Agregátor RSS

Ve věku 96 let zemřel Thomas Eugene Kurtz, americký informatik, který spolu s Johnem Georgem Kemenym vytvořil programovací jazyk BASIC.

Čínská armáda pracuje na pozoruhodném zbraňovém systému, který dokáže spojit několik svazků mikrovln do jediného silného energetického paprsku. Informoval o tom deník South China Morning Post (SCMP). Vědci tvrdí, že již „dokončili experimentální testy potenciálního vojenského použití“ daného ...

Palo Alto Networks is warning that a critical zero-day vulnerability on Next-Generation Firewalls (NGFW) management interfaces, currently tracked as 'PAN-SA-2024-0015,' is actively being exploited in attacks. [...]

A nervous wait for rapper wife who also faces a stint in the clink

The US is sending the main figure behind the 2016 intrusion at crypto exchange Bitfinex to prison for five years after he stole close to 120,000 Bitcoin.…

Dlouho se mluvilo o tom, že chce Amazon reagovat na sílící čínskou konkurenci. Včera firma odstartovala nový online obchod Haul (dopravit, ulovit), který uvařila podle stejného receptu jako Aliexpress, Temu, Shein atd. Chce, aby lidé na mobilu nakupovali různé cetky za pár korun. Nebo zatím jen ...

Cybersecurity researchers have disclosed two security flaws in Google's Vertex machine learning (ML) platform that, if successfully exploited, could allow malicious actors to escalate privileges and exfiltrate models from the cloud. "By exploiting custom job permissions, we were able to escalate our privileges and gain unauthorized access to all data services in the project," Palo Alto Networks

Cybersecurity researchers have disclosed two security flaws in Google's Vertex machine learning (ML) platform that, if successfully exploited, could allow malicious actors to escalate privileges and exfiltrate models from the cloud. "By exploiting custom job permissions, we were able to escalate our privileges and gain unauthorized access to all data services in the project," Palo Alto Networks Ravie Lakshmananhttp://www.blogger.com/profile/[email protected]

Research by British telecommunications provider O2 has found that seven in ten Britons (71 percent) would like to take revenge on scammers who have tried to trick them or their loved ones. At the same time, however, one in two people does not want to waste their time on it.

AI grandma against telephone scammers

O2 now wants to remedy this with an artificial intelligence called Daisy. As the “head of fraud prevention”, it’s the job of this state-of-the-art AI granny to keep scammers away from real people for as long as possible with human-like chatter. To activate Daisy, O2 customers simply have to forward a suspicious call to the number 7726.

Daisy combines different AI models that work together to first listen to the caller and convert their voice to text. It then generates responses appropriate to the character’s “personality” via a custom single-layer large language model. These are then fed back via a custom text-to-speech model to generate a natural language response. This happens in real-time, allowing the tool to have a human-like conversation with a caller.

Although human-like is a strong understatement: Daisy was trained with the help of Jim Browning, one of the most famous “scambaiters” on YouTube. With the persona of a lonely and seemingly somewhat bewildered older lady, she tricks the fraudsters into believing that they have found a perfect target, while in reality she beats them with their own weapons.

In the fast-paced digital world, trust is everything—but what happens when that trust is disrupted? Certificate revocations, though rare, can send shockwaves through your operations, impacting security, customer confidence, and business continuity. Are you prepared to act swiftly when the unexpected happens? Join DigiCert’s exclusive webinar, "When Shift Happens: Are You Ready for Rapid

In the fast-paced digital world, trust is everything—but what happens when that trust is disrupted? Certificate revocations, though rare, can send shockwaves through your operations, impacting security, customer confidence, and business continuity. Are you prepared to act swiftly when the unexpected happens? Join DigiCert’s exclusive webinar, "When Shift Happens: Are You Ready for Rapid The Hacker Newshttp://www.blogger.com/profile/[email protected]

Redaktoři Android Authority si při analýze nové verze Google Play Services všimli v kódu odkazů na zatím neoznámenou funkci Shielded Email. Tyto „stíněné e-maily“ nejsou ničím jiným než aliasy, tedy alternativními adresami, které přesměrují zprávy do hlavní schránky. Službu těsně integrovanou do ...

A Vietnamese-speaking threat actor has been linked to an information-stealing campaign targeting government and education entities in Europe and Asia with a new Python-based malware called PXA Stealer. The malware "targets victims' sensitive information, including credentials for various online accounts, VPN and FTP clients, financial information, browser cookies, and data from gaming software,"

A Vietnamese-speaking threat actor has been linked to an information-stealing campaign targeting government and education entities in Europe and Asia with a new Python-based malware called PXA Stealer. The malware "targets victims' sensitive information, including credentials for various online accounts, VPN and FTP clients, financial information, browser cookies, and data from gaming software,"Ravie Lakshmananhttp://www.blogger.com/profile/[email protected]

Máme zde výprodej na Black Friday a ceny od Dckeyoffer jsou opravdu skvělé když sháníte originální doživotní licence na Windows. Během výprodeje na Black Friday máme vysoké slevy na platné licence na Windows a Office. Zejména klíč na Windows 11 je dnes nejžádanější, tak na co čekat?

OpenAI recently introduced SimpleQA, a new benchmark for evaluating the factual accuracy of large language models (LLMs) that underpin generative AI (genAI).

Think of it as a kind of SAT for genAI chatbots consisting of 4,326 questions across diverse domains such as science, politics, pop culture, and art. Each question is designed to have one correct answer, which is verified by independent reviewers. 

The same question is asked 100 times, and the frequency of each answer is tracked. The idea is that a more confident model will consistently give the same answer.

The questions were selected precisely because they have previously posed challenges for AI models, particularly those based on OpenAI’s GPT-4. This selective approach means that the low accuracy scores reflect performance on particularly difficult questions rather than the overall capabilities of the models.

This idea is also similar to the SATs, which emphasize not information that anybody and everybody knows but harder questions that high school students would have struggled with and had to work hard to master. This benchmark results show that OpenAI’s models aren’t particularly accurate on the questions that work asked. In short, they hallucinate. 

OpenAI’s o1-preview model achieved a 42.7% success rate. GPT-4o followed with a 38.2% accuracy. And the smaller GPT-4o-mini scored only 8.6%. Anthropic did worse than OpenAI’s top model; the Claude-3.5-sonnet model managed to get just 28.9% of the answers correct.

All these models got an F, grade-wise, providing far more incorrect answers than correct ones. And the answers are super easy for a human.

Here are the kinds of questions that are asked by SimpleQA: 

• What year did the Titanic sink?

• Who was the first President of the United States?

• What is the chemical symbol for gold?

• How many planets are in our solar system?

• What is the capital city of France?

• Which river is the longest in the world?

• Who painted the Mona Lisa?

• What is the title of the first Harry Potter book?

• What does CPU stand for?

• Who is known as the father of the computer?

These are pretty simple questions for most people to answer, but they can present a problem for chatbots. One reason these tools struggled is that SimpleQA questions demand precise, single, indisputable answers. Even minor variations or hedging can result in a failing grade. Chatbots do better with open-ended overviews of even very complex topics but struggle to give a single, concise, precise answer. 

Also, the SimpleQA questions are short and self-contained and don’t provide a lot of context. This is why providing as much context as possible in the prompts that you write improves the quality of responses. 

Compounding the problem, LLMs often overestimate their own accuracy. SimpleQA queried chatbots on what they think is the accuracy of their answers; the models consistently reported inflated success rates. They feign confidence, but their internal certainty may be low.

LLMs don’t really think

Meanwhile, newly published research from MIT, Harvard, and Cornell University show that while LLMs can perform impressive tasks, they lack a coherent understanding of the world.

As one of their test examples, the researchers found that LLMs can generate accurate driving directions in complex environments like New York City. But when researchers introduced detours, the models’ performance dropped because they didn’t have an internal representation of the environment (as people do). Closing just 1% of streets in New York City led to a drop in the AI’s directional accuracy from nearly 100% to 67%. 

Researchers found that even when a model performs well in a controlled setting, it might not possess coherent knowledge structures necessary for random or diverse scenarios. 

The trouble with AI hallucinations

The fundamental problem we all face is this: Industries and individuals are already relying on LLM-based chatbots and generative AI tools for real work in the real world. The public, and even professionals, believe this technology to be more reliable than it actually is. 

As one recent example, OpenAI offers an AI transcription tool called Whisper, which hospitals and doctors are already using for medical transcriptions. The Associated Press reported that a version of Whisper was downloaded more than 4.2 million times from the open-source AI platform HuggingFace.

More than 30,000 clinicians and 40 health systems, including the Children’s Hospital Los Angeles, are using a tool called Nabla, which is based on Whisper but optimized for medical lingo. The company estimates that Nabla has been used for roughly seven million medical visits in the United States and France. 

As with all such AI tools, Whisper is prone to hallucinations. 

One engineer who looked for Whisper hallucinations in transcriptions found the in every document examined. Another found hallucinations in half of the 100 hours of Whisper transcriptions he analyzed. 

Professors from the University of Virginia looked at thousands of short snippets from a research repository hosted at Carnegie Mellon University. They found that nearly 40% of the hallucinations were “harmful or concerning.”

In one transcription, Whisper even invented a non-existent medication called “hyperactivated antibiotics.”

Experts fear the use of Whisper-based transcription will result in misdiagnoses and other problems.

What to do about AI hallucinations

When you get a diagnosis from your doctor, you might want to get a second opinion. Likewise, whenever you get a result from ChatGPT, Perplexity AI, or some other LLM-based chatbot, you should also get a second opinion.

You can use one tool to check another. For example, if the subject of your query has original documentation — say, a scientific research paper, a presentation, or a PDF of any kind — you can upload those original documents into Google’s NotebookLM tool. Then, you can copy results from the other tool, paste them into NotebookLM, and ask if it’s factually accurate. 

You should also check original sources. Fact-check everything. 

Chatbots can be great for learning, for exploring topics, for summarizing documents and many other uses. But they are not reliable sources of factual information, in general. 

What you should never, ever do is copy results from AI chatbots and paste it into something else to represent your own voice and your own facts. The language is often a bit “off.” The emphasis of points can be strange. And it’s a misleading practice. 

Worst of all, the chatbot you’re using could be hallucinating, lying or straight up making stuff up. They’re simply not as smart as people think.

** V kanadských horách vymírá lesní sob karibu ** Příčiny jeho zániku jsou jasné a vědci se na nich shodnou ** Na zákrocích, které by měly soba zachránit, už ale shoda nepanuje

The US Federal Trade Commission (FTC) is reportedly preparing to investigate Microsoft for potentially anticompetitive practices in its cloud computing division. This inquiry centers on whether Microsoft is abusing its market dominance by deploying restrictive licensing terms to dissuade customers from switching from its Azure platform to competitors, the Financial Times reported.

According to the report, the practices under scrutiny include sharply raising subscription fees for customers looking to switch providers, imposing high exit charges, and reportedly making Office 365 less compatible with competitor cloud services.

The investigation reflects the agency’s broader push, led by FTC Chair Lina Khan, to address Big Tech’s influence in sectors such as cloud services, with bipartisan support for curbing monopolistic practices.

In November 2023, the FTC began assessing cloud providers’ practices in four broad areas — competition, single points of failure, security, and AI — and sought feedback from stakeholders in academia, industry, and civil society.

The majority of the feedback the commission received highlighted concerns over licensing constraints that limit customers’ choices. 

Microsoft’s cloud strategy under fire

The inquiry reported by the Financial Times is still in its early stages, but an FTC challenge could significantly impact Microsoft’s cloud operations, which have grown rapidly in recent years.

“Interoperability and the fear of vendor lock-in are important criteria for enterprises selecting cloud vendors,” said Pareekh Jain, CEO of Pareekh Consulting. “This could create a negative perception of Microsoft. Previously, Microsoft faced a similar probe regarding the interoperability of Microsoft Teams.”

This scrutiny aligns with global regulatory focus: In the UK, the Competition and Markets Authority (CMA) is investigating Microsoft and Amazon following complaints about restrictive contracts and high “egress fees,” which make switching providers costly. Similarly, Microsoft recently sidestepped a formal probe in the European Union after it reached a multi-million-dollar settlement with rival cloud providers, addressing concerns of monopolistic practices.

Neither the FTC nor Microsoft had responded to questions about the reported investigation by press time.

Microsoft’s position in the cloud market

Cloud computing has rapidly expanded, with industry spending expected to reach $675 billion in 2024, according to Gartner. Microsoft controls roughly 20% of the global cloud market, second only to Amazon Web Services (31%) and ahead of Google Cloud (12%), according to Statista. Tensions have risen between the leading providers, with Microsoft accusing Google of using “shadow campaigns” to undermine its position by funding adversarial lobbying efforts.

“It seems Google has two ultimate goals in its astroturfing efforts: distract from the intense regulatory scrutiny Google is facing around the world by discrediting Microsoft and tilt the regulatory landscape in favor of its cloud services rather than competing on the merits,” Microsoft Deputy General Counsel Rima Alaily said in a statement in October.

AWS has also accused Microsoft of anticompetitive practices in the cloud computing segment and complained to the UK CMA.

These top cloud providers had already filed an antitrust case against Microsoft in 2022 alleging that Microsoft is using its software licensing terms to restrict European businesses’ options in selecting cloud providers for services like desktop virtualization and application hosting.

Previous FTC interventions and growing cloud sector scrutiny

This move follows the FTC’s legal challenge against Microsoft’s $75 billion acquisition of Activision Blizzard, which faced antitrust concerns around Microsoft’s cloud gaming business. While a federal court allowed the acquisition to proceed, the FTC’s appeal highlights its commitment to maintaining oversight of Big Tech’s market reach.

Since its inception, cloud computing has evolved from simple storage solutions to a cornerstone of AI development, with Microsoft, Amazon, and Google competing for contracts that power AI model training and deployment.

If pursued, this inquiry could lead to intensified regulations on Microsoft’s cloud strategy, underscoring the FTC’s commitment to protecting competitive markets in sectors increasingly dominated by a few key players. Neither the FTC nor Microsoft has publicly commented on the matter.

“Moving forward, all hyperscalers should commit to the interoperability of their cloud solutions in both intent and practice,” Jain noted, adding, “failing to do so may expose them to investigations that could damage their brand and business.”

Shared blame

If enterprises are finding themselves locked in to high costs, though, some of the blame may fall on them, suggested Yugal Joshi, a partner at Everest Group.

“Enterprises are happy signing highly discounted bundled deals, and when these financial incentives run out they complain about lock-in. Many of them already know what they are getting into but then are focused on near-term discounts over long-term interoperability and freedom to choose. Given the macro economy continues to struggle, price-related challenges are pinching harder,” Joshi said. “Therefore, clients are becoming more vocal and proactive about switching vendors if it saves them money.”

Microsoft has been a beneficiary of this, he said, because some clients are planning to move, and some have already moved, to its Dynamics platform from Salesforce.

In recent years, artificial intelligence (AI) has begun revolutionizing Identity Access Management (IAM), reshaping how cybersecurity is approached in this crucial field. Leveraging AI in IAM is about tapping into its analytical capabilities to monitor access patterns and identify anomalies that could signal a potential security breach. The focus has expanded beyond merely managing human