Author: Admin

Posted by Nathan Parker, Chrome security team Chrome has been advancing the web’s security for well over 15 years, and we’re committed to meeting new challenges and opportunities with AI. Billions of people trust Chrome to keep them safe by default, and this is a responsibility we take seriously. Following the recent launch of Gemini in Chrome and the preview of agentic capabilities, we want to share our approach and some new innovations to improve the safety of agentic browsing. The primary new threat facing all agentic browsers is indirect prompt injection. It can appear in malicious sites, third-party content…

“The Royal Academy of Engineering’s Green Future Fellowships provide academics, entrepreneurs, innovators and engineers, the space and time to transform their cutting-edge ideas into scalable, commercially viable, technologies to secure a greener, fairer future.” New technologies to store renewable energy, power data centres and computers more efficiently and multiply the power of batteries four-fold are among the engineering solutions awarded a share of £39 million by the Royal Academy of Engineering today (18 December). A total of 13 Green Future Fellows have been awarded £3 million each to scale ambitious ideas and cutting-edge engineering over the next decade into commercially…

Apple’s engineers may have long recognized the widespread desire among iOS developers to recreate the elegant hero animation featured in the App Store app. Understanding the complexity and time investment typically required to implement such animations from scratch, Apple has incorporated this feature into the iOS 18 SDK.With this update, you can now achieve a similar animated transition effect in your own apps using just a few lines of code. This significant enhancement empowers developers to create visually appealing and seamless transitions, elevating the overall user experience of their apps.In this tutorial, we’ll explore how to leverage the new NavigationTransition protocol and…

Automotive Grade Linux (AGL) has announced a reference platform that addresses issues holding back software-defined vehicles (SDVs).As manufacturers attempt to transition toward SDVs, the traditional reliance on specific hardware availability often creates bottlenecks that delay feature deployment and increase integration costs. Named SoDeV, AGL’s open-source platform is designed to enable software-first development that is not bound by hardware constraints.Developed through a collaboration led by Panasonic Automotive Systems, Honda, and the AGL SDV Expert Group, SoDeV seeks to standardise how manufacturers approach hardware abstraction and cloud integration.Getting software-defined vehicles on the roadThe current automotive development model often forces software teams to…

True Ventures co-founder Jon Callaghan doesn’t think we’ll be using smartphones the way we do now in five years — and maybe not at all in 10. For a venture capitalist whose firm has had some big winners over its two decades – from consumer brands like Fitbit, Ring, and Peloton, to enterprise software makers HashiCorp and Duo Security – that’s more than armchair theorizing; it’s a thesis on which True Ventures is actively betting. True hasn’t gotten this far by following the crowd. The Bay Area firm has largely operated under the radar despite managing roughly $6 billion across…

Researchers in Japan have developed an innovative, scalable method to adjust thermal conductivity in thin films through the application of femtosecond lasers. Study: Scalable Thermal Engineering via Femtosecond Laser-Direct-Written Phononic Nanostructures. Image Credit: VVVproduct/Shutterstock.com The study was recently published in Advanced Functional Materials. It illustrates how their method could be pivotal in achieving both laboratory-scale precision and industrial-scale throughput simultaneously. The researchers reported how femtosecond laser-induced periodic surface structures effectively manage thermal conductivity in thin film solids. Their technique uses high-speed laser ablation to generate parallel nanoscale grooves with a throughput that is 1,000 times greater than traditional methods, thereby strategically modifying…

Software engineering techniques continue to evolve and improve, with SEI research playing a key role in helping the Department of Defense (DoD) gain the benefits of these improvements and drive changes to revolutionize software acquisition. Our work helps simplify rules, adapt practices to enable decisive actions, improve workforce capability, and accelerate adoption of commercial and scientific innovations. Since its inception, the SEI has been leading the DoD’s embrace of modern software engineering methods, such as Agile and DevSecOps. The SEI contributed its expertise to the development of the Software Acquisition Pathway (SWP), which was issued in 2020 as Department of…

This fall’s weekslong government shutdown only added to concerns about the state of federal cybersecurity—creating the possibility of blind spots or gaps in monitoring while so many workers were furloughed and contributing in general to the already extensive IT backlog at agencies across the government.“Federal IT workers, they are good jobs, there’s not enough resources for the issues that they have to deal with,” one former national security official, who requested anonymity because they are not authorized to speak to the press, told WIRED. “It’s always underfunded. They always have to catch up.”Amélie Koran, a cybersecurity consultant and former chief…

More and more, there are Android tips — and then there are Pixel tips. Owning a Google Pixel phone has become a ticket of sorts to a uniquely top-tier type of Android experience. With Google’s pure vision for the way the operating system itself should work (and none of the experience-harming and often even privacy-compromising layers other device-makers love to lard into the software) — not to mention all the extra bits of exceptionally helpful Googley goodness that are available only in the Pixel environment — the Google Pixel increasingly represents Android at its best. And as anyone who’s spent…

import dataclassesimport os import datasetsimport tokenizersimport torchimport torch.distributed as distimport torch.nn as nnimport torch.nn.functional as Fimport torch.optim.lr_scheduler as lr_schedulerimport tqdmfrom torch import Tensorfrom torch.distributed.checkpoint import load, savefrom torch.distributed.checkpoint.state_dict import StateDictOptions, get_state_dict, set_state_dictfrom torch.distributed.pipelining import PipelineStage, ScheduleGPipe  # Build the model@dataclasses.dataclassclass LlamaConfig:    “””Define Llama model hyperparameters.”””    vocab_size: int = 50000  # Size of the tokenizer vocabulary    max_position_embeddings: int = 2048  # Maximum sequence length    hidden_size: int = 768  # Dimension of hidden layers    intermediate_size: int = 4*768  # Dimension of MLP’s hidden layer    num_hidden_layers: int = 12  # Number of transformer layers    num_attention_heads: int = 12  # Number of attention heads    num_key_value_heads: int = 3  # Number of key-value heads for GQA  class RotaryPositionEncoding(nn.Module):    “””Rotary position encoding.”””     def __init__(self, dim: int,…