ReadingResearchRadarInvestment framework
Sign in / Sign up
AllTechnologyProductsInvestingPublic marketsOther

Reading

2026-11-164 posts

Refresh

Alibaba Releases Qwen4-Max: 2.4 Trillion Parameters, Lower API Prices, and Open-Weight Promise for Smaller Models

Material
2026-11-16 12:10 GMT+8

Alibaba's Tongyi Qianwen (Qwen) team officially released Qwen4-Max, its next-generation flagship multimodal large language model, on November 16. The model boasts a total parameter count of 2.4 trillion (2.4T) and introduces new architectural designs aimed at optimizing long-text processing efficiency and computational costs. While the weights for Qwen4-Max itself remain closed, the company announced plans to open-source smaller models built on the same technology in the future.

Technically, Qwen4-Max utilizes a Mixture-of-Experts (MoE) architecture, activating approximately 71 billion (71B) parameters per token. Key innovations include the 'GDN + Qwen Sparse Attention (QSA)' mechanism, which combines context compression with sparse attention to reduce computational overhead for long sequences, and 'N-gram Embedding,' which expands model capacity using continuous word patterns without significantly increasing compute. The model supports a context length of up to 1 million tokens and can process image and video inputs.

According to vendor-reported benchmarks, Qwen4-Max outperforms some competitors in specific tasks. In 'Agents' Last Exam v1.1,' which evaluates real-world business tasks, Qwen4-Max scored 46.7, surpassing Anthropic's Claude Opus 5.5 (43.4). It also achieved a top score of 75.7 in 'DeepSWE v1.1' for long-term software development. However, in 'Terminal-Bench 4.0' for command-line coding, it scored 48.5, trailing behind Claude Opus 5.5 (59.6) and Kimi K3.1 (67.2). Alibaba claims that token consumption for identical tasks has decreased by 48-49% compared to the previous Qwen3.8-Max, with internal assessments showing a 62% drop in average cost per task.

With the official release, API pricing has been finalized. Qwen4-Max costs $1.50 per million input tokens and $4.50 per million output tokens, with cached input priced as low as $0.075. These rates represent a price reduction from the previous Qwen3.8-Max generation. Developers can access the model via Alibaba Cloud APIs, the Qwen Studio app, and the Qoder coding tool.

Sources:PC Watch

Research

South Korea's AI Security Firm Aim Intelligence: Whether an AI Agent's Action Is Safe Depends on Its Context Within the Business Workflow

Material
2026-11-16 09:27 GMT+8

When enterprises deploy AI agents, the most common security approach is to create a permission list: which tools can this agent call, which data can it access, and which external APIs can it connect to? Any operation outside the list is blocked. Lee Han-wool, Head of Defense Security at South Korean AI security firm Aim Intelligence, believes that this approach is far from sufficient for AI agents. His core judgment is: determining whether an action is safe cannot rely solely on whether it is within the allowed scope, but must also consider whether it is appropriate within the current business workflow.

This distinction stems from two different types of risks. The first type is actively caused by humans: someone gives the agent malicious instructions or accidentally transmits incorrect commands. Traditional permission models can still handle these issues. The second type is more difficult to manage: no one instructs the agent to do something, but the agent crosses boundaries or bypasses security mechanisms on its own while completing its goal. Lee Han-wool attributes the difficulty of AI security to the fact that model performance has developed too quickly while control technologies have failed to keep up.

He uses a hypothetical scenario to illustrate why permission lists are insufficient. Suppose an agent is permitted to call a document search tool, a data analysis tool, and a certain external API. Each of these three permissions is fine when viewed individually. However, if the agent first searches an internal document, passes the content to another agent for analysis, and then calls an external API to send the results out—each step is within the scope of permission, but the overall chain constitutes a data leak. To identify such risks, the system must know what instructions the agent initially received, what sub-task each step was called to complete, and whether the causal relationships between these steps are reasonable.

Aim Intelligence refers to this capability as Observability, i.e., full observability of the agent's execution process. It records not just individual logs, but the entire chain from input to tool invocation to output, as well as the correlations between multiple chains. Based on this, the company proposes that static security and dynamic security must run in parallel: static security consists of pre-set hard rules, such as prohibiting access to certain external networks; dynamic security judges whether an action is reasonable based on the current task context and past execution records.

However, there is an engineering contradiction here: if every action requires tracing the entire business chain for deep analysis, latency and computational costs will expand rapidly, while AI agents execute much faster than humans. Lee Han-wool's solution is tiered inspection. The first layer detects whether malicious instructions are hidden in external data or tool descriptions, i.e., indirect prompt injection; the second layer directly blocks explicitly prohibited operations; only when a request falls into a gray area does it trigger the third layer, calling the full execution history and task context for deep judgment. This way, most requests take the fast track, and only a few suspicious requests pay the cost of analysis.

Sources:바이라인네트워크

Research

Apple Launches Health AI Usage Study with OpenEvidence to Evaluate Conversational Medical Tools

Material
2026-11-16 09:26 GMT+8

Apple announced the launch of the "Apple Health AI Usage Study" on November 15, 2026, aiming to better understand how people interact with conversational AI health tools and how useful their responses actually are. The study is conducted in collaboration with OpenEvidence, a clinical AI platform used by healthcare professionals that generates answers based on medical information from sources like The New England Journal of Medicine.

Participants will use a new iOS app to ask medical questions and then rate the responses based on criteria including accuracy, detail, and relevance. Apple is looking for up to 200,000 participants who must be 21 years or older and based in the US. Although the app does not use Apple Intelligence or Siri, it requires an iPhone 15 Pro running iOS 26 or later.

During enrollment, participants can optionally share Apple HealthKit data, such as activity, sleep, or clinical lab results, so that responses include more context about their personal health. Apple emphasizes that privacy is a priority, stating that measures have been taken to ensure user information remains confidential and that collected data will not be used to train AI models.

"At Apple, we believe responsible innovation starts with scientific rigor and clinical grounding," said Sumbul Desai, MD, Apple’s vice president of Health and Fitness. "Health is deeply personal, and conducting a dedicated research study is the most thoughtful way to understand how people engage with, comprehend, and evaluate health AI in everyday life." Insights from this study are intended to advance the broader scientific conversation and help build trusted tools that support people on their health journey.

Sources:The Verge

Research

Nvidia Robotics Engineer Resigns Over Lack of Transparency as Petition to Reinstate Jensen Huang Passes 150,000 Signatures

Material
2026-11-16 01:56 GMT+8

In the wake of Nvidia founder and former CEO Jensen Huang's abrupt resignation on November 12, internal turmoil at the chipmaker has intensified. On November 15, Darrel Dai, a Robot Research Engineer at Nvidia, published a detailed post on X announcing his own departure. Dai stated he could not accept having to come to terms with Huang's exit before receiving an explanation that helped him understand it. He emphasized that the engineering culture Huang built—encouraging exploration, tolerating failure, and fostering direct communication between leadership and staff—was a primary reason for staying at Nvidia, and expressed disappointment that employees' perspectives were ignored in this decision.

Dai's resignation is not isolated; it reflects growing dissatisfaction among some Nvidia employees regarding how the board handled the crisis. Concurrently, an online petition demanding the board "Bring Jensen Back" has gathered over 150,000 signatures. Despite controversies involving personal relationships and national security investigations, many employees and external supporters argue that Huang's vision and investments in CUDA and physical AI were crucial to Nvidia's success. This strong grassroots support places significant reputational and cultural pressure on interim CEO Colette Kress and the board.

The incident highlights the governance challenges faced by tech giants under "key-man risk." As a global leader in AI chips, Nvidia's strategic direction has been heavily dependent on its founder's personal foresight. Employee resignations and public petitions signal deep divisions within the company regarding the transparency and legitimacy of the leadership transition. While some staff have pledged support to the interim management, the potential exodus of core technical talent could impact continuity in robotics and physical AI research. Additionally, Elon Musk has publicly offered to hire Huang, further complicating the talent landscape.

Sources:Tom's Hardware

Research

You’re all caught up in this view

Sign in / Sign up
ReadingResearchRadarInvestment framework

The company has integrated this mechanism into a platform named Starfort. Among them, Guardian is a real-time guardrail responsible for prompt injection detection, sensitive information leakage interception, and policy violation checks; Opticon is responsible for recording and tracking execution flows; Stratum is responsible for classifying and reviewing outgoing documents. Additionally, there is a Kill Switch function that can cut off AI-related communications by project or organizational unit. Lee Han-wool also mentioned that the company simultaneously develops an attack tool called Stinger for red team testing internally, and the discovered attack patterns are directly fed back into Guardian's defense strategies.

An counter-intuitive argument made during the interview is worth highlighting separately. When the number of agents deployed by an enterprise increases, the volume of operational requests requiring manual approval will surge. Lee Han-wool points out that once approval requests exceed human processing capacity, people will start approving without looking at the background or impact, making humans the new vulnerability themselves. His suggestion is: enterprises must first define which tasks are truly important and require human confirmation, and which can be left to agents for autonomous execution; the focus of approval process design lies not in having someone ultimately confirm the action itself, but in clarifying what tasks, at what timing, and by whom they should be approved, and ensuring that approvers can see sufficient information needed to make judgments.