AI News

Follow the latest AI and automation news

14 May 202627 Dhuʻl-Qiʻdah 1447 AH

China's 'Jiuzhang 4' Quantum Computer Achieves 10^54 Speedup Over Supercomputers

Researchers from the University of Science and Technology of China, led by Pan Jianwei, have developed a new quantum computing prototype named "Jiuzhang 4". This system features 1,024 squeezed state inputs and 8,176 modes, enabling it to manipulate quantum states of up to 3,050 photons. Jiuzhang 4 solves Gaussian boson sampling tasks at a speed 10^54 times faster than the fastest supercomputer, El Capitan. Published in Nature on May 13, this research marks a significant advancement in quantum computing capabilities.

Agents

14 May 202627 Dhuʻl-Qiʻdah 1447 AH

SenseTime Launches Embodied AI Convenience Store with Robotic Staff in Shanghai

SenseTime, through its subsidiary SenseMart, has launched a new AI convenience store named "Xiaomai Go" in the Baoshan district of Shanghai. The store attracted customers during the Labor Day holiday, utilizing robotic assistants to fulfill orders in as little as 15 seconds. It can handle 400 transactions daily without the need for cashiers or on-site managers. The store's system is based on five years of collected business data, including over 300,000 2D product models and 100,000 3D assets.

Models

14 May 202627 Dhuʻl-Qiʻdah 1447 AH

‘Basically a Chinese racket’: how China streaked ahead at a global AI conference

Chinese universities have overtaken American institutions in research output at the International Conference on Learning Representations (ICLR) held last month in Rio de Janeiro. They accounted for approximately 44% of the accepted papers, with Tsinghua University leading the contributions. This shift highlights China's rapid advancements in artificial intelligence, driven by significant government investments and a growing number of innovative research initiatives. The increasing presence of Chinese institutions at major conferences indicates a competitive edge in AI research.

💡

Helping ChatGPT better recognize context in sensitive conversations

How might safety updates influence ChatGPT's role in sensitive conversations?

Agents

Microsoft’s Edge Copilot update uses AI to pull information from across your tabs

13 May 202626 Dhuʻl-Qiʻdah 1447 AH

Agents

13 May 202626 Dhuʻl-Qiʻdah 1447 AH

Microsoft’s Edge Copilot update uses AI to pull information from across your tabs

Microsoft has introduced a new feature in its Edge browser, allowing its AI assistant, Copilot, to gather information from all open tabs. Users can now ask Copilot questions about the content in these tabs, compare products, and summarize articles. This feature enables users to choose their preferred experiences or disable those they do not need. Microsoft also announced the discontinuation of the previous Copilot mode, which offered similar features but with more autonomy, such as scheduling appointments.

Agents

Notion just turned its workspace into a hub for AI agents

13 May 202626 Dhuʻl-Qiʻdah 1447 AH

Agents

13 May 202626 Dhuʻl-Qiʻdah 1447 AH

Notion just turned its workspace into a hub for AI agents

Notion has launched a new developer platform that allows teams to connect AI agents with external data sources. This initiative aims to enhance productivity through agent-based software. The platform also supports the integration of custom code, providing greater flexibility for users. The platform facilitates access to data from multiple sources, improving collaborative efficiency. Teams can now leverage AI agents to streamline workflows, enabling them to focus on more critical tasks. This feature marks a significant shift in how software is utilized in workplace environments.

Agents

Google Innovates User Interaction with AI through Contextual Engineering

13 May 202626 Dhuʻl-Qiʻdah 1447 AH

Agents

13 May 202626 Dhuʻl-Qiʻdah 1447 AH

Google Innovates User Interaction with AI through Contextual Engineering

Google's AI subsidiary, DeepMind, is innovating user interaction with AI by transforming the mouse pointer into a key element of what it calls 'contextual engineering.' This approach reduces reliance on traditional text commands. Researchers Adrian Baranes and Rob Marchant from DeepMind explain that this new method allows users to interact more seamlessly with intelligent systems, enhancing the effectiveness of AI in daily applications. This shift comes as reliance on AI grows across various sectors.

Governance

What It Will Take to Make AI Sustainable

13 May 202626 Dhuʻl-Qiʻdah 1447 AH

Governance

13 May 202626 Dhuʻl-Qiʻdah 1447 AH

What It Will Take to Make AI Sustainable

Researcher Sasha Luccioni emphasizes the need for better emissions data and understanding AI usage. She argues that current data fails to accurately reflect the environmental impact of modern technologies. Improved mechanisms for data collection and analysis are essential for ensuring sustainable AI practices. Luccioni points out that existing carbon emissions data from AI is insufficient, complicating the assessment of its environmental effects. There is an urgent need to clarify how AI is utilized across various sectors, which will aid in developing more effective reduction strategies.

Infrastructure

mimalloc: A new, high-performance, scalable memory allocator for the modern era

13 May 202626 Dhuʻl-Qiʻdah 1447 AH

Infrastructure

13 May 202626 Dhuʻl-Qiʻdah 1447 AH

mimalloc: A new, high-performance, scalable memory allocator for the modern era

Microsoft's RiSE group has launched mimalloc, an open-source memory allocator designed as a drop-in replacement for malloc and free. It boasts fast allocation times and reduced contention, making it ideal for large applications. Mimalloc supports memory footprints exceeding 500 GiB and has significantly improved response times in services like Bing. Its efficient design allows it to be used effectively in various contexts, from small-scale applications to large services.

Business

Mark Zuckerberg announces ‘completely private’ encrypted Meta AI chat

13 May 202626 Dhuʻl-Qiʻdah 1447 AH

Business

13 May 202626 Dhuʻl-Qiʻdah 1447 AH

Mark Zuckerberg announces ‘completely private’ encrypted Meta AI chat

Meta CEO Mark Zuckerberg announced the launch of Incognito Chat, claiming it is the first major AI product that does not store any logs of user conversations. Unlike other chatbots, this service does not save messages in chat history, providing a similar experience to incognito modes but with added end-to-end encryption. Zuckerberg emphasized that Incognito Chat ensures true privacy, as no one, not even Meta, can read users' conversations. He noted that while other apps have introduced similar modes, they still have access to incoming questions and outgoing answers.