Forgot your password?
typodupeerror
AI

Teams of Coordinated GPT-4 Bots Can Exploit Zero-Day Vulnerabilities, Researchers Warn (newatlas.com) 27

New Atlas reports on a research team that successfuly used GPT-4 to exploit 87% of newly-discovered security flaws for which a fix hadn't yet been released. This week the same team got even better results from a team of autonomous, self-propagating Large Language Model agents using a Hierarchical Planning with Task-Specific Agents (HPTSA) method: Instead of assigning a single LLM agent trying to solve many complex tasks, HPTSA uses a "planning agent" that oversees the entire process and launches multiple "subagents," that are task-specific... When benchmarked against 15 real-world web-focused vulnerabilities, HPTSA has shown to be 550% more efficient than a single LLM in exploiting vulnerabilities and was able to hack 8 of 15 zero-day vulnerabilities. The solo LLM effort was able to hack only 3 of the 15 vulnerabilities.
"Our findings suggest that cybersecurity, on both the offensive and defensive side, will increase in pace," the researchers conclude. "Now, black-hat actors can use AI agents to hack websites. On the other hand, penetration testers can use AI agents to aid in more frequent penetration testing. It is unclear whether AI agents will aid cybersecurity offense or defense more and we hope that future work addresses this question.

"Beyond the immediate impact of our work, we hope that our work inspires frontier LLM providers to think carefully about their deployments."

Thanks to long-time Slashdot reader schwit1 for sharing the article.

Submission + - AI Companies Would Be Required To Disclose Copyright Training Data Under New Law (theverge.com)

An anonymous reader writes: Two lawmakers filed a bill requiring creators of foundation models to disclose sources of training data so copyright holders know their information was taken. The AI Foundation Model Transparency Act — filed by Reps. Anna Eshoo (D-CA) and Don Beyer (D-VA) — would direct the Federal Trade Commission (FTC) to work with the National Institute of Standards and Technology (NIST) to establish rules for reporting training data transparency. Companies that make foundation models will be required to report sources of training data and how the data is retained during the inference process, describe the limitations or risks of the model, how the model aligns with NIST’s planned AI Risk Management Framework and any other federal standards might be established, and provide information on the computational power used to train and run the model. The bill also says AI developers must report efforts to “red team” the model to prevent it from providing “inaccurate or harmful information” around medical or health-related questions, biological synthesis, cybersecurity, elections, policing, financial loan decisions, education, employment decisions, public services, and vulnerable populations such as children.

The bill calls out the importance of training data transparency around copyright as several lawsuits have come out against AI companies alleging copyright infringement. It specifically mentions the case of artists against Stability AI, Midjourney, and Deviant Art, (which was largely dismissed in October, according to VentureBeat), and Getty Images’ complaint against Stability AI. The bill still needs to be assigned to a committee and discussed, and it’s unclear if that will happen before the busy election campaign season starts. Eshoo and Beyer’s bill complements the Biden administration’s AI executive order, which helps establish reporting standards for AI models. The executive order, however, is not law, so if the AI Foundation Model Transparency Act passes, it will make transparency requirements for training data a federal rule.

Submission + - Ryzen vs. Meteor Lake: AMD's AI often wins, even on Intel's hand-picked tests (tomshardware.com)

Velcroman1 writes: Intel's new generation of "Meteor Lake" mobile CPUs herald a new age of "AI PCs," computers that can handle inference workloads such as generating images or transcribing audio without an Internet connection. Officially named "Intel Core Ultra" processors, the chips are the first to feature an NPU (neural processing unit) that's purpose-built to handle AI tasks. But there are few ways to actually test this feature at present: Software will need to be rewritten to specifically direct operations at the NPU.

So Intel has steered testers toward its Open Visual Inference and Neural Network Optimization (OpenVINO) AI toolkit. With those benchmarks, Tom's Hardware tested the new Intel chips against AMD — and surprisingly, AMD chips often came out on top, even on these handselected benchmarks. Clearly, optimization will take some time!

Google

Google Is Retiring Its Gmail Basic HTML View In January 2024 (bleepingcomputer.com) 79

Bill Toulas writes via Bleeping Computer reports: Google is notifying Gmail users that the webmail's Basic HTML view will be deprecated in January 2024, and users will require modern browsers to continue using the service. After that date, all users of the popular webmail service will automatically be redirected to the more modern Standard view, which supports all the latest usability and security features.

The basic HTML view is a stripped-down version of Gmail that does not offer users chat, spell checking, keyboard shortcuts, adding or importing contacts, setting custom "from" addresses, or using rich text formatting. This feature is designed for people living in areas with internet access, using older hardware with limited memory, or using legacy web browsers that do not support current HTML features.

However, one of the biggest reasons users use HTML view is that text-to-speech tools used by users with visual impairment are more reliable, as the Standard view introduces technical complexities that are harder for these tools to manage. Nonetheless, Google has decided to retire Gmail's HTML view without providing specific reasons.

Programming

WebAssembly 2023 Survey Finds Enthusiasm - and Some Challenges (infoworld.com) 34

An anonymous reader shared this report from InfoWorld: The uses of WebAssembly, aka Wasm, have grown far beyond its initial target of web applications, according to The State of WebAssembly 2023 report. But some developers remain skeptical. Released September 6 by the Cloud Native Computing Foundation (CNCF) and SlashData, in collaboration with the Linux Foundation, the report finds mostly optimism among software developers about future adoption of Wasm for web and non-web environments... However, about 22% of participants in the report indicated pessimism about Wasm adoption for either the web or non-web environments. Further, 83% of the respondents reported challenges with Wasm including difficulties with debugging and troubleshooting, different performance between runtimes, lack of consistent developer experiences between runtimes, lack of learning materials, and compatibility issues with certain browsers.

The report finds that respondents are using WebAssembly across a wide range of software projects including data visualization (35%), internet of things (32%, artificial intelligence (30%), games (28%), back-end services (27%), edge computing (25%), and more. While Wasm is still primarily used to develop web applications (58%), this is changing thanks to WASI (WebAssembly System Interface), which provides a modular interface for Wasm...

Other findings of the State of WebAssembly 2023 report:

- When migrating existing applications to Wasm, 30% of respondents experience performance benefits of more than 50%.
- JavaScript is the most popular language used with Wasm applications. But Rust stands out in popularity in Wasm projects compared to other use cases...

The article says WebAssembly developers were attracted by "faster loading times, the ability to explore new use cases and technologies, and the ability to share code between projects. Improved performance over JavaScript and efficient execution of computationally intensive tasks also were cited."
Businesses

Troubled Toshiba Announces Buyout Offer Led By Japan Businesses (apnews.com) 18

Toshiba announced a 2 trillion yen ($14 billion) tender offer on Monday in a move that would take it private, as the scandal-tarnished Japanese electronics and energy giant seeks to turn itself around. From a report: The tender offer led by a buyout fund of major Japanese banks and companies called Japan Industrial Partners starts Tuesday and is priced at 4,620 yen ($32) a share. Chairperson Akihiro Watanabe asked shareholders to back the proposal, saying it is the only option for Toshiba to return to its former strength.

[...] Tokyo-based Toshiba also reported a 25 billion yen ($176 million) loss for the April-June quarter on 704 billion yen ($5 billion) in sales, down nearly 5% from the previous year. It did not give a full fiscal year profit projection, citing uncertainties in its computer chip business. If the proposal succeeds, it will be a major step in Toshiba's yearslong turnaround effort, allowing it to delist from the Tokyo Stock Exchange.

The Courts

'Extremely Remorseful' Lawyers Confronted by Judge Over 'Legal Gibberish' Citations from ChatGPT (apnews.com) 78

The Associated Press reports: Two apologetic lawyers responding to an angry judge in Manhattan federal court blamed ChatGPT Thursday for tricking them into including fictitious legal research in a court filing... [Attorney Steven A. Schwartz] told U.S. District Judge P. Kevin Castel he was "operating under a misconception ... that this website was obtaining these cases from some source I did not have access to." He said he "failed miserably" at doing follow-up research to ensure the citations were correct.

"I did not comprehend that ChatGPT could fabricate cases," Schwartz said...

The judge confronted Schwartz with one legal case invented by the computer program. It was initially described as a wrongful death case brought by a woman against an airline only to morph into a legal claim about a man who missed a flight to New York and was forced to incur additional expenses. "Can we agree that's legal gibberish?" Castel asked.

Schwartz said he erroneously thought that the confusing presentation resulted from excerpts being drawn from different parts of the case. When Castel finished his questioning, he asked Schwartz if he had anything else to say. "I would like to sincerely apologize," Schwartz said. He added that he had suffered personally and professionally as a result of the blunder and felt "embarrassed, humiliated and extremely remorseful."

He said that he and the firm where he worked — Levidow, Levidow & Oberman — had put safeguards in place to ensure nothing similar happens again.

An attorney for the law firm also told the judge that lawyers have historically had a hard time with technology, particularly new technology. "And it's not getting easier."
IOS

Apple Announces iOS 17 With StandBy Charging Mode, Better Autocorrect (theverge.com) 44

At WWDC today, Apple debuted iOS 17. "Highlights include new safety features, a built-in journaling app, a new nightstand mode, redesigned contact cards, better auto-correct and voice transcription, and live voicemail," reports The Verge. "And you'll be able to drop the 'hey' from 'Hey Siri.'" From the report: Your contact book is getting an update with a new feature called posters, which turns contact cards into flashy marquee-like images that show up full-screen on your recipient's iPhone when you call them. They use a similar design language as the redesigned lock screens, with bold typography options and the ability to add Memoji, and will work with third-party VoIP apps. There's also a new live transcription feature for voicemail that lets you view a transcript of the message a caller is leaving in real time. You can choose to ride it out or pick up the call, and it's all handled on-device. You'll also be able to leave a message on FaceTime, too.

Some updates to messages include the ability to filter searches with additional terms, a feature that jumps to the most recent message so you can catch up more easily, transcriptions of voice messages -- similar to what the Pixel 7 series introduced -- and a series of new features called Check In that shares your live location and status with someone else. It can automatically send a message to a friend when you've arrived home, and it can share your phone's battery and cell service status to help avoid confusion if you're in a dead zone. Stickers are getting an overhaul, too, with the ability to add any emoji or photo cutout as a "sticker" positioned on iMessages or anywhere within the system. Live photos can be turned into animated stickers, too, and you can now add effects to stickers.

AirDrop gets an update to send contact information -- cleverly called NameDrop -- which will send your selected email addresses and phone numbers (and your poster) just by bringing two iPhones near each other. It also works between an iPhone and an Apple Watch. Photos can be shared the same way, and if the file is a big one, it's now possible to move out of range while continuing the download. iOS 17 also includes keyboard updates, including enhancements to autocorrect. It now relies on a new language model for better accuracy, plus an easier shortcut to revert to the original word you wrote if necessary. There's now in-line predictive typing and sentence-level autocorrections to correct more grammatical mistakes. It'll finally learn your favorite cuss words, too; Apple's Craig Federighi even made a "ducking" joke onstage. Dictation uses a new AI model, too, that's more accurate.

A new app called Journal automatically suggests moments that you might want to commemorate in a journal entry. Your entries can include photos, music, and activities, and you can schedule reminders for yourself to start writing. It's end-to-end encrypted, too, to keep things private. StandBy is a new mode for charging that turns the screen into a status display with the date and time. It can show information from Live Activities, widgets, and smart stacks and automatically turns on when your phone is in landscape mode while charging. You can swipe to the right to see some of your highlighted photos, and it comes with customizable clockfaces. Siri will surface visual results in StandBy, and the display shifts to a red tone at night to avoid disrupting sleep. Last but not least, Siri gets a boost, too, and finally lets you drop the "hey" from "Hey Siri." It will also recognize back-to-back commands.
iOS 17 is available to developers today, with a public beta released next month.

Comment Something ambiguous (Score 1) 44

"Package managers haven't been sexy. Until now." , how the term "sexy" could be applied to a package manager? It means appealing?, in which sense?. assuming it means something that appeals (in some way). What could be more appealing that it works the way it was intended to be?

Slashdot Top Deals

It seems intuitively obvious to me, which means that it might be wrong. -- Chris Torek

Working...