Accenture and Unilever Collaborate to Advance Artificial Intelligence Research and Technology Applications


The California state government has partnered with tech companies to invest $250 million to support journalism and AI research, with a plan to implement this over five years, including an initial allocation of $100 million in the first year. The funding is primarily aimed at developing California news organizations and AI research projects. This initiative aims to boost local news vitality and provide job opportunities, seen as a crucial breakthrough for the survival of journalism. The agreement has sparked controversy between the tech industry and lawmakers, with the initial proposal requiring tech companies to pay a percentage of advertising revenue in exchange for linking content, which faced strong opposition. The California News Publishers Association supports the agreement, believing it will benefit the local community.
An AI researcher from Oxford University has pointed out that large language models (LLMs) may pose a threat to scientific integrity. The research calls for a change in the use of LLMs, suggesting they be used as 'zero-shot translators' to ensure factual accuracy in outputs. Relying on LLMs as a source of information could jeopardize scientific truth, leading to calls for responsible use of LLMs. The study warns that casual use of LLMs in scientific papers could result in significant harm.
Qingdao has established the country's first "Individual + AI" venture capital fund, supporting individuals to start AI ventures with as little as 50,000 yuan, exploring the future model of "one-person companies," demonstrating forward-looking financial innovation and business model strategies.
Hume AI opensources the TADA speech generation model, which uses a text-acoustic dual alignment architecture, significantly improving the efficiency and reliability of TTS systems. By achieving 1:1 strict synchronization between text tokens and acoustic representations, it effectively solves the content hallucination problem in traditional LLM-based TTS. The model has been validated through more than a thousand samples and shows excellent performance.
Ali Tongyi Lab has recently undergone an organizational restructuring, splitting the Qwen team into multiple parallel lines such as pre-training and post-training. Subsequently, Yu Bowen, the former head of the post-training team of Qwen, was reported to have joined ByteDance, taking on the role of post-training lead for visual models and multimodal interaction within the Seed team. ByteDance has not officially responded yet.