1. Google releases VideoPoet, a zero-sample video generation model
Google announced on its Web site the release of VideoPoet, a Large Language Model (LLM) for zero-sample video generation capable of performing a variety of video generation tasks, including text-to-video, image-to-video, video stylization, video repair, video expansion, and video-to-audio. The model uses an approach that seamlessly integrates multiple video generation functions into a single large model, rather than relying on separate training components specialized for each task.
Website: https://sites.research.google/videopoet/
Paper address: https://arxiv.org/abs/2312.14125
2 IBM Acquires Two AI Data Platforms for $2.33 Billion. Data Are Key.
IBM announced that it would acquire StreamSets and webMethods, the enterprise integration platforms from German independent software vendor Software AG, for $2.33 billion.
“Together with IBM’s watsonx AI and data platform, as well as its application modernization, data fabric and IT automation products, StreamSets and webMethods will help clients unlock the full potential of their applications and data,” said Rob Thomas, IBM’s senior vice president of software and chief commercial officer.
3. Arcfox and Deepal Granted Permission for L3 Autonomous Driving Tests in China
In a significant development in the autonomous vehicle industry, Arcfox and Deepal have been granted permission to test vehicles equipped with L3 autonomous driving systems in China.
This announcement comes after several Chinese government departments jointly issued a notice last month, permitting cars supporting L3 and L4 autonomous driving to conduct pilot road tests. This notice also marked the first time the determination of accident liability was clarified.
Arcfox, a brand affiliated with the Beijing Automotive Group, emerged as one of the first companies authorized to conduct L3 autonomous driving tests on Beijing’s highways. This authorization came to light after the city announced rules for road testing of vehicles with conditional autonomous driving capabilities on December 16.
Meanwhile, Deepal, the new energy vehicle (NEV) division of Changan Automobile, also made an announcement yesterday. They have been authorized to conduct L3 autonomous driving road tests on highways and expressways in Chongqing, where the company’s headquarters are located.
This news follows the announcements from other automakers, including BMW, Mercedes-Benz, and IM Motors, who have also been authorized to conduct L3 autonomous driving tests following the new regulations announced by China last month.
4. OpenAI Forms New Team to Prevent AI Risks
OpenAI has released an AI security preparedness framework (in beta) detailing OpenAI’s process for tracking, assessing, predicting, and preventing risks. According to the document, OpenAI will establish a dedicated Preparedness team to oversee the operational structure of technical work and security decisions. The Preparedness team will drive the technical work, examine the limits of the capabilities of the state-of-the-art model, conduct assessments, and summarize reports. OpenAI will create a cross-functional group of security advisors who will review all reports and present them to both leadership and the board. While the executive team is the decision maker, the board has the power to override decisions.
Document address:
cdn.openai.com/openai-preparedness-framework-beta.pdf
5. NewsGuard: 614 AI fake news sites identified worldwide
News data site NewsGuard reported yesterday that to date, its team has identified 614 unreliable AI-generated news and information sites in 15 languages. That’s a 1,153 percent increase in the number of fake news sites compared to 49 in April this year. Most of these sites use information such as iBusiness Day, Ireland Top News, Daily Time Update, and the latest AI-generated news and information. “These sites have seemingly authoritative names such as ‘iBusiness Day’, ‘Ireland Top News’, ‘Daily Time Update’ and others, which can be difficult for consumers to recognize. These sites publish articles on topics such as politics, technology, entertainment, and travel, and contain false content such as celebrity death hoaxes, fabricated events, and old events presented as if they just happened.
