Australia embraces artificial intelligence and seeks to become a world leader in the field, Minister of Industry and Science says
Australia's Minister of Industry and Science Tim Ayres stressed that artificial intelligence (AI) is key to the country's future economic growth and competitiveness. The government will promote the development and application of AI technology, accelerate the transformation of the digital economy, promote high value-added industries and employment, while paying attention to social equity and risk management to ensure that Australia takes the initiative in the global digital age.
Sam Altman invests in automatic delivery cargo robots and raises another $80 million
Los Angeles startup Coco Coco Robotics focuses on last-mile autonomous delivery robots and recently completed a US$80 million financing, bringing its total financing to more than US$120 million137. The round was led by venture capital firm SNR, with return investors including OpenAI CEO Sam Altman and his brother Max Altman, Pelion Venture Partners and Offline Ventures also participating136.
Coco Robotics’ electric delivery robot is zero-emission and has a capacity of approximately 90 liters. Since it was put into use in 2020, it has completed more than 500,000 deliveries, serving cities including Los Angeles, Chicago, Miami and Helsinki, and partners include Subway, Wingstop, Jack in the Box, DoorDash and Uber Eats.
Google's "Project Green Light" project uses artificial intelligence (AI) and machine learning (ML) technologies, especially by analyzing traffic data from Google Maps, to optimize the control of urban traffic lights, thereby reducing traffic congestion and frequent vehicle starts and stops.
【Tech Express】ByteDance's blockbuster release! Doubao 1.6 and Seedance 1.0 pro models debut, aiming at the new era of AI agents
ByteDance officially released the new Doubao 1.6 large model and Seedance 1.0 professional video generation model at its "Force Conference" yesterday (June 11). This release is another major progress of ByteDance in the field of AI, aiming to comprehensively accelerate the development and large-scale application of AI agents through performance improvements and more models.
Magistral series: Europe's first "reasoning model"
The Magistral series of models continues to attract attention. Among them, Magistral Small is an open source model with 2.4 billion parameters, using the Apache 2.0 license agreement, and is now available for download on Hugging Face. Magistral Medium is a closed-source enterprise-level model designed for logical reasoning and complex tasks. It is developed based on reinforcement learning technology and its performance is comparable to similar products of OpenAI and DeepSeek.
The company behind it is valued at approximately US$6.2 billion, and the total amount of the latest round of financing exceeded US$600 million. In addition to Magistral, the company has also launched a number of models: including Mistral Small and Large, Pixtral Large that supports multimodality, and Codestral for code tasks, which are fully adapted to cloud, edge computing and multi-language application scenarios.
Alibaba Tongyi Lab Releases MaskSearch
MaskSearch is a general pre-training framework launched by Alibaba Tongyi Lab in May 2025, which aims to improve the active retrieval and multi-step reasoning capabilities of large language models (LLMs). Its core innovation is the RAMP (Retrieval Augmented Mask Prediction) task, which encourages the model to fill in the masked key information in the text and obtain the required knowledge by actively calling the search engine.
Disney and Universal sue AI firm Midjourney for copyright infringement
On June 11, 2025, Disney and Universal jointly filed a copyright infringement lawsuit in a federal court in Los Angeles against AI image generator Midjourney. They allege the platform unlawfully trained its models using copyrighted content owned by the studios and produced AI-generated images that replicate iconic characters—such as Darth Vader, the Minions, Spider-Man, Shrek, Elsa, Yoda, Buzz Lightyear, Princess Elsa, and Homer Simpson—without authorization
The complaint describes Midjourney as a “copyright free-rider” and “a bottomless pit of plagiarism,” accusing it of creating “endless unauthorized copies” by scraping copyrighted works from across the internet. Even generic prompts like "Superhero fight scene" can trigger outputs closely resembling Disney’s Spider‑Man. Despite repeated requests to cease generating infringement, Midjourney allegedly continued to promote and distribute these images
This case marks a landmark escalation in the legal battle between Hollywood studios and generative AI technologies. It could redefine legal precedent around AI training data and copyright. Disney and Universal are seeking injunctive relief, damages, and a jury trial. The Motion Picture Association and the RIAA have publicly supported the lawsuit, noting the necessity of safeguarding intellectual property in an era of rapid AI progress
Wikipedia pauses AI-generated summaries pilot after editors protest
Wikipedia has paused its experiment of using AI to generate article summaries on its platform following pushback from editors. The experiment was available to users with the Wikipedia browser extension who opted in, showing AI-generated summaries at the top of articles with a yellow "unverified" label that users could click to expand. Editors quickly criticized the pilot, expressing concerns it could harm Wikipedia's credibility because AI summaries often contain errors due to AI "hallucinations." Similar experiments by news outlets like Bloomberg have led to corrections and scaling back of tests. While the pilot is paused, Wikimedia remains interested in AI-generated summaries for use cases such as improving accessibility
Meta Unveils Advanced AI Model V-JEPA 2 to Enable Robots to "Think Before They Act
Meta has introduced V-JEPA 2, a state-of-the-art AI "world model" trained primarily on video data, designed to help robots and AI agents understand the physical world and predict how it will respond to their actions. Trained on over one million hours of video, the model learns patterns of object motion, human-object interaction, and object-object interaction, enabling robots to perform tasks like reaching, picking up, and placing objects in new environments with success rates between 65% and 80%. V-JEPA 2 uses a two-stage training process: an initial actionless self-supervised phase to learn world dynamics, followed by action-conditioned training linking visual inputs to robot control. Meta also released three new benchmarks to evaluate physical reasoning in AI models. By open-sourcing V-JEPA 2, Meta aims to accelerate research in robotics and embodied AI toward advanced machine intelligence that can think before it acts
Amazon’s Chief Security Officer Steve Schmidt, along with other major tech leaders, has urged a hands-off approach to AI regulation, warning that government intervention could hinder innovation and progress. Schmidt emphasized that regulation tends to slow development and that industry standards should be set by market demand rather than government mandates. This stance comes as the U.S. Senate debates whether to keep a federal provision blocking states from imposing new AI rules. Previously, the Trump administration rescinded Biden’s AI regulatory executive order, shifting toward a more pro-growth and innovation-friendly policy. While the tech industry calls for a national AI regulatory framework to avoid patchwork rules, critics argue Congress has failed to pass meaningful legislation, leaving states to fill the regulatory void.