06 August 2017

Yoon Ha Lee – Povestiri din ‘Machineries of Empire’

in Bucharest, Romania

De vreme ce tocmai terminasem Ninefox Gambit, m‑am gândit să continui cu două povestiri din același univers. Acestea relatează evenimente mult mai vechi, de pe vremea când Hexarhat‑ul conținea o a șaptea facțiune, Liozh, între timp epurată pentru erezie, și Shuos Jedao era încă în viață, ca asasin și spion Shuos. Scenele de aici nu impactează semnificativ povestea principală din roman, și bănuiesc că ar fi greu de urmărit fără să cunoști contextul, așa că eu aș recomanda să le citiți după primul roman, chiar dacă cronologic îl preced.

04 August 2017

The New York Times: “Artificial Intelligence is Stuck. Here’s How to Move it Forward.”

To get computers to think like humans, we need a new A.I. paradigm, one that places “top down” and “bottom up” knowledge on equal footing. Bottom-up knowledge is the kind of raw information we get directly from our senses, like patterns of light falling on our retina. Top-down knowledge comprises cognitive models of the world and how it works.

Deep learning is very good at bottom-up knowledge, like discerning which patterns of pixels correspond to golden retrievers as opposed to Labradors. But it is no use when it comes to top-down knowledge. If my daughter sees her reflection in a bowl of water, she knows the image is illusory; she knows she is not actually in the bowl. To a deep-learning system, though, there is no difference between the reflection and the real thing, because the system lacks a theory of the world and how it works. Integrating that sort of knowledge of the world may be the next great hurdle in A.I., a prerequisite to grander projects like using A.I. to advance medicine and scientific understanding.

Gary Marcus

Almost exactly what I wrote a couple of months ago in response to a long article about Google’s AI initiatives. The research cannot go forward if it’s tied to commercial, short-term goals, where each company is trying to protect its own data and methods, instead of collaborating as researchers in fundamental physics do.

03 August 2017

Super Ventures Blog: “Why is ARKit better than the alternatives?”

All the above use the same VIO system (Tango & ARKit even use the same code base originally developed by FlyBy!). Neither Hololens or Tango use the Depth Camera for tracking (though I believe they are starting to integrate it to assist in some corner cases). So why is ARKit so good?

The answer is that ARKit isn’t really any better than Hololens (I’d even argue that Hololens’ tracker is the best on the market) but Hololens hardware isn’t widely available. Microsoft could have shipped the Hololens tracker in a Windows smartphone, but I believe they chose not to for commercial reasons (I.e. it would have added a fair bit of cost & time to calibrate the sensors for a phone that would sell in low volumes, and a MSFT version of ARKit would not by itself convince developers to switch from iOS/Android)

Google also could easily have shipped Tango’s VIO system in a mass market Android phone over 12 months ago, but they also chose not to. If they did this, then ARKit would have looked like a catch up, instead of a breakthrough. I believe (without hard confirmation) that this was because they didn’t want to have to go through a unique sensor calibration process for each OEM, where each OEMs version of Tango worked not as well as others, and Google didn’t want to just favor the handful of huge OEMs (Samsung, Huawei etc.) where the device volumes would make the work worthwhile.

Matt Miesnieks

I’ve seen way too much excitement around ARKit on Twitter since Apple announced it and not enough useful demos.

I think there are two key questions developers should answer before investing in new technologies:

02 August 2017

The New York Times: “Apple’s Silence in China Sets a Dangerous Precedent”

Search Apple’s website for a letter from Mr. Cook issuing a public rebuke of China’s intrusion into his customers’ privacy and freedom of expression — you won’t find it. The company has not fully tested its political and economic leverage in China. It hasn’t tested the public’s immense love of its products. It hasn’t publicly threatened any long-term consequences — like looking to other parts of the world to manufacture its products.

The company’s silence may be tactical; the Chinese government, the conventional thinking goes, does not take well to public rebuke. Yet Apple’s quiet capitulation to tightening censorship in one of its largest markets is still a dangerous precedent.

Farhad Manjoo

What a different Apple than the company that last year took a stance against government intrusion by refusing to break iOS encryption at the request of the FBI. I for one think the explanation is simple enough: back then, Apple figured there was little cost to resisting the US government, that they will eventually back down, while the public image benefit would be high, especially considering their armies of fans in the US, ready to defend their every decision.

01 August 2017

TechCrunch: “Grab gets $2B from Didi and SoftBank to fuel bid to defeat Uber in Southeast Asia”

Essentially, both Didi and SoftBank are doubling down on the belief that Grab has what it takes to defeat Uber in Southeast Asia, the same way that Didi did in China when Uber agreed to sell its local business last August. That hope of defeating the U.S. firm was reignited this month when Uber agreed to sell its business in Russia to local rival Yandex.


Grab operates in 36 cities across seven countries in Southeast Asia, where it claims over 50 million downloads from users and 1.1 million drivers on its platform. Its services are primarily focused on licensed taxis and private cars, but Grab also offers motorbike taxis, shuttle bus services and carpooling in a selection of countries.

Jon Russell

Interesting dynamic in the ride-sharing market: after trying to hyperexpand and control the worldwide market, Uber is being forced into a series of tactical retreats across the board. After ceding China to Didi, Russia to Yandex, the latest news show rivals are challenging Uber in Southeast Asia and Didi is also investing in Taxify, an Eastern European startup. The Chinese company seems intent on challenging Uber on multiple fronts, now that their American rival is hampered by internal scandals, lost its CEO and is dealing with a complicated lawsuit brought on by Waymo.

31 July 2017

Kirkville: “Batch Processing in Apple Photos”

Apple’s Photos app does not allow you to perform batch processing. However, there is a way that you can quickly apply the same changes to multiple photos.

Start with any photo in edit mode; to edit a photo, select it and press Return. Photos switches to its editing interface with controls at the right side of the window. Make whatever changes you want to the photo: adjust the color, contrast, brightness, or apply a filter.


You can then switch to another photo in edit mode and paste these adjustments. To do this, choose Image > Paste Adjustments, press Command-Shift-V, or right-click on the photo and choose Paste Adjustments. Photos applies all the adjustments you made to the first photo, with the exception of cropping or rotation.

Kirk McElhearn

If I understand this correctly, you can’t actually select multiple photos and paste adjustments at once, you have to: select each individual photo, apply, repeat. To call this ‘batch processing’ is a massive overstatement. On top of that, you have to manually rotate and crop photos later.

To be fair, Microsoft’s modern Photos app doesn’t seem to be any better at this, but the older Photo Gallery (now discontinued) allowed users to select multiple photos and perform adjustments on all at once.

ZDNet: “New details emerge on Fruitfly, a near-undetectable Mac backdoor”

Apple released security patches for Fruitfly earlier this year, but variants of the malware have since emerged. The core of the malware is an obfuscated perl script using antiquated code, with indicators in the code that suggest the malware may go back almost half a decade or more, the security firm said. Nevertheless, the malware still works well on modern versions of macOS, including Yosemite. Fruitfly connects and communicates with a command and control server, where an attacker can remotely spy on and control an infected Mac.


He found that he could take complete control of an infected Mac, including its keyboard and mouse, take screenshots of the display, remotely switch on the webcam, and modify files. The malware can also run commands in the background, and even kill the malware’s process altogether – likely in an effort to avoid detection.

Zack Whittaker

If five-year-old malware can successfully infect macOS, what does that say about Apple’s security – or about the level of resources dedicated to macOS development?