Intel gets slapped with a class action lawsuit for overlooking a security vulnerability in its processors, Microsoft tries harder to improve AI’s reasoning capabilities, New York based healthcare company faces ransomware attack, and HPE and Nvidia offer the building blocks of a supercomputer to customers.
These and more top tech stories on Hashtag Trending
I’m your host, James Roy.
A new class action lawsuit is accusing Intel of blatantly ignoring a major security flaw in its processors for five years, and continuing to sell defective products to customers.
The lawsuit alleges that Intel was made aware of the flaw, dubbed Downfall, by two separate reports in 2018 but did not fix it, leaving customers with vulnerable CPUs, which were later crippled because of performance-killing limitations.
Google researchers who discovered the flaw also explained that the vulnerability can cause secret user data to be revealed.
Apparently, the company was busy dealing with the Spectre and Meltdown flaws in its CPU architecture at the time, and decided to overlook the Downfall vulnerability.
The class action states that the flaw was not fixed for three more generations of Intel’s x86 chips, which means that now customers that use software for photo and video editing, gaming, and encryption are unfairly reaping the consequences with products that are either “egregiously vulnerable” to attacks or must be slowed down “beyond recognition” to fix the Downfall flaw.
Intel has declined to comment on the lawsuit.
Big Tech has been hell bent on making LLM’s reason better, so Microsoft has devised yet another system.
It’s called Everything of Thought and it combines different techniques, including reinforcement learning and Monte Carlo Tree Search (MCTS) to allow language models to generalise efficiently when presented with unknown problems.
Microsoft tried out the system on a number of challenging tasks including Game of 24, the 8-Puzzle, and the Pocket Cube, which reportedly yielded impressive results and even outperformed contemporaries.
The company has not expressed plans to integrate the new system into its products. Plus, despite its advances, Everything of Thought has not reached a state of 100 per cent reliability.
This crusade to improve LLM’s reasoning is longstanding, but we only found out after ChatGPT struggled painfully with Math. Subsequently, Microsoft released an approach titled “Algorithm of Thoughts” (AoT), to refine AI’s algorithmic reasoning, and later even went on to assess the ethical decision-making skills of its models.
Google and Meta are also pursuing research to improve AI’s reasoning.
Source: Analytics India Mag
Russian-linked ransomware gang, BlackCat, has threatened to leak healthcare solutions provider Henry Schein’s data, which was stolen during an apparent ransomware attack.
Henry Schein, based in New York, has over 300 products and a customer base of over 1 million.
BlackCat claims it stole 35 terabytes of data and threatened to publish it online if Henry Schein refused to pay an undisclosed ransom amount.
Ransom negotiations are falling through, according to the gang, and they claim they will continue to release more data daily.
The company had taken certain systems offline to contain the apparent ransomware attack, negatively impacting a significant portion of its manufacturing, and distribution business. However, it did not say whether the system outage was because of the attack.
CISA revealed that between November 2021 and March 2022, BlackCat had compromised over 60 organizations worldwide.
Michigan-based healthcare company, McLaren Health Care, also confirmed that the sensitive personal and health information of 2.2 million patients was compromised during a cyberattack earlier this year. BlackCat claimed responsibility for the attack.
McLaren currently faces at least three class action lawsuits related to the cyberattack.
Google came up with the IP Protection proposal to prevent marketers from tracking Chrome users across different websites by anonymizing IP addresses. And guess who’s not happy: a marketing advocacy group.
The group, called The Movement for an Open Web, has filed a complaint with the UK’s Competition and Markets Authority (CMA), claiming it’s harmful to rival internet advertising businesses.
Google’s plan is designed to run Chrome browser connections through two proxies, one operated by Google and one operated by a third-party (eg, Cloudflare), so that the true public IP address of the user is obscured.
The company plans to make this the IP protection default for the Chrome browser.
The organization contends that the proposal will make the provision of child protection services more difficult for ISPs, and that Google is acting in a non-competitive manner by denying data to its rivals, while continuing to make use of it.
In a comment to The Register, Google said, “Critics claiming that our IP Protection proposal is self-preferential to Google are either knowingly misrepresenting the facts, or simply don’t understand what is being proposed. The IP Protection proposal includes a two-hop proxy system, with one proxy being operated by a third-party. This means that Google, along with the rest of the industry, will not be able to see both the client IP address and user destination.”
IP Protection is expected to appear in a future version of Chrome, possibly as soon as January 2024 with the debut of Chrome 122.
Source: The Register
HPE and Nvidia are teaming up to offer customers the building blocks to produce a mini version of an AI supercomputer.
The offering is designed to be simpler for organizations to get up and running with AI training, thanks to a preconfigured and pretested stack.
The modular machine is based on HPE’s architecture and Nvidia’s Grace Hopper Superchip, and includes a software stack comprising tools from both.
The software stack includes HPE Machine Learning Development Environment, a platform to train up generative AI models and Nvidia’s AI Enterprise suite, including TensorFlow, PyTorch, Nvidia’s RAPIDS and more. Customers also get HPE’s Cray Programming Environment, a bunch of tools for developing, porting, and debugging code.
Juston Hotard, HPE’s executive vice president for HPC AI and Labs, said that AI training usually requires an entire data center, running over hundreds and thousands of nodes, operating at a scale that’s consistent with what we see in supercomputers.
Naturally, the offering is designed to deliver on that, but HPE declined to say how much it would cost.
Source: The Register
Hashtag Trending goes to air 5 days a week with a special weekend interview show we call “the Weekend Edition.”
You can get us anywhere you get audio podcasts and there is a copy of the show notes at itworldcanada.com/podcasts
For those interested in cybersecurity you can also check out our hit cybersecurity podcast featuring Howard Solomon and called CybersecurityToday. It’s rated as one of North America’s top 10 tech podcasts.
I’m your host, James Roy – have a Terrific Tuesday!