previous versions emphasizing that the content generated by the useful content system is "written for humans" and original and useful. However, the latest version of the update focuses on "original and useful content created by the useful content system for people." This adjustment reflects Google's growing recognition of the challenges in regulating AI-generated content, emphasizing that content should be human-centric regardless of the specific content creation method. In addition, OpenAI has removed its low-accuracy AI recognizer and plans to research more effective text source technologies to help users understand whether audio or visual content is generated by artificial intelligence.

Metaverse Recommended Reading:

"AI Field: Recent Technological Highlights and Investment Opportunities"

The development of artificial intelligence technology is advancing rapidly, with many AI startups standing out in the field. In this article, let's take a look at which promising AI startups have been chosen by investors such as Sequoia Capital and KPCB.

https://www.8btc.com/article/6833194

"First-hand | Kuaishou's self-developed text-to-image large model 'Kolors' has arrived and started internal testing recently"

Since the second half of this year, Kuaishou has been frequently strengthening its large model business. GenAI New World first learned that Kuaishou's latest development in the AIGC field, the self-developed text-to-image large model "Kolors," has been fully launched for internal testing within the company. Following the launch of the large language model "KwaiYii" last month, Kuaishou has launched the text-to-image large model "Kolors" within less than a month, filling the gap in the large model business after "Wenshengwen" with "Wenshengtu."

https://www.8btc.com/article/6833162

"Unveiling the Transformer in the iPhone: Based on the GPT-2 architecture, the tokenizer contains emojis, produced by MIT alumni"

The "secrets" of Apple's Transformer have been uncovered by enthusiasts. A person named Jack Cook turned over the macOS Sonoma beta and found quite a bit of fresh information: in terms of model architecture, Cook believes that Apple's language model is more like one built on the GPT-2. In terms of the tokenizer, emojis are prominent.

https://www.8btc.com/article/6833200

免责声明:本文章仅代表作者个人观点,不代表本平台的立场和观点。本文章仅供信息分享,不构成对任何人的任何投资建议。用户与作者之间的任何争议,与本平台无关。如网页中刊载的文章或图片涉及侵权,请提供相关的权利证明和身份证明发送邮件到support@aicoin.com,本平台相关工作人员将会进行核查。

Bybit: $50注册体验金,$30,000储值体验金
Ad
Share To
APP

X

Telegram

Facebook

Reddit

CopyLink