标签：#compute

Once upon a time, Microsoft had to write off $900 million betting an Arm-based Nvidia chip could power its first flagship Windows portable, the original Microsoft Surface. But today, it's trying again. Microsoft and Nvidia have just announced the Surface Laptop Ultra, a computer with a new Arm-based Nvidia chip at its core. There's a […]

This fall, Nvidia will officially become a consumer PC chipmaker like Intel, AMD, Apple, and Qualcomm, putting a complete computing chip - not just graphics - into the very heart of laptops and mini-PCs. After many months of leaks, it's finally announcing the RTX Spark, the first in a family of chips that will meet […]

[Paper Notes] JEPA: Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecture 🔗 TL;DR: JEPA learns a a generalized semantic representation with less data pairs by predicting missing information in the embedding space , which helps it disregard unnecessary noisy from input(pixel)-level details and learns at a higher abstraction level with good semantic generalization. 1. Innovation & Significance The Bottleneck: Image-text data pair labels are hard to find Pixel level pre-training paired & data augmentation are strongly biased towards trained data distribution, hard to determine proper generalization and level of abstraction. JEA's (Joint Embedding Architecture) collapse probelm: encoder & decoder attempts to cheat by always landing on trivial constant when predicting itself (reconstruction) and gets away with an easy Error=0. The Solution: > Chain-of-thought ⭕ Mask pre-training to reduce data & generalize↓❌ Bad/lower semantic representation without semantic target, could be learning noisy local pixel correlation↓⭕ Learn at the embedding level to omit pixel input and generalize⭕ Adds context encoder & positional encoding to inject context and force model to pick up image inherent structure from reconstructing multiple masked patches with one target.↓❌ JEAs wants to cheat: if I always map all pixels to a constant for both the predictor and end target encoder then the reconstruction error is always collapsed to zero! Hehe~ ↓ ⭕ EMA (Exponential moving avg.): Update target encoder parameters from the EMA of context encoders. This 'delays' the target encoder to prevent collapsing (a trick from the BYOL paper[2020], proven essential to training JEAs with ViT). 2. Model & High-Level Intuitions 2.1 Model Architecture Input: randomly samples block masks from original image within certain aspect ratio changes, and apply mask for context image 2.1.2 Context Context Encoder: ViT encodes context image to embedding SxS_x S x ​ Mask Token : an [1,D] random

Dell, Microsoft, and others are unveiling new laptops to compete directly with the Neo, but not all are learning the right lessons from Apple.

Dell is making good on its tease from CES and finally announcing a new XPS 13. The XPS 13 returns as a budget-friendly option, launching in July at a promotional student price of $599 - though that introductory deal only runs until September for back-to-school shopping; it'll start at $699 for everyone else. The $599 […]

Alienware is taking to this year's Computex 2026 in Taipei to announce some cool gaming monitors, most notably two exciting OLED options that are coming at different points this year. First off, the company is debuting the successor to its very first QD-OLED gaming monitor from 2022 with a refreshed design and high-end specs that's […]

NVIDIA's CEO Jensen Huang is set to take the stage for his GTC Taipei keynote at 8PM PT / 11PM ET. You can watch all the announcements here and embedded below. Rumors have been flying about what to expect from today's presentation, but the big one is the possibility of a partnership with Microsoft and […]

In This Article The Question The Intuition Trap Building the State Machine for HH Solving the System: E[HH] = 6 Building the State Machine for HTH Solving the System: E[HTH] = 10 Why Overlapping Patterns Change Everything Python Simulation: 100,000 Trials Business Application: Credit Migration & Web Ranking The Question You flip a fair coin — one with probability 1/2 of landing heads and 1/2 of landing tails — repeatedly, recording every result. What is the expected number of flips required until the sequence HH appears for the first time as consecutive results? What is the expected number of flips required until HTH appears for the first time? Both questions have the same surface structure: you want a specific consecutive pattern, and you want to know, on average, how many flips it takes to observe it. The coin is fair, the flips are independent, and the patterns are short. These seem like they should yield similar answers. They do not. HH takes exactly 6 flips on average. HTH takes exactly 10. The four-flip gap between those two answers is not a rounding artifact or a computational error — it is a precise consequence of the internal structure of each pattern, and deriving it rigorously is one of the cleanest demonstrations of absorbing Markov chain analysis you will encounter. This problem appears frequently in quantitative finance interviews — at firms like Jane Street, Citadel, and Two Sigma — precisely because it separates candidates who understand Markov structure from those who rely on heuristic reasoning. Getting the answer right, and being able to explain it, requires building a state machine, writing the system of first-step equations, and solving it algebraically. That is exactly what we will do. The Intuition Trap Before the formal derivation, it is worth examining why intuition fails here. The most common wrong answer from candidates is that both expected values should be "similar" because the patterns are comparable in length. This intuition imports th

My eyes have seen the PC gaming promised land, and it's a beautifully bright world without a shred of blurriness. It's warm, it looks lovely, and it's impeccably sharp. Also, it's expensive as hell. I've dipped my toe in this world by testing a pre-production version of the upcoming Asus ROG Strix Scar 18, which […]

Find out whether a single Wi-Fi router or a mesh system makes the most sense for your home network.

It's the world's worst kept secret that Nvidia is about to announce its own Arm-powered laptop chips at Computex this weekend, and now Microsoft, Nvidia, and Arm are all openly teasing the announcement. The Windows and Nvidia GeForce accounts on X both posted "A new era of PC" earlier today, and now Arm has followed […]

I pondered the other day what's next for Microsoft's Surface PC lineup, and it looks like we're about to find out. Windows and Surface chief Pavan Davuluri has just teased "something new is coming for developers," complete with a mysterious image of what looks like a curved display edge. Davuluri notes that whatever is coming […]

Made for those who hate having a TV in the home, Samsung’s Movingstyle monitor-on-wheels brings the entertainment when you need it and hides away in a closet when you don’t.

Acer is aiming to take on the MacBook Neo more directly with a new Swift Air 14, a new 14-inch laptop that starts at $699. The laptop will use Intel's lower-cost Core Series 3 "Wildcat Lake" chips - either Core 5 or Core 7 models, both of which are six-core processors. It comes in green, […]

I'm one of those people who enjoy on planes and in public spaces in general. But the issue of privacy was always a hurdle.

To get to $600, Apple and HP made different compromises. If you demand more power and performance, the HP Omnibook 3 delivers.

Elon Musk is publicly reframing xAI’s massive Anthropic compute deal as short-term and cancellable, despite SpaceX’s own S-1 filing describing payments through May 2029.

General Compute is betting SambaNova will be the next breakout chipmaker.

Why your age-gating algorithm is probably doomed to fail in the wild For developers building in the computer vision and biometrics space, there is a massive gap between a model that passes a NIST benchmark and a system that survives the "child-with-a-VPN" test. Recent data indicates that roughly 32% of children are successfully bypassing age-gating tech. As engineers, our first instinct is often to blame the model—to tweak the weights, gather more training data, or tighten the threshold. But the technical reality is more sobering: the failure isn't in the algorithm; it's in the deployment architecture. The Problem with Probabilistic Logic in Binary Workflows Most age estimation models rely on analyzing biometric markers—skin texture, bone structure ratios, and periocular geometry. They produce a probabilistic age range. However, according to NIST's evaluation of age estimation software, to maintain a low false-positive rate, systems often need to set a "challenge age" between 29 and 33 years. If you are a dev tasked with keeping 17-year-olds off a platform, you are essentially forced to build a "buffer zone" of over a decade. If the system flags anyone who might be under 30, the UX becomes a nightmare. If you lower the threshold to 18, the false-negative rate skyrockets. This is the fundamental trade-off of probabilistic facial analysis: precision and recall are at constant war, and in a high-traffic production environment, the "noise" of real-world variables (poor lighting, low-res sensors, off-axis angles) makes consistency nearly impossible. The Breakdown of the Identity Handoff Beyond the model, there are three technical failure points that no amount of Euclidean distance analysis can fix if the pipeline is broken: The Signal-to-Noise Ratio at Source: Evaluation datasets are clean. Production images are taken on scratched lenses in low-light bedrooms. The delta between training distribution and inference-time reality is where the first 10% of accuracy vanishes.

These WIRED-tested computer speakers, from stereo speakers to surround sound, will suit any budget.