AI News

Follow the latest AI and automation news

12 May 202625 Dhuʻl-Qiʻdah 1447 AH

[AINews] Thinking Machines' Native Interaction Models - TML-Interaction-Small 276B-A12B - advances SOTA Realtime Voice and kills standard VAD

Thinking Machines has launched new interactive models, TML-Interaction-Small, featuring 276 billion parameters. These models aim to enhance AI interaction with users by processing audio and visual inputs in under 200 milliseconds. The new models introduce internal standards for measuring temporal awareness and simultaneous translation, such as TimeSpeak and CueSpeak. These standards enable the model to interact more accurately with users, thereby improving the effectiveness of AI in everyday applications.

Business

12 May 202625 Dhuʻl-Qiʻdah 1447 AH

M42 Announces New Clinical LLM to Transform the Future of AI in Healthcare

M42 has launched Med42, an open-access clinical language model capable of processing 70 billion parameters, significantly enhancing patient care outcomes. Developed using a curated medical dataset from M42, Med42 delivers high-quality answers to medical inquiries. Tests reveal that it outperforms larger closed models, such as ChatGPT 3.5, achieving a score of 72% on the U.S. Medical Licensing Exam. Med42 represents a crucial advancement in clinical decision-making and increases access to focused medical knowledge for healthcare professionals and patients, potentially transforming healthcare delivery in the region.

Business

12 May 202625 Dhuʻl-Qiʻdah 1447 AH

Kuaishou plans to spin off Kling AI in $2 billion funding round with Tencent

Kuaishou is planning to spin off its AI model, Kling AI, aiming to raise $2 billion at a $20 billion valuation. Tencent is reportedly in discussions to join this funding round, highlighting the significant interest from major players in AI development. This spin-off could enhance Kuaishou's focus on advancing Kling AI, potentially opening new avenues in the AI market and strengthening its competitive position.

💡

What Parameter Golf taught us about AI-assisted research

What innovations might emerge from this conference?

Business

AutoScout24 scales engineering with AI-powered workflows

12 May 202625 Dhuʻl-Qiʻdah 1447 AH

Business

12 May 202625 Dhuʻl-Qiʻdah 1447 AH

AutoScout24 scales engineering with AI-powered workflows

AutoScout24 Group has announced its use of Codex and ChatGPT to enhance code quality and accelerate development cycles. These tools improve development team efficiency by providing intelligent solutions that speed up coding processes and reduce errors. The adoption of AI in development reflects a growing trend towards innovation in the tech sector. This move is significant for tech companies aiming to boost performance and productivity. By integrating AI, AutoScout24 can enhance its competitive edge in the market.

Agents

How NVIDIA engineers and researchers build with Codex

12 May 202625 Dhuʻl-Qiʻdah 1447 AH

Agents

12 May 202625 Dhuʻl-Qiʻdah 1447 AH

How NVIDIA engineers and researchers build with Codex

Teams are utilizing Codex alongside GPT-5.5 to transform research ideas into executable experiments. This integration accelerates the development of production systems, enhancing overall work efficiency. The anticipated impact includes significant improvements in team performance and productivity. Codex is part of a suite of AI tools that empower developers to create advanced systems. By leveraging GPT-5.5, teams can harness deep learning capabilities to enhance software quality. This collaboration signifies a crucial step towards innovation in software development.

Agents

Cursor in Microsoft Teams

11 May 202624 Dhuʻl-Qiʻdah 1447 AH

Agents

11 May 202624 Dhuʻl-Qiʻdah 1447 AH

Cursor in Microsoft Teams

Microsoft has launched the Cursor feature in Teams, enabling users to delegate tasks to a cloud agent. By mentioning @Cursor in any Teams channel, users can pull information or execute tasks seamlessly. Cursor automatically selects the appropriate repository and model based on user prompts and recent agent activity. It reads the entire thread for context before implementing solutions and creating a pull request for team review.

Governance

OpenAI just released its answer to Claude Mythos

11 May 202624 Dhuʻl-Qiʻdah 1447 AH

Governance

11 May 202624 Dhuʻl-Qiʻdah 1447 AH

OpenAI just released its answer to Claude Mythos

OpenAI has launched Daybreak, an initiative aimed at detecting and patching security vulnerabilities before attackers exploit them. This initiative utilizes the Codex Security AI agent, introduced in March, to create a threat model based on an organization's code and identify potential attack paths. Daybreak automates the detection of higher-risk vulnerabilities by validating likely threats, allowing organizations to prioritize their security efforts effectively. This launch follows Anthropic's announcement of Claude Mythos, a security-focused AI model deemed too dangerous for public release.

Business

GM just laid off hundreds of IT workers to hire those with stronger AI skills

11 May 202624 Dhuʻl-Qiʻdah 1447 AH

Business

11 May 202624 Dhuʻl-Qiʻdah 1447 AH

GM just laid off hundreds of IT workers to hire those with stronger AI skills

A new set of job openings has been announced, focusing on AI development, data engineering, and analytics. These positions cover areas such as cloud engineering, model development, and prompt engineering, highlighting the growing demand for AI-related skills. The need for professionals skilled in advanced data analysis and engineering reflects current market trends. Companies are increasingly seeking experts who can navigate complex AI technologies, emphasizing the importance of these roles in driving innovation.

Models

Here’s what Mira Murati’s AI company is up to

11 May 202624 Dhuʻl-Qiʻdah 1447 AH

Models

11 May 202624 Dhuʻl-Qiʻdah 1447 AH

Here’s what Mira Murati’s AI company is up to

Thinking Machines, founded by former OpenAI CTO Mira Murati, has announced its development of new interaction models. These models aim to allow users to collaborate with AI in a more natural way by continuously processing audio, video, and text in real time. The interaction models will enable simultaneous information processing, allowing AI to understand user behavior and provide immediate responses. This shift in how models interact with users is a significant step toward enhancing user experience and improving AI effectiveness.