Latest AI News

Get the fastest, most accurate AI news every morning

Total articles: 13,454|Added todayToday: 116

Category

Search

South Korea and AMD Review AI Chip Partnership

South Korea’s science ministry and AMD reviewed AI chip cooperation at a committee meeting on the 4th.
Plans include linking Korean NPUs with AMD CPUs and GPUs, an AI research center, and university support.
AMD held its Advancing AI event in South Korea for the first time; registration closed early after about 1,300 sign-ups.

ETNewsRead more · 1 min →

Qualcomm Validates HBC for AI250 Accelerator

Qualcomm completes initial hardware validation of first-generation HBC for its Dragonfly AI250 accelerator
Company reports more than 5× HBM4 bandwidth-per-watt efficiency and 18 times external LPDDR5X bandwidth
AI250 sampling is scheduled for the first half of 2027; AI300 sampling is expected in 2028

TrendForceRead more · 1 min →

SemiAnalysis Unveils GDPx Work Benchmark

SemiAnalysis released GDPx, assessing 206 tasks across 11 professional fields.
Claude Opus 5.5 led at 12.0%, followed by Gemini 4 Carbon at 10.9%.
No model fully completed 25 finance runs; all four models failed the legal task.

AI Times KoreaRead more · 2 min →

LG AI Research Wins Award for EXAONE Nexus

LG AI Research won the National Intellectual Property Commission Chairperson’s Award for EXAONE Nexus, which assesses copyright, licensing and privacy risks.
The system analyzed one dataset 45 times faster than legal experts, with 26% higher accuracy and costs at about one-thousandth the level.
LG plans to apply the tool across its internal AI development cycle and pursue global technology and law firm clients; timing and deal size remain undisclosed.

ETNewsRead more · 2 min →

Tohoku University and Aisin Develop STT-MRAM AI Chip

Tohoku University and Aisin developed a 16-nanometer AI chip with built-in STT-MRAM
The chip starts its OS 38.59 times faster, cuts AI processing time to about one-tenth, and uses about one-fiftieth the inference energy
A hybrid design cut Tiny YOLOv3 weight data by 53%, while mAP fell from 0.325 to 0.316

ITmedia AI+Read more · 2 min →

Cainz Cuts Demand Forecast Error by 19%

Cainz used AlphaEvolve to improve long-term demand forecasts for overseas-made private-label products.
Forecast error improved 19% over the previous model after hundreds to thousands of algorithm searches.
Cainz plans to measure stockout reduction and inventory optimization in categories with improved forecast accuracy.

IT LeadersRead more · 2 min →

Today'sSafety & Ethics

Anno Calls for Domestic Cybersecurity AI

Tak Hirohiro Anno warned that running foreign open-weight models domestically can still pose risks.
Domestic servers and data sovereignty do not eliminate the possibility of degraded performance or data leaks.
Anno called for three measures: access to foreign closed models, secured foreign open-weight models, and domestic development.

ITmedia AI+Read more · 1 min →

Trending Keywords

Apple Opens Siri AI to Developers

Apple opened Siri AI app integrations to all developers in betas for iOS 27.3, iPadOS 27.3, macOS 27.3 and visionOS 27.3.
App Intents enable custom app actions, while skills let Siri combine several actions into one experience for complex requests.
App Store submissions open in the spring, ahead of customer release in a few months; Siri AI supports six languages.

AppleRead more · 1 min →

LangSmith BYOC Expands to Three Clouds

LangChain made LangSmith BYOC available on AWS, Azure, and Google Cloud, with AWS generally available and the other two in beta.
BYOC keeps traces, prompts, datasets, experiments, and deployments in customer accounts while LangChain operates the platform data plane.
Migration tooling transfers workspace resources; trace migration is available on AWS, and the LLM Gateway supports 11 providers.

LangChainRead more · 2 min →

Noahlogy and AVEVA Partner on Ship Design

Noahlogy signed a partner agreement with UK-based AVEVA on November 2, 2026.
The companies are testing data reading, visualization and design checks using AVEVA 3D models.
They aim to develop practical support shaped by shipyards’ and design teams’ needs.

ITmedia AI+Read more · 1 min →

Team Air Reports 2,000 Hours Saved With AI Agents

Team Air agents cut inventory and customer-credit searches from 15 to 20 minutes to about 60 seconds
Invoice-processing automation reached 75%, above the 60% to 70% target, saving more than 2,000 hours annually
Team Air is exploring purchasing and procure-to-pay automation after accounts payable and receivable

Constellation ResearchRead more · 1 min →

IdeaTaste Finds LLMs Trail Researchers on Ideas

IdeaTaste compares research proposals from six language models with 2026 papers across four fields
Across more than 90,000 proposals, agents reached 26% of human citation matching and 18% of idea matching
Expert reviews found the largest gap in expected success, while richer context and larger populations changed little

HuggingFaceRead more · 1 min →

Salesforce Reports Agentforce Memory Gains

Salesforce AI Research reports Agentforce Memory policy recalls 1.8 times as many stated facts and 2.2 times as many changing facts
Live-system evaluation tests facts across long, multi-session conversations using Salesforce’s ConvoMem benchmark, released in 2025
Updated policy treats customer updates as corrections; Salesforce says the underlying Agentforce Memory architecture stayed the same

SalesforceRead more · 1 min →

Today'sMarketing

Evernote Tests ChatGPT Ads for Subscriptions

Evernote used ChatGPT Ads to guide prospective users from ad clicks to account registrations, free trials and subscriptions.
About 25% of desktop visits led to new accounts; trial-to-paid conversion was 10–20% above a historical average.
Evernote called first-month results directional and plans additional use cases, creative and dedicated landing pages.

OpenAIRead more · 1 min →

Yesterday'sFinance

AMD Revenue Jumps 45%, Shares Rise 7%

AMD's third-quarter revenue rose 45%, while its shares gained 7% in after-hours trading.
Data center revenue nearly doubled; AMD forecast fourth-quarter revenue of $17 billion.
Investors weighed AI growth against speculation risks and Anthropic's potential IPO.

EdailyRead more · 2 min →

Chinese Models Trail in Revised AI Rankings

Artificial Analysis’ revised ranking puts open-weight models 12 points behind the leader, up from 3 points under its previous system.
Publicly released-weight models scored 32 points, while Claude Opus 5.5 scored 58; Claude Haiku 5.5 beat all of them.
Dax Raad estimated one rack of Nvidia computers costing $6 million to $7 million could handle almost all prior-day DeepSeek traffic on OpenRouter.

Business InsiderRead more · 2 min →

Yesterday'sFinance

Investors Push Anthropic IPO Valuation Lower

Institutional buyers are pushing Anthropic’s IPO valuation toward $1.5 trillion after last week’s tech selloff.
SEC filing reports $33.2 billion revenue in 2026’s first nine months, up 1,206% year over year.
Investors cite market volatility, AI exposure concerns and $895.4 billion in commitments and leases as valuation factors.

sainteur.comRead more · 2 min →

Yesterday'sFinance

AI Stock Declines Temper Wall Street Gains

U.S. stocks ended mixed as AI and tech share declines offset gains across much of Wall Street
Nasdaq-100 entered its third correction this year as investors questioned Big Tech’s hundreds of billions in AI spending
Brookfield’s planned Drax data center could top £20 billion, pending approvals, a tenant and final investment decisions

sainteur.comRead more · 1 min →

Yesterday'sInfra

Pennsylvania Township Approves Data Center District

Salem Township approved roughly 4,500 acres for a data center overlay district, an early step toward planned and proposed projects.
About 13 percent of roughly 2,180 households may sell property; developers promised free well testing and about $12 million for a micro-hospital.
Residents split over data centers as officials faced questions about water, property values, conflicts of interest and public-comment rules.

GovTech AIRead more · 2 min →

Six AI Models Launch Across Voice, Robotics and Design

Rime launched Prism, a speech-to-speech model with a response start time of about 0.57 seconds.
Hume AI reported 70.7% agreement with human evaluators and an 82% average across voice benchmarks.
Canva unveiled Brush, while Nace AI and Unsloth released decision-model weights and browser-control features.

AI Times KoreaRead more · 3 min →

Trending Keywords

Last 7 Days