Daily Archives: March 26, 2018

New Slack Tool Lets Your Boss Potentially Access Far More of Your Data Than Before, without notification

Posted on by

According to Slack’s new guidelines, however, Compliance Exports will be replaced by “a self-service export tool” on April 20th. Previously, an employer had to request a data dump of all communications to get access to private channels and direct messages. This new tool should streamline things so they can archive all your shit-talk and time-wasting with colleagues on a regular basis. The tool not only makes it easy for an admin to access everything with a few clicks, it also enables automatic exports to be scheduled on a daily, weekly, or monthly basis. An employer still has to go through a request process to get the tool, but Slack declined to elaborate on what’s involved in that process.

What’s particularly concerning is that Compliance Exports were designed so they notified users when they were enabled, and future exports only covered data that was generated after that notification. A spokesperson for Slack confirmed to Gizmodo that this won’t be the case going forward. The new tool will be able to export all of the data that your Slack settings previously retained. Whereas before, if you were up on Slack policy, you could feel pretty comfortable that your private conversations were private unless you got that Compliance Exports notification. After the notification, you’d want to make sure you didn’t discuss potentially sensitive topics in Slack. Now, anyone who was under the impression that they were relatively safe might have some cause to worry.

Source: New Slack Tool Lets Your Boss Potentially Access Far More of Your Data Than Before

2 + 2 = 4, er, 4.1, no, 4.3… Nvidia’s Titan V GPUs spit out ‘wrong answers’ in scientific simulations

Posted on by

Nvidia’s flagship Titan V graphics cards may have hardware gremlins causing them to spit out different answers to repeated complex calculations under certain conditions, according to computer scientists.

The Titan V is the Silicon Valley giant’s most powerful GPU board available to date, and is built on Nv’s Volta technology. Gamers and casual users will not notice any errors or issues, however folks running intensive scientific software may encounter occasional glitches.

One engineer told The Register that when he tried to run identical simulations of an interaction between a protein and enzyme on Nvidia’s Titan V cards, the results varied. After repeated tests on four of the top-of-the-line GPUs, he found two gave numerical errors about 10 per cent of the time. These tests should produce the same output values each time again and again. On previous generations of Nvidia hardware, that generally was the case. On the Titan V, not so, we’re told.

We have repeatedly asked Nvidia for an explanation, and spokespeople have declined to comment. With Nvidia kicking off its GPU Technology Conference in San Jose, California, next week, perhaps then we’ll get some answers.

All in all, it is bad news for boffins as reproducibility is essential to scientific research. When running a physics simulation, any changes from one run to another should be down to interactions within the virtual world, not rare glitches in the underlying hardware.

[…]

Unlike previous GeForce and Titan GPUs, the Titan V is geared not so much for gamers but for handling intensive parallel computing workloads for data science, modeling, and machine learning.

And at $2,999 (£2,200) a pop, it’s not cheap to waste resources and research time on faulty hardware. Engineers speaking to The Register on condition of anonymity to avoid repercussions from Nvidia said the best solution to these problems is to avoid using Titan V altogether until a software patch has been released to address the mathematical oddities.

Source: 2 + 2 = 4, er, 4.1, no, 4.3… Nvidia’s Titan V GPUs spit out ‘wrong answers’ in scientific simulations • The Register

This kind of reminds me of when Intel brought out the Pentium. They couldn’t count either.

Siri Can Expose Your Hidden Notifications Even When Your Phone Is Locked

Posted on by

With iOS 11, Apple added a new setting that lets you choose whether you want previews of your notifications to appear on your lock screen. By default, iOS shows a preview of your notifications only when your phone is unlocked, via some form of authentication like Face ID. But Siri will read your notifications from third-party apps aloud even if your phone is locked. This means anyone with physical access to your phone could hear messages meant just for you. MacMagazine first reported the issue after one of its readers noticed the peculiar behavior.

We tested the issue with some texts and Facebook Messenger exchanges. When my partner pressed the iPhone’s side button and asked Siri to “read my notifications,” the snitch of a voice assistant read the contents of my Facebook Messenger notifications aloud.

However, notifications from Apple’s own Messages app remained properly hidden behind the locked screen, leaving my texts secure. If you ask Siri to read your messages from Apple’s app aloud, you’ll be greeted by Siri telling you to unlock your iPhone if you want those juicy deets.

We’ve reached out to Apple for comment.

Notification contents in iOS 11 are hidden on locked devices by default. With an iPhone X, that means you can look at your phone (or tap the fingerprint sensor on other iOS devices) and watch the contents of your notifications appear. You can edit the option by visiting Settings > Notifications and toggling between the “Always,” “Never,” and “When Unlocked” options, although changing the setting to “Never” does not appear to address the issue. For now, your best bet may simply be to only allow Siri to be activated when your phone is unlocked.

Source: Siri Can Expose Your Hidden Notifications Even When Your Phone Is Locked [Updated]

IBM claims its machine learning library is 46x faster than TensorFlow • The Register

Posted on by

Analysis IBM boasts that machine learning is not just quicker on its POWER servers than on TensorFlow in the Google Cloud, it’s 46 times quicker.

Back in February Google software engineer Andreas Sterbenz wrote about using Google Cloud Machine Learning and TensorFlow on click prediction for large-scale advertising and recommendation scenarios.

He trained a model to predict display ad clicks on Criteo Labs clicks logs, which are over 1TB in size and contain feature values and click feedback from millions of display ads.

Data pre-processing (60 minutes) was followed by the actual learning, using 60 worker machines and 29 parameter machines for training. The model took 70 minutes to train, with an evaluation loss of 0.1293. We understand this is a rough indicator of result accuracy.

Sterbenz then used different modelling techniques to get better results, reducing the evaluation loss, which all took longer, eventually using a deep neural network with three epochs (a measure of the number of times all of the training vectors are used once to update the weights), which took 78 hours.

[…]

Thomas Parnell and Celestine Dünner at IBM Research in Zurich used the same source data – Criteo Terabyte Click Logs, with 4.2 billion training examples and 1 million features – and the same ML model, logistic regression, but a different ML library. It’s called Snap Machine Learning.

They ran their session using Snap ML running on four Power System AC922 servers, meaning eight POWER9 CPUs and 16 Nvidia Tesla V100 GPUs. Instead of taking 70 minutes, it completed in 91.5 seconds, 46 times faster.

They prepared a chart showing their Snap ML, the Google TensorFlow and three other results:

IBM_Snap_ML_chart

A 46x speed improvement over TensorFlow is not to be sneezed at. What did they attribute it to?

They say Snap ML features several hierarchical levels of parallelism to partition the workload among different nodes in a cluster, takes advantage of accelerator units, and exploits multi-core parallelism on the individual compute units

  1. First, data is distributed across the individual worker nodes in the cluster
  2. On a node data is split between the host CPU and the accelerating GPUs with CPUs and GPUs operating in parallel
  3. Data is sent to the multiple cores in a GPU and the CPU workload is multi-threaded

Snap ML has nested hierarchical algorithmic features to take advantage of these three levels of parallelism.

Source: IBM claims its machine learning library is 46x faster than TensorFlow • The Register

22 Ambassadors Recommend the One Book to Read Before Visiting Their Country

Posted on by

Preparing for a visit to a foreign country can often be overwhelming, with no shortage of things to learn before you go. Where should you eat? Where should you stay? What do you tip? More so than this service information, though, is a sense of cultural understanding that’s hard to put your finger on. With this in mind, language learning app Babbel asked foreign ambassadors to the U.S. to pick the book they believe first-time visitors to their country should read before they arrive. Their answers may surprise you.

Source: 22 Ambassadors Recommend the One Book to Read Before Visiting Their Co – Condé Nast Traveler

The Hilarious (and Terrifying?) Ways Algorithms Have Outsmarted Their Creators

Posted on by

. As research into AI grows ever more ambitious and complex, these robot brains will challenge the fundamental assumptions of how we humans do things. And, as ever, the only true law of robotics is that computers will always do literally, exactly what you tell them to.

A paper recently published to ArXiv highlights just a handful of incredible and slightly terrifying ways that algorithms think. These AI were designed to reflect evolution by simulating generations while other competing algorithms conquered problems posed by their human masters with strange, uncanny, and brilliant solutions.

The Surprising Creativity of Digital Evolution: A Collection of Anecdotes from the Evolutionary Computation and Artificial Life Research Communities covers some 27 anecdotes from various computer science projects and is worth a read on its own, but here are a few highlights:

  • A study designed to evolve moving creatures generated ‘hackers’ that would break their simulation by clipping into the ground and using the “free energy” of the simulation’s correction to speed towards their goal.
  • An AI project which pit programs against each other in games of five-in-a-row Tic-Tac-Toe on an infinitely expansive board surfaced the extremely successful method of requesting moves involving extremely long memory addresses which would crash the opponent’s computer and award a win by default.
  • A program designed to simulate efficient ways of braking an aircraft as it landed on an aircraft carrier learned that by maximizing the force on landing—the opposite of its actual goal—the variable holding that value would overflow and flip to zero, creating a practically catastrophic, but technically perfect solution.
  • A test that challenged a simulated robot to walk without allowing its feet to touch the ground saw the robot flip on its back and walk on its elbows (or knees?) as shown in the tweet above.
  • A study to evolve a simulated creature that could jump as high as possible yielded top-heavy creatures on tiny poles that would fall over and spin in mid-air for a technically high ‘jump.’

While the most amusing examples are clearly ones where algorithms abused bugs in their simulations (essentially glitches in the Matrix that gave them superpowers), the paper outlines some surprising solutions that could have practical benefits as well. One algorithm invented a spinning-type movement for robots which would minimize negative effect of inconsistent hardware between bots, for instance.

As the paper notes in its discussion—and you may already be thinking—these amusing stories also reflect the potential for evolutionary algorithms or neural networks to stumble upon solutions to problems that are outside-the-box in dangerous ways. They’re a funnier version of the classic AI nightmare where computers tasked with creating peace on Earth decide the most efficient solution is to exterminate the human race.

The solution, the paper suggests, is not fear but careful experimentation. As humans gain more experience in training these sorts of algorithms, and tweaking along the way, experts gain a better sense of intuition. Still, as these anecdotes prove, it’s basically impossible to avoid unexpected results. The key is to be prepared—and to not hand over the nuclear arsenal to a robot for its very first test.

Source: The Hilarious (and Terrifying?) Ways Algorithms Have Outsmarted Their Creators

AI software that can reproduce like a living thing? Yup, boffins have only gone and done it • The Register

Posted on by

A pair of computer scientists have created a neural network that can self-replicate.

“Self-replication is a key aspect of biological life that has been largely overlooked in Artificial Intelligence systems,” they argue in a paper popped onto arXiv this month.

It’s an important process in reproduction for living things, and is an important step for evolution through natural selection. Oscar Chang, first author of the paper and a PhD student at Columbia University, explained to The Register that the goal was to see if AI could be made to be continually self improving by mimicking the biological self-replication process.

“The primary motivation here is that AI agents are powered by deep learning, and a self-replication mechanism allows for Darwinian natural selection to occur, so a population of AI agents can improve themselves simply through natural selection – just like in nature – if there was a self-replication mechanism for neural networks.”

The researchers compare their work to quines, a type of computer program that learns to produces copies of its source code. In neural networks, however, instead of the source code it’s the weights – which determine the connections between the different neurons – that are being cloned.

The researchers set up a “vanilla quine” network, a feed-forward system that produces its own weights as outputs. The vanilla quine network can also be used to self-replicate its weights and solve a task. They decided to use it for image classification on the MNIST dataset, where computers have to identify the correct digit from a set of handwritten numbers from zero to nine.

[…]

The test network required 60,000 MNIST images for training, another 10,000 for testing. And after 30 runs, the quine network had an accuracy rate of 90.41 per cent. It’s not a bad start, but its performance doesn’t really compare to larger, more sophisticated image recognition models out there.

The paper states that the “self-replication occupies a significant portion of the neural network’s capacity.” In other words, the neural network cannot focus on the image recognition task if it also has to self-replicate.

“This is an interesting finding: it is more difficult for a network that has increased its specialization at a particular task to self-replicate. This suggests that the two objectives are at odds with each other,” the paper said.

Chang explained he wasn’t sure why this happened, but it’s what happens in nature too.

Source: AI software that can reproduce like a living thing? Yup, boffins have only gone and done it • The Register

SpaceX blasted massive plasma hole in Earth’s ionosphere

Posted on by

A SpaceX rocket ripped a humongous hole in Earth’s ionosphere during a launch in California last year and may have impaired GPS satellites.

The Falcon 9 rocket was blasted from Vandenberg Air Force Base on 24 August last year. It was carrying the Formosat-5, an Earth observation satellite, built by the Taiwan’s National Space Organization.

As the rocket reached supersonic speeds minutes after liftoff, it sent gigantic circular shock acoustic waves (SAWs) rippling through the atmosphere. These SAWs continued to extend outwards for about 20 minutes at a whopping speed of about 629 to 726 meters per second – equivalent between 0.021 and 0.0242 per cent of the maximum velocity of a sheep in a vacuum in Reg units.

It’s the largest rocket-induced SAW on record, according to a paper published in the Advancing Earth and Space Science journal. The plume tore a gigantic hole, approximately 900 kilometers (559 miles) in diameter stretching to 1,770,000 square kilometers (1,099,827 square miles), more than four times the total area of California.

The ionosphere is a region of the Earth’s upper atmosphere that contains a soup of particles that have been ionized from the Sun’s rays. The researchers estimate that the SAW blasted electrons away, causing the total electron content – the concentration of electrons along a one-meter squared region – to deplete by as much as 70 per cent.

The researchers reckon the fluctuations were probably pretty small and could have led to a range of errors in GPS navigation of up to a meter – not significant enough to cause major problems until the SAW dissipated.

The particularly large circular size of the shock wave was down to the way the Falcon 9 rocket flew. It had a nearly vertical trajectory, compared to most satellite launches that fly over a horizontal trajectory before the satellites are booted into orbit.

Disruptions in the ionosphere are to be expected for every rocket launch and are also detected during volcano blasts and solar flares.

“Understanding how the rocket launches affect our upper atmosphere and space environment is important as these anthropogenic space weather events are expected to increase at an enormous rate in the near future,” the paper concluded.

Source: SpaceX blasted massive plasma hole in Earth’s ionosphere • The Register

‘R2D2’ stops disk-wipe malware before it executes evil commands

Posted on by

Purdue University researchers reckon they’ve cracked how to protect data against “disk-wipe” malware.

Led by Christopher Gutierrez, the team has created a shim of software that analyses write buffers before they reach storage, and if the write is destructive, it steps in to preserve the data targeted for destruction.

Dubbed R2D2 – “Reactive Redundancy for Data Destruction Protection” – their work will be published in the May issue of the journal Computers & Security.

In this [PDF] pre-press version of the paper, the researchers explained their technique. The inspection is implemented in the virtual machine monitor (VMM) using virtual machine introspection (VMI).

“This has the benefit that it does not rely on the entire OS as a root of trust”, they wrote, and they claimed a latency penalty of between 1 and 4 per cent for batch tasks, and 9 to 20 per cent for interactive tasks.

'R2D2' architecture

Click to enlarge

The system has been tested against various secure delete tools and malware like Shamoon and Stonedrill, and they claim complete success against “all the wiper malware samples in the wild that we experimented with”.

R2D2 intercepts the open file and write file system calls on a guest VM. When it detects an open file request, it checks “all open system calls” to see if the file is already open for writing.

“If the system call requests a write permission, a policy determines if the file should be protected based on a blacklist or whitelist,” they wrote.

Whitelisted files are those not protected; if a blacklisted file is requested, “If the file is on the blacklist, we take a snapshot of the file system because the file is considered critical to system stability.”

If the attacker tries to open a file on neither list, “R2D2 takes a temporary checkpoint of the file system, and subsequent write system calls are analysed, according to analysis policy, to determine if the write is suspect”.

Source: ‘R2D2’ stops disk-wipe malware before it executes evil commands • The Register

How to Find Out Everything Facebook Knows About You

Posted on by

If you can’t bring yourself to delete your Facebook account entirely, you’re probably thinking about sharing a lot less private information on the site. The company actually makes it pretty easy to find out how much data it’s collected from you, but the results might be a little scary.

When software developer Dylan McKay went and downloaded all of his data from Facebook, he was shocked to find that the social network had timestamps on every phone call and SMS message he made in the past few years, even though he says doesn’t use the app for calls or texts. It even created a log of every call between McKay and his partner’s mom.

To get your own data dump, head to your Facebook Settings and click on “Download a copy of your data” at the bottom of the page. Facebook needs a little time to compile all that information, but it should be ready in about 10 minutes based on my own experience. You’ll receive a notification sending you to a page where you can download the data—after re-entering your account password, of course.

The (likely huge) file downloads onto your computer as a ZIP. Once you extract it, open the new folder and click on the “index.html” to view the data in your browser.

Be sure to check out the Contact Info tab for a list of everyone you’ve ever known and their phone number (creepy, Facebook). You can also scroll down to the bottom of the Friends tab so see what phase of your life Facebook thinks you’re in —I got “Starting Adult Life.”

Source: How to Find Out Everything Facebook Knows About You