Rolling coverage from San Jose, including NVIDIA CEO Jensen Huang’s keynote, news highlights, live demos and on‑the‑ground color through March 19.

"더 크면 더 좋다"는 공식이 깨졌다.

2026년 AI의 트렌드는 더 큰 모델이 아니라 더 현명한 배포다. SLM 아키텍처가 개선되고 양자화 기술이 발전하면서 특정 작업에서 대형 모델과 소형 모델 간 성능 격차가 좁아지고 있다. 대형 이커머스 업체가 GPT-3.5 API를 fine-tuned Mistral 7B로 교체했더니 90% 비용 절감, 3배 빠른 응답, 그리고 일반 질문에서 동등하거나 더 나은 정확도를 기록한 사례가 있다. 복잡한 쿼리만 GPT-4로 에스컬레이션하고, 75%의 티켓은 SLM이 처리한다.

Every time you send a support ticket, log file, or user message to an LLM API, you're potentially leaking emails, phone numbers, credit card numbers, API keys, and more to a third party. Most developers know this is a problem. Most do nothing about it because building a sanitization layer from sc

If you're on pay-per-token plans, you've probably been surprised by a bill at least once. Here's how the main options compare:

1. Provider Dashboard (OpenAI, Anthropic, etc.)

• Updates every few hours
• No per-session breakdown
• Free
• → Lire l'article

It’s a feature of the architecture

The post Hallucinations in LLMs Are Not a Bug in the Data appeared first on Towards Data Science.

Optimizing the cost and latency of your LLM calls with Prompt Caching

The post Why Care About Prompt Caching in LLMs? appeared first on Towards Data Science.

A deep dive into exactly how text-only language models are finetuned to *see* images

The post How Vision Language Models Are Trained from “Scratch” appeared first on → Lire l'article

With no enforcement and questionable economics, it may not make a difference.

OpenAI's new GPT‑5.3‑Codex‑Spark is 15 times faster at coding than its predecessor.

Nvidia added two leading Chinese automakers, BYD and Geely, to its robotaxi program, as the chipmaker seeks to put its stamp on the growing autonomous vehicle market worldwide. At its GTC conference today, Nvidia announced that BYD and Geely, as well as Isuzu and Nissan, would use the chipmaker's Dr

Nvidia announced DLSS 5 on Monday during its GTC conference, and based on early reactions, it's going to be a divisive update, with some reactions calling it "slop" that unacceptably alters artistic intent. Nvidia CEO Jensen Huang is calling this the "GPT moment for graphics - blending hand-crafted

Railway, a San Francisco-based cloud platform that has quietly amassed two million developers without spending a dollar on marketing, announced Thursday that it raised $100 million in a Series B funding round, as surging demand for artificial intelligence applic

Dell expands Alienware 16 Area-51, Alienware 16X Aurora, and Alienware 18 Area-51 laptops with new Intel Arrow Lake-HX Refresh processor options.

The HBM4 36GB 12H stack runs at over 11 Gb/s pin speeds, delivering bandwidth greater than 2.8 TB/s.

Six commercial space companies are understood to have already deployed the platform.

"Open models are the lifeblood of innovation and the engine of global participation in the AI revolution," said Nvidia CEO Jensen Huang.

Roche's new deployment spans more than 3,500 NVIDIA Blackwell GPUs across its worldwide operations and embedded across the entire value chain, massively scaling R&D productivity, next-generation diagnostics and manufacturing efficiencies.

Setting up AI factories in simulation — decreasing deployment time from months to days — is  accelerating the next industrial revolution.  Nowhere was that more apparent than at GTC 2026, in San Jose, where NVIDIA founder and CEO Jensen Huang introduced NVIDIA DSX Air. Part of NVIDIA DSX Sim in the

UNESCO reports that artificial intelligence is reshaping education worldwide, enabling personalized learning experiences for millions of students.

The promise of AI in the classroom goes far beyond simple automation. Adaptive learning platforms analyze each learner’s progress, strengths,

Nvidia CEO Jensen Huang said he expects $1 trillion worth of orders for the chips.

Sen. Elizabeth Warren noted that Grok, xAI's controversial chatbot, has created harmful outputs for users and poses a potential national security risk.

Nvidia’s new DLSS 5 uses generative AI and structured graphics data to make video games more realistic. CEO Jensen Huang says the approach could eventually spread to other industries.

GTC is Nvidia's flagship annual event, where the chipmaker typically announces new products, partnerships, and its vision for the future of computing. Huang's keynote will focus on Nvidia's role in the future of computing and AI.

At Nvidia CEO Jensen Huang's urging, Frore developed liquid-cooling tech for chips. That shift helped it raise $143 million.

Cursor is quietly becoming more than an IDE. With background agents that run autonomously in the background, it's turning into a platform where code can buy things on its own — data feeds, skill protocols, intelligence endpoints — without a human in the loop.

The payment layer th

In the span of four days in mid-March 2026, the AI ecosystem shipped three fundamentally incompatible visions of how the next generation of intelligent systems should work. Nobody's talking about the contradiction. They should be.

On March 11, Baidu launched → Lire l'article

Add Visual Audit Trails to Your Windsurf Agent in 5 Minutes

Your Windsurf agent just completed a complex task. It edited code, ran tests, committed to GitHub, opened a PR. You see the task output.

Your compliance officer asks: "Prove the agent made the right changes. S

NTT DATA has announced an initiative to deliver NVIDIA-powered platforms designed to give organisations a repeatable, production-ready model for scaling AI. The offering integrates NVIDIA’s GPU-accelerated computing and high-performance networking with NVIDIA AI Enterprise software, includi

E.SUN Bank is working with IBM to build clearer AI governance rules for how artificial intelligence can be used inside a bank. The effort reflects a wider shift in finance. Many firms already use AI for fraud checks and credit scoring, and some also use it to handle customer service queries. The

That guest network you set up for your neighbors may not be as secure as you think.

Apple has just announced the AirPods Max 2 with better noise cancellation and sound quality. The second generation of Apple's over-ear headphones starts at $549 and comes equipped with an H2 chip that powers a bunch of new features, including AI-powered live translation, which translates speech to y

OpenAI's delayed "adult mode" for ChatGPT is expected to support saucy text conversations at launch, but not the chatbot's ability to generate images, voice, or video. Speaking to The Wall Street Journal, an unnamed OpenAI spokesperson described content that will be provided by the upcoming feature

Facebook Marketplace is adding a bunch of new AI-powered tools that are supposed to make selling items on the platform a little more efficient. One feature will use Meta AI to automatically respond to those annoying "Is this still available?" messages. You can toggle on the auto-reply option when cr

Alfred Wahlforss was running out of options. His startup, Listen Labs, needed to hire over 100 engineers, but competing against Mark Zuckerberg's → Lire l'article

The startup also announced a $5 million "rescue fund" to help credit unions ditch legacy software for its AI-native platform.

PhonePe has paused its planned IPO, delaying a major liquidity event for investors including Tiger Global and Microsoft.

Google and Accel say about 70% of AI startup pitches tied to India were "wrappers" as they reviewed more than 4,000 applications for their Atoms cohort.

The deal comes after Unacademy’s valuation plunged from $3.5B to under $500M, as India’s once-booming edtech sector struggles.

Shardul Shah of Index Ventures walks us through Google's biggest acquisition ever.

Meta AI just crossed a billion monthly users.
It's free, embedded in Instagram/WhatsApp/Facebook/Messenger, and generates images through Imagine.

For developers and builders, though, the limitations are structural:
No persistent memory — context doesn't

Europe’s factory floors have a new kind of colleague. BMW Group has deployed humanoid robots in manufacturing in Germany for the first time, launching a pilot project at its Leipzig plant with AEON–a wheeled humanoid built by Hexagon Robotics.  It is the first automotive deploymen

Industrial AI, digital twins, AI physics and accelerated AI infrastructure are empowering companies across industries to accelerate and scale the design, simulation and optimization of products, processes and facilities before building in the real world.

ADLINK Technology has signed a strategic alliance and joint development agreement with Under Control Robotics, the company behind the robotics startup Noble Machines. The two firms will combine ADLINK’s edge AI platforms with Noble Machines’ autonomy software to create a new generation of general