跳至主要内容

2023-05-30

Let us serve you, but don't bring us down

  • Archive.org experienced an hour-long outage due to tens of thousands of requests per second launched from 64 virtual hosts on Amazon Web Services, which brought the site down for all users.
  • Engineers scrambled on a Sunday afternoon on a holiday weekend to work on the issue, and the site got back up by blocking those IP addresses, but another 64 addresses started the same type of activity a couple of hours later.
  • Those wanting to use Archive.org's materials in bulk should start slowly and ramp up, and if starting a large project, should contact the organization at info@archive.org to avoid overwhelming the site.

Industry Reactions

  • System administrator shares experience of adding a recaptcha token to thwart scraping attack that worked like DDoS from multiple IP addresses
  • Commenters suggest using Proof-of-Work algorithms, including Bitcoin's, as alternative to CAPTCHAs to prevent bot attacks, but concerns raised about initial hurdle for users who don't own cryptocurrency and limitations of alternative PoW algorithms
  • Discussion centers around strategies to prevent DDoS-style attacks, including rate limiting and delaying responses to bots, as well as concerns about overload on Internet Archive servers and potential rate limiting for excessive scraping or downloads

India ruling party's IT cell used AI to show smile on arrested protesters' faces

  • Some of India's top wrestlers were detained by Delhi Police after demanding the arrest of the WFI president and BJP MP Brij Bhushan Sharan Singh over allegations of sexual harassment.
  • A photo of detained female wrestlers allegedly smiling in a police vehicle started circulating on social media, but it was later uncovered that the photo had been edited using an AI app to show them smiling.
  • This incident highlights the potential misuse of social media and AI technology for misleading propaganda purposes, and the importance of fact-checking and critical thinking when consuming news on social media platforms.

Industry Reactions

  • AI-generated images can be used to spread false narratives, eroding trust in visual information
  • There is a need for developing ways to authenticate images and videos to combat digital forgeries
  • Experts fear people will become more vulnerable to subtle yet effective tricks as newer technology makes it easier to produce authentic yet false narratives

HBO Max new Captcha system

  • This text does not contain any meaningful content to summarize.
  • This website is best experienced by visiting the website directly.
  • Please visit the website to read the article.

Industry Reactions

  • HBO Max's new Captcha system, powered by Arkose Labs, has received criticism for its poor design and difficulty level.
  • Users and experts discuss the need for alternative, more effective and accessible verification systems to prevent bot attacks and ensure online security.
  • Some commenters express frustration with the rebranding of HBO streaming service, along with glitches and bugs in the platform.

300ms Faster: Reducing Wikipedia's total blocking time

  • Total Blocking Time (TBT) measures the sum of all long tasks on a browser's main thread that blocks user interaction between First Contentful Paint and Time to Interactive.
  • Long JavaScript execution and attaching event listeners to every element can lead to high TBT.
  • By removing unnecessary JavaScript and optimizing existing JavaScript, Wikipedia reduced TBT by nearly 300ms on low-end devices, improving its overall performance.

Industry Reactions

  • Wikipedia is faster for users who are not logged in
  • Low-hanging fruit suggestions have been made to improve website performance, such as using Astro and speculative prefetching
  • Discussions include whether Wikipedia should serve the mobile version by default or detect device capabilities

Setris – Tetris with Sand Physics

  • 'Setris' is a new game that combines the mechanics of Tetris with sand physics.
  • The game includes three different play modes and five levels of difficulty.
  • Players can download the game for free or name their price to support the developer.

Industry Reactions

  • Setris, a new Tetris-style game developed with Java and LibGDX, is popular among programmers and gamers.
  • Some users want the game to run in a browser instead of as a download, others suggest names like Sandtris and Brick Game to avoid trademark issues.
  • Users have reported issues running the game on specific systems and the download size is large due to the bundled JRE required by the game's Java-based development.

An open-source alternative to QuickBooks

  • This website is best experienced by visiting the website directly.
  • Please visit the website to read the article.

Industry Reactions

  • Users express frustration with Intuit and discuss open-source alternatives to QuickBooks.
  • Concerns are raised about the project's use of Discord as a support channel.
  • Commenters discuss challenges with interoperability with different countries' banking and taxes.

Wikipedia had wrong Vatican flag for years – now incorrect flags are everywhere

  • The Vatican flag, featuring a yellow and white background and a papal coat of arms, is often displayed by Catholics around the world as a symbol of their identity.
  • However, many of the flags available for purchase or use may be incorrect, as multiple versions with minor differences have been circulated online, including one sourced from an erroneous Wikipedia file.
  • This situation highlights a lack of clear specifications from the Vatican regarding the official design of the flag, and the need for greater standardization and clarity in the design specifications, particularly as the Vatican flag is seen globally.

Industry Reactions

  • Wikipedia displayed the wrong Vatican flag for years, leading to widespread usage of the incorrect design due to lack of specificity and multiple acceptable versions of the flag.
  • Discussions on the accuracy of symbols on flags and coats of arms on Wikipedia, with mixed views on the importance of factual accuracy and the challenges of maintaining uniformity in flag designs globally.
  • An ongoing edit war and discussion on Wikipedia regarding the correct flag designs for certain countries, highlighting the limitations of crowdsourced information and the challenges of maintaining accurate and reliable data in the digital age.

Donut: OCR-Free Document Understanding Transformer

  • Donut is an OCR-free document understanding transformer that achieves state-of-the-art performance on visual document understanding tasks without off-the-shelf OCR engines or APIs.
  • SynthDoG, a synthetic document generator, helps Donut's pre-training be flexible on various languages and domains.
  • Pre-trained models and web demos for various visual document understanding tasks are available, such as document parsing, document classification, document VQA, and text reading task.

Industry Reactions

  • Donut, an OCR-Free Document Understanding Transformer, has been released on Github.
  • Donut received mixed reviews from Hacker News users compared to other OCR tools such as Google Cloud OCR or Azure OCR.
  • Despite criticisms, Donut showed potential for future document understanding and classification applications.

Removing support for forwarded ports

  • This text appears to be the header of a website, likely belonging to Mullvad VPN.
  • It includes links to various pages, such as help, servers, pricing, blog, and information about privacy and VPNs.
  • Without further context or information, it is unclear if there is anything new or exciting relevant to tech news.

Industry Reactions

  • Mullvad is removing its port forwarding feature, which is used for peer-to-peer file sharing and gaming, due to abuse by malicious actors.
  • Private trackers and seedboxes are suggested as alternatives for torrenting, though accessing private trackers may require an invitation from established members.
  • Some users are concerned about losing privacy and the ability to access geographically restricted content without port forwarding. Alternative VPN services that support port forwarding such as AirVPN and Hide.me were recommended by users.

Production AI systems are hard

  • Radiologists have a complex job that requires a 3D world model of the brain and its dynamics, making it nearly impossible to replace them entirely with AI systems.
  • Automated systems cannot make determinations like radiologists because radiologists have grounded brain models, making it challenging to train AI machines to replicate their performance.
  • Even creating an algorithm that could automate the work of radiologists could prove difficult given the extensive domain-specific knowledge necessary, making it unlikely that anyone will succeed in doing so.

Industry Reactions

  • AI can pre-draft 90% of diagnostic studies in radiology, but concerns exist about radiologists becoming complacent and misreading due to not reviewing output
  • Challenges of sharing medical data exist due to legal and regulatory issues, differences in imaging protocols and disease prevalence between hospitals, and concerns over privacy
  • AI can assist radiologists in triaging interpretation priorities and flagging abnormal scans, but human diagnostician input remains crucial for rare and challenging conditions.