MAROKO133 Eksklusif ai: The Amount Google’s AI Knows About You Will Cause an Uncomfortable

📌 MAROKO133 Eksklusif ai: The Amount Google’s AI Knows About You Will Cause an Unc

We all know that tech companies keep tabs on everything about our online habits. But it’s another thing to actually be confronted with just how much data they have on you.

This was the experience of tech journalist Pranav Dixit, who experimented with using Google’s new “Personal Intelligence” feature for Gemini and its search engine’s AI Mode. And boy, did things get personal. The AI was able to dig up everything from his license plate to his parents’ vacation history, sometimes without it being directly requested.

“Personal Intelligence feels like Google has been quietly taking notes on my entire life and finally decided to hand me the notebook,” Dixit wrote in a piece for Business Insider.

Google rolled out Personal Intelligence to subscribers of Google AI Pro and AI Ultra last week. Once you opt in, the AI can scour your Gmail and Google Photos accounts, and a more powerful version released for the Gemini app earlier this month goes even deeper, raking your Search and YouTube history, too. In short, if you’ve ever used Google for anything, it can probably dig it up.

This represents one way Google intends to keep its edge in the AI race. Unlike competitors such as OpenAI, it has decades’ worth of user data on billions of people. It can infer plenty from your Google searches alone, and your Gmail account is probably littered with confirmations and reminders for all kinds of life events, ranging from doctor’s appointments to hotel bookings to online purchases.

If the idea of letting an AI prowl through all this sounds like a privacy nightmare to you, you’re probably not wrong. Google, for its part, maintains that it’s being careful with your personal secrets, with VP Josh Woodward insisting in a recent blog post that it only trains its AI on your prompts and the responses they generate — not stuff like your photos and emails.

“We don’t train our systems to learn your license plate number,” he summarized. “We train them to understand that when you ask for one, we can locate it.”

Whatever the ethics, Dixit’s estimation is that giving the AI access to your data at least makes for a genuinely useful — and “scary-good,” in his phrasing — personal assistant.

When asked to come up with some sightseeing ideas for his parents, Personal Intelligence correctly inferred that they’d already done plenty of hikes on previous trips to the Bay Area, and suggested some museums and gardens instead.

Gemini told Dixit that it had deduced this from “breadcrumbs” including emails, photos of a forest they trekked in, a parking reservation in Gmail, and a Google search for “easy hikes for seniors.” It also figured out his license plate number based on photos stored in his Google library and scanned his emails to correctly report when his car insurance was up for renewal.

Privacy isn’t the only concern that the feature raises. With the data, chatbots can sound more humanlike, giving the impression that they’re intimately familiar with users’ personal lives. This is a dangerous road to go down amid reports of many people falling down delusional mental health spirals as they come to believe the AIs are trustworthy companions; Dixit touches on this when he complains about how’d he’d “pour my soul into ChatGPT and get a smart answer,” only for it to “forget I existed like a genius goldfish.” Experts have focused on ChatGPT’s “memory” as allowing it to seem too lifelike by drawing on what you’ve said in previous conversations.

More on AI: AI Is Causing Cultural Stagnation, Researchers Find

The post The Amount Google’s AI Knows About You Will Cause an Uncomfortable Prickling Sensation on Your Scalp appeared first on Futurism.

🔗 Sumber: futurism.com

📌 MAROKO133 Update ai: World’s first exascale supercomputer shows how worn turbine

Scientists have used one of the world’s most powerful supercomputers to find out how microscopic damage to turbine blades undermines jet engine performance, fuel efficiency and durability.

The project brought together researchers from the University of Melbourne, GE Aerospace, and the Oak Ridge National Laboratory (ORNL), who ran simulations on the Frontier supercomputer. The system is the first exascale supercomputer for open science, capable of more than one quintillion calculations per second.

Known as the Hewlett Packard Enterprise Frontier (OLCF-5), the Frontier, which is the world’s most powerful supercomputer for open science analyzed how surface degradation on high-pressure turbine (HPT) blades affects aerothermal efficiency and heat transfer inside jet engines.

“Degradation happens at the microscale, and that makes it very difficult to simulate because of the discrepancy in time and length scales – you’ve got a big blade, but then you’ve got all these minute changes to the surface,” Richard Sandberg, chair of computational mechanics in the University of Melbourne’s Department of Mechanical Engineering, said.

Tiny microscopic flaws

High-pressure turbines in jet engines operate under extreme conditions, with gas temperatures exceeding 3,600 degrees Fahrenheit (2,000 degrees Celsius). Over time, turbine blades are exposed to surface roughness due to erosion, oxidation and mechanical wear.

“This roughness can significantly increase aerodynamic loss, which leads to worse fuel efficiency, and heat flux, which leads to reduced durability and more frequent engine maintenance,” Greg Sluyter, a Turbine Aerodynamics team senior engineer at GE Aerospace, noted.

While this degradation is unavoidable, predicting its impact on engine efficiency has long challenged engineers. To address the issue, the team used the Frontier’s exascale computing power to perform simulations containing between 10 and 20 billion grid points in size with 10¹⁷ degrees of freedom.

They revealed that the previous notions of how roughness affects viscous flow in simple geometries do not apply well to the geometries of turbine engines. “All of our understanding of roughness effects has been built on what we call canonical problems,” Thomas Jelly, PhD, professor at the University of Melbourne and first author of the study, stated.

“But when you look at roughness effects on a blade, it’s actually quite different because there are a lot of fluid dynamic and thermodynamic phenomena that are absent in these canonical cases but present inside jet engines,” he continued.

The novel simulations showed that roughness effects on turbine blades behave very differently. This is largely due to the fact that the blades transition between laminar and turbulent flow.

An engineering challenge

According to the team, surface roughness was found to accelerate this transition, significantly increasing heat transfer to the blade and raising aerodynamic losses. Both effects reduced engine efficiency and shortened component lifespan. It led to higher fuel consumption and more frequent maintenance.

The simulations relied on direct numerical simulation, a method that resolves all relevant turbulence scales without using modeling assumptions. To enable this, the team upgraded its in-house computational code, the High-Performance Solver for Turbulence and Aeroacoustics Research (HiPSTAR).

They then optimized it for Frontier’s AMD GPU architecture. Individual simulation cases took weeks to complete. Running the same calculations on a standard laptop would have required more than a thousand years.

GE Aerospace engineers are already using these insights to next-gen HPT designs. This includes joint work with NASA on the Hybrid Thermally Efficient Core Project to improve fuel efficiency in commercial engines.

The research also supports broader efforts to reduce aviation’s fuel consumption and emissions. More efficient turbines mean less fuel burned for the same thrust, which directly reduces operating costs and environmental impact. The team is also exploring better cooling strategies.

“In the long term, we will develop models that can better predict this so that the designers can have more confidence in their predictions and therefore design a more efficient engine,” Sandberg concluded in a press release.

The study has been published in the ASME Journal of Turbomachinery.

🔗 Sumber: interestingengineering.com