1. Runway Announces General World Models Research and Development
AI startup Runway announced on its website that it is conducting long-term research around General World Models, aiming to represent and model a wide range of interactions consistent with those in the real world. According to Runway, its current video generation systems, such as Gen-2, can be viewed as very early General World Models, with some understanding of physics and motion. However, they still need to work on dealing with issues such as complex camera or object motion. However, there are still difficulties in dealing with complex issues such as camera or object motion.
To build generic world models, Runway is conducting several publicly available studies, including generating maps of environments, navigating and interacting in environments, capturing the dynamics of the world and its inhabitants, and modeling realistic human behavior.
Watching the video to find out more:
2. Microsoft releases Phi-2, a 2.7-billion-parameter language model
Microsoft released Phi-2, a 2.7 billion parameter language model that is said to match or outperform models up to 25 times larger, thanks to innovations in model scaling and training data curation.
With its compact size, Phi-2 is an ideal playground for researchers, including exploring mechanistic interpretability, improving confidence, or fine-tuning experiments on a variety of tasks.
3. ChatGPT Plus subscription reopens
OpenAI CEO Altman announced on social media platform X that ChatGPT Plus subscriptions have been reopened and that “we have found more GPUs”. Previously, on November 15, he announced that he was suspending new subscriptions after ChatGPT usage spiked beyond the company’s capacity.
4. Google launches Imagen 2, a text-to-image AI Model
Google has announced the launch of Imagen 2, a new text-to-image model that is now generally available for Vertex AI customers on the allowlist (i.e., approved for access).
Imagen 2 supports the following functions:
- Generating high-quality, photorealistic, high-resolution, aesthetically pleasing images from natural language prompts;
- Text rendering in multiple languages to create images with accurate text overlays;
- Logo generation to create company or product logos and overlay them in images;
- Visual questions and answering for generating captions from images, and for getting informative text responses to questions about image details.
5. BMW licensed to test Level 3 vehicles in Shanghai
BMW Group has announced that it has been granted a test license for vehicles equipped with Level 3 autonomous driving functions on certain highways in Shanghai. The move is aimed at promoting the operation of driverless vehicles in the world’s largest automotive market.
BMW is the first automaker to comply with the United Nations Automated Lane Keeping System regulations and receive approval for high-speed road testing in China. The new license will allow BMW to expand the areas where it can test advanced autonomous driving technologies in Shanghai. In 2018, BMW received a license that allowed it to conduct tests on 5.6 kilometers of public roads in the city.
