Grasp the pulse of AIGC and grasp the pulse of science and technology. TechNode collects and summarizes the progress and hot spots of global AIGC every day, and takes you to understand AIGC for 5 minutes every day, hoping to play AIGC with you, decode the new trend of industry development, and open a new era of wisdom!
Text|TechNodeTypography|Li TingtingThis article is expected to take 5 minutes to read
01
Meta has released a new generation of AI training and inference chips with three times the performance of the original chip
Meta Platforms released the latest version of its Training and Inference Accelerator Project (MTIA), a family of custom chips designed by Meta specifically for AI workloads, on the 10th local time. According to reports, the new generation of MTIA released this time has significantly improved performance compared with the first generation MTIA, and helps strengthen the content ranking and ad recommendation model. Its architecture is fundamentally focused on providing the right balance of compute, memory bandwidth, and memory capacity. The chip can also help improve training efficiency and make inference (i.e., actual inference tasks) easier.
02
A 10-second summary YouTube video in which Jia Yangqing, the former chief AI scientist at Alibaba, builds Elmo, a browser plug-in
Jia Yangqing, the former chief AI scientist of Alibaba, shared the plug-in Elmo on X, which can summarize the Google Next keynote speech in 10 seconds, generating a one-sentence summary, summary, and main ideas. The plugin was created by Lepton AI, an AI company founded by Jia Yangqing last year. According to Jia, Elmo uses DBRX, an open-source model from Databricks, a data company. It is reported that DBRX has 132 billion parameters and uses a MoE architecture, surpassing GPT-3.5 and some other open-source models in performance.
03
South Korea plans to invest 9.4 trillion won in AI semiconductors by 2027, aiming to become the top three in AI technology According to Yonhap News Agency, South Korean President Yoon Suk-yeol said at South Korea's "Semiconductor Field Pending Issues Conference" held yesterday that the South Korean government plans to invest 9.4 trillion won in the field of AI semiconductors by 2027. In addition to direct investment, South Korea plans to create a fund worth 1.4 trillion won to help the development of AI semiconductor startups.
04
Google Launches CodeGemma AI Model Google Inc. recently released CodeGemma Advanced Large Language Models (LLMs), which are designed to generate code, understand and track instructions, etc., with the aim of making high-quality code assistance tools more accessible to developers around the world. CodeGemma is launched as an open-access series of Gemma models that are optimized and fine-tuned for code development scenarios. The suite includes three different models: ◾️the 2 billion parameter base model: Optimization Completion and Code Generation to provide a fast and efficient solution for the code, primarily with a focus on latency and privacy. ◽️7 Billion Parameter Base Model: Integrates code completion and natural language processing to enhance its usefulness in completing code and understanding language generation tasks. ◾️ 7 Billion Parameter Models for Guided Traces: Users engage in conversations about code, programming, and mathematical reasoning, and are a valuable resource for developers seeking guidance or comments.
05
Up to $7.25/min, Adobe buys video to build AI models to compete with Sora
Adobe has begun sourcing video to build its AI-powered text-to-video generator, and after OpenAI demonstrated a similar technology (Sora), Adobe tried to catch up with competitors. The software house is offering $120 each to its network of photographers and artists, and Adobe requires them to submit videos of people engaging in everyday activities, such as walking, or expressing various emotions, including joy and anger. The company said it was doing this to collect data for AI training. Adobe asked for more than 100 short films of people involved in action and emotion, as well as simple anatomical footage of feet, hands, or eyes. The company also wants to get videos of people "interacting with objects such as smartphones or gym equipment." It warns against providing copyrighted material, nudity, or other "offensive content." Users are paid around $2.62 per minute for submitting videos, up to $7.25 per minute. Adobe tries to differentiate its models by training marketers and creative agencies with its vast library of stock media. If its inventory is insufficient, it gets images directly from contributors.
06
OpenAI CEO Altman wants to build a global AI alliance to accelerate the development of artificial intelligenceOpenAI CEO Altman met with investors and government officials in the UAE this week, according to Bloomberg, Altman has been working to establish a global AI alliance between the government and industry leaders to increase the supply of chips, energy and data center capacity needed to develop AI technology, more infrastructure investment, and promote the development of artificial intelligence. Altman discussed with investors and government officials how the private sector can work with countries, and he also spoke with officials from a number of Western countries and will meet in Washington later this week. In an attempt to solve the problem of poor supply of chips and other critical infrastructure, Altman has been trying to raise billions of dollars from global investors to build a network of factories that make semiconductors.
07
At the just-concluded spring press conference, Huawei applied the Huawei Pangu model for the first time in the new MateBook X Pro, empowering Huawei's smart PC in all scenarios. Huawei's AI Spatial feature allows users to directly access a wide range of AI applications with one click, selecting 100+ agents, and creating a one-stop AI capability aggregation portal. Huawei is also working with third-party partners to build more comprehensive intelligence capabilities for the MateBook X Pro. The super terminal has been upgraded again, and the multi-screen collaboration supports P3 wide color gamut projection, and P3 wide color gamut projection with Huawei mobile phones, and the color and detail performance of the picture are consistent. In addition, the new MateBook X Pro has been deeply optimized for a series of scenarios such as office, meeting, learning, creation, and life.
This article is organized by TechNode, and may not be reproduced without authorization, if you need to reprint or open up, please reply "reprint" in the background.
- - - - - - END - - - - - - If you want to get industry information and share your experience with like-minded technology enthusiasts, then scan the code to add "Movedian" to join the group chat!
Wonderful article is worth recommending!
One-click forwarding, poke and watch!
