1. Stable Diffusion 3.0 with Sora-like architecture released
According to the official website of AI startup Stability AI, Stability AI has released a new version of its AI text-generated mapping tool, Stable Diffusion 3, with an early preview application channel. Stable Diffusion 3 features significant performance improvements in multi-subject cueing, image quality, and text rendering capabilities. Currently, the Stable Diffusion 3 model suite has parameters ranging from 800M to 8B, and the model uses the Sora-like Diffusion Transformer architecture.
Trial application address: http://stability.ai/stablediffusion3
2. Google Suspends Image Generation For Gemini Models
Google's artificial intelligence model, Gemini, has sparked controversy due to its alleged inability to generate historical images of white people accurately. As a result, Google has suspended the feature that allows Gemini to produce images of people. The decision came after Gemini users reported on social media that the model generated inaccurate images of historical figures, displaying them as people of color instead of white. This has raised concerns about the possibility of racial discrimination in AI.
3. Google launched the Open-source Model Gemma
Google recently released an open-source model called Gemma, which is focused on being lightweight and high-performance. It has two parameter scales of 2 billion and 7 billion and can be run on different platforms such as laptops, desktops, IoT devices, mobile devices, and the cloud. Regarding performance, Gemma outperformed the current mainstream open-source models Llama 2 and Mistral in 18 benchmarks, especially in math and code capabilities. It has also topped the Hugging Face open-source big model list.
Gemma open source address: https://www.kaggle.com/models/google/gemma/code
Technical report address: https://goo.gle/GemmaReport
4. Tesla Shares Video of Humanoid Robot Optimus Walking
Tesla's second-generation humanoid robot, Optimus, recently showcased its improved walking ability in a 1-minute and 18-second video shared on social media platforms. The video displays the robot walking around a test site more stably and smoothly than in the previous video released a few weeks ago.
5. NVIDIA AI Chip Rentals Getting Easier
Several companies that use AI chips have reported that renting NVIDIA's most advanced chips from cloud providers has become easier, according to an article in The Information. A large consumer company that rents servers from AWS and Google discovered that a few months ago, it could take weeks to get NVIDIA AI chips, but more recently the wait time has decreased.
AWS has introduced a new service that simplifies the process of renting NVIDIA chips, enabling customers to schedule GPU rentals for several days or weeks at a time. A Google spokesperson stated that Google can now meet almost all customer demands and is continuously expanding its online capacity.
6. Microsoft Open-sources Generative AI Risk Identification Tool PyRIT
Microsoft has recently launched an open-source automation framework called PyRIT. It is a Python-based risk identification toolkit that assists security professionals and machine learning engineers in identifying risks in generative AI systems. The tool has been used by Microsoft's AI Red Team to detect risks in generative AI systems, including the Copilot.
Here's how the PyRIT framework works: the PyRit Agent sends malicious cue words to the target generative AI system. When the generative AI system responds, the PyRit Agent sends the response to the PyRIT scoring engine. The scoring engine then sends the response back to the PyRit Agent, which uses the feedback to generate a new cue word. This automated process continues until the security expert obtains the desired result.
7. India's AI market to reach $17 billion by 2027, says report
India's AI market is expected to reach $17 billion by 2027, growing at a compound annual growth rate of 25-35 percent between 2024 and 2027, according to IT industry organization Nasscom and consulting firm BCG, citing a joint report examining AI technology in the country, Reuters reports. The growth is mainly due to increased spending on technology by businesses, the country's growing AI talent pool, and increased investment in AI, the report said.
8. Automotive Software And Electronics Market To Reach USD 469 Billion By 2033 With A CAGR of 5.5%
The global automotive software and electronics market is expected to grow at a CAGR of 5.5% from USD 274 Billion in 2023 to USD 469 Billion by 2033. The integration of electronics and software in automotive engineering has enhanced vehicle performance and safety, and boosted fuel efficiency. However, the high cost of infrastructure and complex safety and legal requirements in the automotive industry could hamper its growth.
