AI News

Follow the latest AI and automation news

28 Apr 202611 Dhuʻl-Qiʻdah 1447 AH

Introducing talkie: a 13B vintage language model from 1930

A team of researchers led by Nick Levin, David Dufinaud, and Alec Radford has launched Talkie, a 13 billion parameter language model trained on 260 billion tokens of historical English texts predating 1931. The project features two models: talkie-1930-13b-base and talkie-1930-13b-it, with the latter designed to operate a chat interface using a new dataset of instruction-response pairs. Both models are licensed under Apache 2.0.

Models

28 Apr 202611 Dhuʻl-Qiʻdah 1447 AH

Alibaba Opens Beta Testing for HappyHorse 1.0 Video Generation Model

On April 27, Alibaba announced the beta launch of its video generation model, HappyHorse 1.0. The model is available to professional creators and enterprise users via registration, while general users can access it through the Qwen App. HappyHorse 1.0 supports text-to-video, image-to-video, and audio generation, producing videos up to 15 seconds long with multi-shot storytelling. The reference price for 720p video generation is RMB 0.9 per second, accessible by updating the Qwen App.

Agents

28 Apr 202611 Dhuʻl-Qiʻdah 1447 AH

Google is testing AI chatbot search for YouTube

Google is testing a new conversational search feature on YouTube, allowing users to receive results that include longform videos and Shorts. This experiment is currently available to YouTube Premium subscribers in the U.S. aged 18 and older. When activated, users will see an 'Ask YouTube' button in the search bar, enabling them to pose questions like 'funny baby elephant clips' or 'summary of volleyball rules.' This shift represents a significant change in user interaction with the platform, highlighting Google's focus on enhancing search experiences.

💡

OpenAI models, Codex, and Managed Agents come to AWS

How might OpenAI's new models impact Arab enterprises?

Governance

Our commitment to community safety

28 Apr 202611 Dhuʻl-Qiʻdah 1447 AH

Governance

28 Apr 202611 Dhuʻl-Qiʻdah 1447 AH

Our commitment to community safety

OpenAI has announced new measures to enhance community safety in ChatGPT. These measures include model safeguards, misuse detection, and policy enforcement in collaboration with safety experts. This initiative is part of the company's ongoing efforts to ensure a safe environment for users. The organization is developing its models to be safer by implementing advanced techniques for detecting misuse. Clear and effective policies are also a priority, helping protect users from potential risks. Collaboration with safety experts strengthens the effectiveness of these policies.

Models

microsoft/VibeVoice

27 Apr 202610 Dhuʻl-Qiʻdah 1447 AH

Models

27 Apr 202610 Dhuʻl-Qiʻdah 1447 AH

microsoft/VibeVoice

Microsoft launched the VibeVoice speech-to-text model on January 21, 2026, featuring MIT licensing and integrated speaker diarization. This model was tested on a Mac using mlx-audio tools, demonstrating its effectiveness with various audio files. VibeVoice allows users to convert audio files into text with high accuracy, evidenced by its application in a podcast episode with Lenny Rachitsky. The results showed a processing time of 524.79 seconds, generating 20,248 tokens at a rate of 38.585 tokens per second, indicating the model's efficiency with lengthy audio.

Business

Elon Musk and Sam Altman are going to court over OpenAI’s future

27 Apr 202610 Dhuʻl-Qiʻdah 1447 AH

Business

27 Apr 202610 Dhuʻl-Qiʻdah 1447 AH

Elon Musk and Sam Altman are going to court over OpenAI’s future

This week, Elon Musk and OpenAI CEO Sam Altman face trial in Northern California, a case that could reshape OpenAI's future. Musk alleges that Altman and President Greg Brockman misled him about the company's shift from a nonprofit to a for-profit entity, seeking damages of up to $134 billion. OpenAI was founded in 2015 as a nonprofit with Musk's $38 million donation, promising to develop open-source technology. However, the company shifted its focus towards profitability, raising concerns about transparency and competition in the AI sector.

Governance

Google Employees Demand Rejection of AI Use in Military Operations

27 Apr 202610 Dhuʻl-Qiʻdah 1447 AH

Governance

27 Apr 202610 Dhuʻl-Qiʻdah 1447 AH

Google Employees Demand Rejection of AI Use in Military Operations

More than 560 Google employees signed an open letter to CEO Sundar Pichai, urging him to reject any collaboration that would allow the U.S. government to use the company's AI technologies in covert military operations. This letter comes amid growing concerns about the military use of AI. The employees emphasize their desire to harness these technologies for the benefit of humanity, rather than for purposes that threaten peace.

Business

China Blocks Meta's $2 Billion Acquisition of AI Startup Manus

27 Apr 202610 Dhuʻl-Qiʻdah 1447 AH

Business

27 Apr 202610 Dhuʻl-Qiʻdah 1447 AH

China Blocks Meta's $2 Billion Acquisition of AI Startup Manus

Chinese authorities have blocked Meta's planned $2 billion acquisition of AI startup Manus. The National Development and Reform Commission cited regulatory requirements for the cancellation, amid rising tech tensions between China and the U.S. Concerns have grown that China might acquire U.S. companies' intellectual property in AI, prompting this decision. This reflects the escalating tensions between the two tech giants as they seek to protect their national interests.

Business

Ineffable Intelligence Secures $1.1 Billion in Major European Funding Round

27 Apr 202610 Dhuʻl-Qiʻdah 1447 AH

Business

27 Apr 202610 Dhuʻl-Qiʻdah 1447 AH

Ineffable Intelligence Secures $1.1 Billion in Major European Funding Round

Ineffable Intelligence, founded by David Silver, has raised $1.1 billion in a seed funding round, marking the largest deal of its kind in Europe. The funding round valued the company at approximately $5.1 billion and was led by Sequoia Capital and Lightspeed Venture Partners, with backing from the UK government. Major players such as Nvidia and Alphabet participated in the round, reflecting the intensifying competition in artificial intelligence.