Google Officially Introduces Ads in AI Mode Search Results
Google has officially announced that it will display ads in the search results of AI Mode, its AI-powered search feature. Announced at Google Marketing Live, the move signals that advertising in AI search—previously in an experimental phase—is now shifting into full gear.
As AI-driven conversational search becomes widespread, where and how to display ads has been a major point of debate. This announcement means Google has now clarified its monetization model for AI search. For users, an experience where ads are mixed into AI responses is expected to become the norm.
The news has drawn considerable attention, picking up 54 points on Hacker News, with some arguing that the spread of AI search could reshape the very structure of the search advertising market.
OpenAI to Provide Security-Specialized AI "GPT-5.5-Cyber" to Japanese Government and Companies
OpenAI has revealed plans to provide GPT-5.5-Cyber, its latest AI model specialized in cybersecurity, to Japanese government agencies and select companies. According to a report by ITmedia.
GPT-5.5-Cyber is believed to be a model optimized for tasks such as security threat detection and vulnerability analysis. What draws attention is that OpenAI will be directly supporting Japan's national challenge of strengthening its cyber defense capabilities.
AI-driven security measures are becoming increasingly important worldwide, and this collaboration with the Japanese government could serve as a reference case for other countries. That said, introducing AI from a foreign company into highly classified domains is expected to face careful scrutiny from the perspectives of data governance and dependency risk.
LDP Unveils Full Outline of AI Investment Strategy Proposal
The Liberal Democratic Party's Japan Growth Strategy Headquarters (headed by former Prime Minister Fumio Kishida) has revealed the full outline of a proposal aimed at promoting investment in 17 strategic areas, including AI and energy security. As an organization reporting directly to Prime Minister Sanae Takaichi, the proposal is expected to be finalized within the month.
The proposal lays out country- and region-specific strategic approaches and calls for maximizing investment in the AI sector to deliver results. It is also set to be reflected in the government's "honebuto" (bold) economic policy package, making it an important document for forecasting the future of Japan's AI policy.
Strengthening the competitiveness of Japan's domestic AI industry is an urgent task, and attention now turns to the details of the concrete investment frameworks and support measures.
AI Judged "More Human Than Humans" in Turing Test — UCSD Study
A study from the University of California, San Diego (UCSD) has found that in classic Turing tests, there are cases where AI is judged to be "more human" than actual humans.
While the results suggest that AI output has reached a level of convincingly mimicking natural human conversation, they also highlight the ironic situation of humans coming across as "machine-like" by comparison. As AI permeates society, the findings once again raise the question of what it means for the line between humans and AI to blur.
Intuit Cuts 17% of Its Workforce — CEO Says It's "Unrelated to AI"
Intuit, a major accounting and financial software company, has carried out a large-scale layoff affecting approximately 17% of its workforce. The CEO explained that the cuts have "nothing to do with AI," but skepticism is spreading within the Hacker News community.
As AI-driven efficiency gains in the workplace accelerate, cases are mounting where the credibility of corporate claims that "layoffs are not the result of AI adoption" is called into question. Reported by CNBC, the incident has become emblematic of the complexity of employment issues in the AI era.
3.125-bit Quantization Bypasses Tensor Cores — Ultra-Low-Bit LLMs Move Forward
An LLM quantization method at an extremely low precision of 3.125 bits has been reported. The approach is characterized by not relying on tensor cores and running data-free.
Making LLMs lighter is a critical area of research directly tied to running models on edge devices and cutting costs. While 4-bit quantization has become mainstream, the fact that quantization in the 3-bit range is approaching practical levels could further fuel the spread of local AI.
Hardware LLM "Taalas" Achieves Over 14,000 TPS with Llama 3.1 8B
Taalas has announced that it achieved a generation speed of over 14,000 tokens per second running Llama 3.1 8B on dedicated hardware.
LLM inference on specialized chips that don't depend on GPUs offers major advantages in terms of cost and power consumption. With TPS (tokens per second) exceeding 14,000, practical text generation reaches a level where wait times become virtually imperceptible. It demonstrates how hardware-software co-design could significantly advance the practical deployment of LLMs.