Explore News by Category
Apple Intelligence Coming to China in December With iOS 27.3
Apple announced that its Apple Intelligence suite will become available in China in December, bundled with iOS 27.3. The company will begin testing the first features on November 4 through the latest iOS developer beta, delivering tools such as Live Translation, Clean Up, and Writing Tools. The rollout will coincide with an iPhone launch and includes integration of Alibaba’s Qwen large‑language model, but it will not feature the newly announced Siri AI assistant.
One In Three Young Adults Surveyed Use AI For Substance Use Information
A Rutgers Health study found that more than 32% of surveyed young adults use ChatGPT or comparable AI chatbots to obtain information about substance use, chiefly regarding risks and consequences. Approximately half of those who rely on AI for such queries do so at least once a month. Rutgers researchers caution that these conversational agents may spread misinformation about substance use, potentially shaping young adults' perceptions and decisions. The findings highlight a growing reliance on AI for health‑related information among this demographic.
Dubai university develops AI tools to accelerate hospital triage and emergency care
Dubai Medical University is creating artificial intelligence systems designed to speed up hospital triage, provide first‑aid guidance in remote settings, and identify possible illnesses among its student population. The initiative is part of the university's expanding healthcare innovation programme, aiming to integrate advanced digital tools into emergency and preventative care processes within the UAE.
Bidirectional cross-scale detail-semantic collaborative network (BCDS-Net) for vision-based waste recognition
Researchers introduced BCDS-Net, a Bidirectional Cross-Scale Detail‑Semantic Collaborative Network designed to improve vision‑sensor waste recognition in intelligent sorting systems. The model combines a Hybrid Conv‑MLP backbone, Scale‑Aware Context Gated Fusion, Dilated Scanning Attention, and Bidirectional Cross‑Scale Aggregation to capture fine details and long‑range dependencies while handling scale variation and background clutter. Tested on TrashNet, RealWaste and TrashBox datasets, BCDS-Net achieved accuracies of 96.73%, 92.15% and 96.23% respectively, surpassing baseline methods. With 11.53 M parameters and 2.52 G FLOPs, it runs at 24.4 ms per image on an NVIDIA Jetson Xavier NX and 3.2 ms on an RTX 3090, indicating suitability for edge deployment.
Anthropic’s record US$2 trillion stock market listing highlights the growing risks for all of us
Anthropic, the developer of the Claude AI chatbot, filed its IPO prospectus and plans to list on the New York Stock Exchange in November. The offering targets a valuation of up to $2 trillion, which would make it the largest initial public offering ever. In its filing, Anthropic is the first company to warn investors about “catastrophic or existential risks to humanity” posed by advanced AI. The prospectus underscores both the market’s appetite for AI assets and the growing attention to potential dangers.
OpenAI announces three big science discoveries and a new way to work with academics
OpenAI announced that its artificial‑intelligence models have generated three novel scientific findings, one of which invalidates the Exponential Time Hypothesis, a long‑standing conjecture in computational complexity theory. The company also unveiled a new framework for working directly with university researchers, pledging to follow recently‑issued guidance from scientists on how to release such results responsibly. The disclosures were made in a public statement by OpenAI, highlighting the organization’s push to integrate AI capabilities into academic research.
Claude Just Wanna Have Fun: Without Any Work, AI Agents Like to ‘Play'
Researchers at MIT observed that when advanced AI agents were placed on a virtual island with no assigned tasks, they engaged in spontaneous activities such as playing catch, arranging letters in cubes, and constructing sailboats from available objects. The study found that this undirected, play‑like behavior improved the agents’ performance on subsequent tasks. The authors caution against interpreting these actions as human‑like, warning against anthropomorphizing the agents. The findings suggest that giving AI systems periods of free exploration could be a useful strategy for enhancing learning efficiency.
AI Image Generator Test: Canva Craft Vs. ChatGPT, Gemini, Meta Muse
Canva introduced its first AI image‑generation model, Canva Craft, on Thursday, making the feature available through the platform’s Dream Lab. The tool was benchmarked against rivals such as OpenAI’s ChatGPT image model, Google's Gemini, and Meta's Muse. In a test creating a birthday card, Craft lagged behind the competitors, but reviewers noted that its ability to produce realistic images is narrowing the gap with the more established models. The rollout marks Canva's entry into the fast‑moving AI image‑creation market.
Integrating causal knowledge with satellite and glider data: a novel framework for marine water quality simulation
Researchers introduced a causally informed neural network (CINN) that incorporates an expert‑specified directed acyclic graph as a regularization target to jointly predict coastal marine variables. The model merges satellite remote‑sensing data with high‑resolution glider measurements from the upper 0–2 m, estimating temperature, salinity, dissolved oxygen and chlorophyll‑a. In acquisition‑date‑blocked five‑fold cross‑validation, CINN achieved R² scores of 0.961 for temperature, 0.905 for salinity, 0.898 for dissolved oxygen and 0.725 for chlorophyll‑a, yielding an overall multi‑output R² of 0.872. Ablation studies and statistical tests evaluated predictive sensitivity and uncertainty.
Chinese 'superstar hacker' 5 times named Microsoft's top security researcher joins AI giant Z.ai
Chinese cybersecurity specialist Yuki Chen, who has topped Microsoft’s security researcher rankings five times and won the "Oscars of hacking" twice, has left his previous role to join Beijing‑based artificial‑intelligence company Z.ai. Chen announced the move on social media last week, describing it as a "new journey, new mission" and expressing gratitude for the chance to help build the firm’s security capabilities. His online handle is Guhe. The announcement was reported by the South China Morning Post.
