Agregátor RSS

Meanwhile, US apparently considers further AI hardware sanctions

Germany's government has named China-controlled actors as the perpetrators of a 2021 cyber attack on the Federal Office of Cartography and Geodesy (BKG) – the official mapping agency.…

Od začátku června tohoto roku zemřelo při nejhorší epidemii viru Chandipura v Indii za posledních 20 let nejméně 38 lidí, většinou dětí a dospívajících. Tyčinkovitý patogen je členem rodiny virů vztekliny, způsobuje encefalitidu – zánět a otok mozku. Šíří ho především písečné mušky, ale mohou ho ...

AMD zveřejnila finanční výsledky za druhý kvartál, které přinesly několik bezprecedentních skoků, ale i pár dílčích poklesů.

Do 22. srpna probíhá na Kickstarteru kampaň na podporu telefonu Volla Phone Quintus. S Volla OS nebo Ubuntu Touch. Za 647 eur.

Scumbags go for the jugular

A ransomware attack against blood-donation nonprofit OneBlood, which services more than 250 American hospitals, has "significantly reduced" the org's ability to take, test, and distribute blood.…

Závěti pořízené během hospitalizace v nemocnici, pobytu v pečovatelském domě, hospici, LDN ve prospěch jejich provozovatelů nebo zaměstnanců jsou zásadně neplatné. Ale kdy jde a kdy nejde o zařízení?

Datová analýza je velmi zajímavý a neustále se rozvíjející se obor. Své místo v něm má i knihovna scikit-learn běžící v ekosystému jazyka Python. Dnes si představíme naprosté základy této knihovny.

Intel oznámil datum akce, na které odhalí energetickou efektivitu nové x86 architektury i výkon integrované grafiky a nového AI akcelerátoru…

Superdiamant BC8 by měl být nejtvrdším materiálem ze všech, co známe. Vědci předpokládají, že se vyskytuje uvnitř planet minimálně dvakrát větších než Země. Superpočítač Frontier se zapotil, ale dal dohromady recept, s nímž snad vědci vybavení výkonnými lasery National Ignition Facility dovedou superdiamant vyrobit.

Small stay of execution in 'exceptional circumstances' promised – amid legal action to pause digital bonfire

As the DigiCert drama continues, we now have a better idea of the size and scope of the problem – with the organization's infosec boss admitting the SSL/TLS certificate revocation sweep will affect tens of thousands of its customers, some of which have warned that the short notice may have real-world safety implications and disrupt critical services.…

Eight-year-old domain hijacking technique still claiming victims

Dozens of Russia-affiliated criminals are right now trying to wrest control of web domains by exploiting weak DNS services.…

Traveling the vast distances between solar systems is well beyond existing technology. But a new ultra-thin lightsail designed with AI could make it possible to reach the nearest star within 20 years.

Launched in 1977, the Voyager 1 probe was the first human-made object to leave our solar system. But at current speeds, it would take over 70,000 years to reach Alpha Centauri, the closest star system to our own.

There is one propulsion technology, however, that could significantly speed things up. A lightsail is a large reflective surface deployed in front of a spacecraft, where it can harness either sunlight or light from an Earth-based laser to continually accelerate the vehicle. In theory, this could make it possible to achieve speeds of 10 to 20 percent of the speed of light.

Building materials that are both reflective and light enough to make this possible has been an outstanding challenge though. Now, researchers have used an AI technique called “neural topology optimization” to create a nanometer-thick sheet of silicon nitride that could bring the idea to life.

“This mission requires lightsail materials that challenge the fundamentals of nanotechnology, requiring innovations in optics, material science, and structural engineering,” the team writes in a preprint posted to arXiv.

“This study underscores the potential of neural topology optimization to achieve innovative and economically viable lightsail designs, crucial for next-generation space exploration.”

The researchers’ technique was inspired by Breakthrough Starshot, a project launched by the Breakthrough Initiatives in 2016. Starshot seeks to design a fleet of around 1,000 tiny spacecraft that use lightsails and an Earth-based laser to reach Alpha Centauri within 20 to 30 years. The probes would carry cameras and other sensors to send back data on arrival.

To reach the required speeds, the spacecraft will have to be incredibly light—the probes themselves will be just centimeters across and weigh a few grams. But to gather enough light, the sails need to measure roughly 100 square feet, so we need new ultralight materials to keep their weight down.

One promising approach involves creating optical nanostructures called “photonic crystals” made up of a repeating grid of tiny holes. Punching millions or billions of these holes into the material reduces its weight significantly, but these repeating structures also create unusual optical effects that can actually enhance the material’s reflectivity.

Working out exactly how to arrange these holes is a complicated process though, so the group from Delft University in the Netherlands and Brown University in the United States enlisted AI to help them. They combined a neural network with a more conventional computational physics program to find the most optimal configuration and shape of the holes to minimize mass and boost reflectivity.

This resulted in a lattice of bean-shaped holes less than 200 nanometers thick. To show the design works as expected, they used an approach called flood lithography, in which a laser uses an incredibly detailed stencil to create holes in a silicon nitride wafer. Using the approach, the team created a 5.5 square inch sample that weighed just a few micrograms.

Lithography is the same technology companies use to make computer chips, so the researchers think the approach could easily be scaled up. The team predict it would take about a day and cost around $2,700 to create a full-sized sail. They’d need to build a dedicated facility though, team leader Richard Norte, from Delft, told New Scientist, because those used for chip fabrication only work with wafers about 15-inches long.

There are still a lot of other engineering challenges to be solved for the Breakthrough Starshot mission to come together, Stefania Soldini at the University of Liverpool told New Scientist, but a cheap and fast way of producing lightsails will be crucial.

NASA is also actively pursuing the approach. Just last week, the agency announced that its Advanced Composite Solar Sail System, which launched earlier this year, is close to hoisting its sails for the first time.

If these projects are successful, we may get our first close-up glimpse of worlds beyond our solar system within many people’s lifetimes.

Image Credit: This 4.5-square-inch sample could lead to a full-sized lightsail lightweight enough to tow tiny spacecraft to another star system / L. Norder, et al via arXiv

Jane Greaves z Cardiff University a její spolupracovníci zveřejnili nová zjištění podporující jejich hypotézu, že na Venuši by se teoreticky mohla vyskytovat nějaká forma mimozemského života. Připomeňme, že tato hypotéza se až dosud opírala především o objev fosfanu, který byl oznámen v roce 2020. ...

I do pětistovky koupíte dobré periferie k počítači • S Logitechem soupeří i privátní značky od Alzy a CZC • Víte, že existují opravdu tiché myši?

Windows users now get macOS-grade secret security

Google says it's enhancing the security of sensitive data managed by Chrome for Windows users to fight the scourge of infostealer malware targeting cookies.…

Google parent Alphabet’s $2 billion investment in AI firm Anthropic has caught the eye of the UK’s antitrust regulator.

On Tuesday, the UK’s Competition and Markets Authority (CMA) opened an inquiry into whether Alphabet’s partnership with Anthropic created a “relevant merger situation” that threatened competition within the fast-growing market for cloud-delivered AI products and services.

Invitation to comment

The CMA is inviting industry comments before a deadline of Tuesday, August 13, in advance of the launch of its formal investigation. The outcome of the inquiry will decide whether or not the regulator orders remedial actions or otherwise intervenes in the market.

The prospective probe is part of more general concerns about competition in the generative AI industry.

Partnerships between other hyperscalers and AI startups have also attracted concerns, including the business relationships between Microsoft and OpenAI as well as links between Amazon Web Services and Anthropic.

Open relationship

Google told Computerworld that Anthropic is free to partner with other cloud technology providers and hyperscalers, effectively arguing that competitive concerns were misplaced.

“Google is committed to building the most open and innovative AI ecosystem in the world,” the tech giant said in a statement. “Anthropic is free to use multiple cloud providers and does, and we don’t demand exclusive tech rights.”

In a statement, Anthropic told Computerworld it intends to “cooperate with the CMA and provide them with the complete picture about Google’s investment and our commercial collaboration.

“We are an independent company and none of our strategic partnerships or investor relationships diminish the independence of our corporate governance or our freedom to partner with others,” the company said. “Anthropic’s independence is a core attribute, integral both to our public benefit mission and to serving our customers wherever and however they prefer to access Claude.”

Smaller players in the cloud computing market argue that powerful partnerships threaten the development of competition in the AI marketplace.

‘Virtual monopoly’

Josh Mesout, chief innovation officer of UK-based cloud computing firm Civo, told Computerworld, “As an industry we should be cautious over powerful partnerships as they pose a threat to the entire ecosystem by suffocating competition and innovation.”

He added, “We cannot surrender AI to a virtual monopoly before it has really started.”

Maintaining a diverse and competitive landscape in artificial intelligence is important not least because of the diverse and far-reaching applications of AI technologies across multiple industry sectors.

“Over-dependence on a handful of major firms could stifle innovation, limit consumer choice, and potentially lead to a monopoly that favors Big Tech,” Mesout warned.

“To keep the market fair and open, regulators should be eyeing these types of partnerships warily,” he said. “Otherwise, we risk AI following the path of cloud, where hyperscalers run unchecked and leave a broken, locked-in, and stifled market in their wake.”

To get a sense of what a smarter Siri in iOS 18.1 might look like once it appears, Open AI just introduced a new voice mode in its app, albeit in limited alpha, meaning not every user will get ahold of the new tech.

Delayed for a month in response to quality concerns, this is a test of the company’s Advanced Voice Mode on ChatGPT; it’s available to iPhone users who subscribe to the $20 per month ChatGPT Plus service using its GPT-4o model. 

The company warns that it might make mistakes and says access and rate limits are subject to change. It isn’t expected to be universally available across all users until the end of the year, and should be available to Mac, iPhone, and iPad users once it appears. Subscribers accepted to the alpha group will get an alert in the app and an email inviting them to take part in the test. 

“We’ll continue to add more people on a rolling basis and plan for everyone on Plus to have access in the fall,” OpenAI said.

What does Advanced Voice Mode do? 

Effectively, it’s a more powerful chatbot that delivers more natural, real-time conversations with a degree of contextual awareness, which means it can understand and respond to emotion and non-verbal cues. It is also capable of processing prompts more swiftly, which significantly reduces the latency within conversations, and lets you interrupt it to get it to change what it says at any time.

OpenAI first demonstrated the new mode in April, when it showed how the tool can recognize different languages simultaneously and translate them in real time. During that demo, employees were able to interrupt ChatGPT, get it to tell stories in different ways, and more. One thing the bot can no longer do is sound like Scarlet Johansson — it now supports only four preset voices in order to prevent it being used for impersonation. OpenAI has also put filters in place to block requests to generate music or other copyrighted audio, reflecting legal challenges raised against song-generating AI firms such as Suno.

Video and screen sharing capabilities are not yet available.

How it works

If you are a ChatGPT Plus subscriber running the latest version of the app, and are accepted to the test, you can access the bot from within the app by tapping the Voice icon at the bottom of the screen. You can then switch between the new Advanced mode and the existing Standard mode (better for longer sessions) using an interface at the top of the screen. Privacy concerns mean many Apple users might prefer to access these features via Apple Intelligence.

What about privacy?

Apple Intelligence puts additional safeguards in place to protect people’s privacy. As Wired points out, ChatGPT’s user agreement at present appears to want to use your voice and images for training purposes. In a remarkably quotable line, AI consultant Angus Allan calls it a “data hoover on steroids. Their privacy policy explicitly states they collect all user input and reserve the right to train their models on this,” he said.

This is less a problem when used with Apple Intelligence, as ChatGPT requests are anonymized and data from those sessions is not used to train ChatGPT models, according to Apple. If that proves true, many Apple users will eventually gravitate to accessing ChatGPT via their Apple AI as the safest way to use it.

All eyes now will turn to Google, which is expected to introduce similar features within Google Gemini AI soon — features that might also end up being integrated inside Apple Intelligence. The battle of the bots is heating up.

Please follow me on Mastodon, or join me in the AppleHolic’s bar & grill and Apple Discussions groups on MeWe.

Open AI on Tuesday began rolling out a preview version of Advanced Voice Mode to paying ChatGPT customers. With the help of the new function, users get access to a “hyper-realistic voice” that sounds exactly like a human.

After a demonstration in May, Open AI was accused of copying the voice of actress Scarlett Johansson without permission, prompting the company to remove the voice in question. Ever since, users have to make do with four voices, which are called Juniper, Breeze, Cove and Ember.

The idea now is that all Plus users will get access to Advanced Voice Mode in the fall, Techcrunch reports.

Watch this Palo Alto Networks keynote to understand the importance of visibility, control and governance in AI application and service development

Sponsored Post  The dawn of artificial intelligence is upon us, but its development has only just begun.…