Connect with us

Biz & IT

China’s Baidu says its answer to Alexa is now on 200M devices

Published

on

A Chinese voice assistant has been rapidly gaining ground in recent months. DuerOS, Baidu’s answer to Amazon’s Alexa, reached over 200 million devices, China’s top search engine announced on its Weibo official account last Friday.

To put that number into context, more than 100 million devices pre-installed with Alexa have been sold, Amazon recently said. Google just announced it expected Assitant to be on 1 billion devices by the end of this month.

Voice interaction technology is part of Baidu’s strategy to reposition itself from a heavy reliance on search businesses towards artificial intelligence. The grand plan took a hit when the world-renown scientist Lu Qi stepped down as Baidu’s chief operating officer, though the segment appears to have scored healthy growth lately, with DuerOS more than doubling from a base of 90 million installs since last June.

When it comes to how many devices actually use DuerOS regularly, the number is much less significant: 35 million machines a month at the time Baidu’s general manager for smart home devices announced the figure last November.

Like Alexa, which has made its way into both Amazon-built Echo speakers and OEMs, DuerOS also takes a platform play to power both Baidu-built and third-party devices.

Interestingly, DuerOS has achieved all that with fewer capabilities and a narrower partnership network than its American counterpart. By the end of 2018, Alexa could perform more than 56,000 skills. Devices from over 4,500 brands can now be controlled with Alexa, says Amazon. By comparison, Baidu’s voice assistant had 800 different skills, its chief architect Zhong Lei revealed at the company’s November event. It was compatible with 85 brands at the time.

This may well imply that DuerOS’s allies include heavy-hitters with outsize user bases. Baidu itself could be one as it owns one of China’s biggest navigation app, which is second to Alibaba’s AutoNavi in terms of number of installs, according to data from iResearch. Baidu said in October that at least 140 million people had activated the voice assistant of its Maps service.

Furthermore, Baidu speakers have managed to crack a previously duopolistic market. A report from Canalys shows that Baidu clocked in a skyrocketing 711 percent quarter-to-quarter growth to become China’s third-biggest vendor of smart speakers during Q3 last year. Top players Alibaba and Xiaomi, on the other hand, both had a sluggish season.

While Baidu deploys DuerOS to get home appliances talking, it has doubled down on smart vehicles with Apollo . The system, which the company calls the Android for autonomous driving, counted 130 OEMs, parts suppliers and other forms of partners as of last October. It’s attracted global automakers Volvo and Ford who want a foothold in China’s self-driving movement. Outside China, Apollo has looked to Microsoft Azure Cloud as it hunts for international partnerships.

Baidu has yet to prove commercial success for its young AI segment, but its conversational data trove holds potential for a lucrative future. Baidu became China’s top advertising business in part by harnessing what people search on its engine. Down the road, its AI-focused incarnation could apply the same data-crunching process to what people say to their machines.

Source link

Source link

Continue Reading
Click to comment

Leave a Reply

Your email address will not be published.

Biz & IT

Better than JPEG? Researcher discovers that Stable Diffusion can compress images

Published

on

Enlarge / These jagged, colorful blocks are exactly what the concept of image compression looks like.

Benj Edwards / Ars Technica

Last week, Swiss software engineer Matthias Bühlmann discovered that the popular image synthesis model Stable Diffusion could compress existing bitmapped images with fewer visual artifacts than JPEG or WebP at high compression ratios, though there are significant caveats.

Stable Diffusion is an AI image synthesis model that typically generates images based on text descriptions (called “prompts”). The AI model learned this ability by studying millions of images pulled from the Internet. During the training process, the model makes statistical associations between images and related words, making a much smaller representation of key information about each image and storing them as “weights,” which are mathematical values that represent what the AI image model knows, so to speak.

When Stable Diffusion analyzes and “compresses” images into weight form, they reside in what researchers call “latent space,” which is a way of saying that they exist as a sort of fuzzy potential that can be realized into images once they’re decoded. With Stable Diffusion 1.4, the weights file is roughly 4GB, but it represents knowledge about hundreds of millions of images.

Examples of using Stable Diffusion to compress images.
Enlarge / Examples of using Stable Diffusion to compress images.

While most people use Stable Diffusion with text prompts, Bühlmann cut out the text encoder and instead forced his images through Stable Diffusion’s image encoder process, which takes a low-precision 512×512 image and turns it into a higher-precision 64×64 latent space representation. At this point, the image exists at a much smaller data size than the original, but it can still be expanded (decoded) back into a 512×512 image with fairly good results.

While running tests, Bühlmann found that images compressed with Stable Diffusion looked subjectively better at higher compression ratios (smaller file size) than JPEG or WebP. In one example, he shows a photo of a candy shop that is compressed down to 5.68KB using JPEG, 5.71KB using WebP, and 4.98KB using Stable Diffusion. The Stable Diffusion image appears to have more resolved details and fewer obvious compression artifacts than those compressed in the other formats.

Experimental examples of using Stable Diffusion to compress images. SD results are on the far right.
Enlarge / Experimental examples of using Stable Diffusion to compress images. SD results are on the far right.

Bühlmann’s method currently comes with significant limitations, however: It’s not good with faces or text, and in some cases, it can actually hallucinate detailed features in the decoded image that were not present in the source image. (You probably don’t want your image compressor inventing details in an image that don’t exist.) Also, decoding requires the 4GB Stable Diffusion weights file and extra decoding time.

While this use of Stable Diffusion is unconventional and more of a fun hack than a practical solution, it could potentially point to a novel future use of image synthesis models. Bühlmann’s code can be found on Google Colab, and you’ll find more technical details about his experiment in his post on Towards AI.

Continue Reading

Biz & IT

Apps can pose bigger security, privacy threat based on where you download them

Published

on

Google and Apple have removed hundreds of apps from their app stores at the request of governments around the world, creating regional disparities in access to mobile apps at a time when many economies are becoming increasingly dependent on them.

The mobile phone giants have removed over 200 Chinese apps, including widely downloaded apps like TikTok, at the Indian government’s request in recent years. Similarly, the companies removed LinkedIn, an essential app for professional networking, from Russian app stores at the Russian government’s request.

However, access to apps is just one concern. Developers also regionalize apps, meaning they produce different versions for different countries. This raises the question of whether these apps differ in their security and privacy capabilities based on region.

In a perfect world, access to apps and app security and privacy capabilities would be consistent everywhere. Popular mobile apps should be available without increasing the risk that users are spied on or tracked based on what country they’re in, especially given that not every country has strong data protection regulations.

My colleagues and I recently studied the availability and privacy policies of thousands of globally popular apps on Google Play, the app store for Android devices, in 26 countries. We found differences in app availability, security, and privacy.

While our study corroborates reports of takedowns due to government requests, we also found many differences introduced by app developers. We found instances of apps with settings and disclosures that expose users to higher or lower security and privacy risks depending on the country in which they’re downloaded.

Geoblocked apps

The countries and one special administrative region in our study are diverse in location, population and gross domestic product. They include the US, Germany, Hungary, Ukraine, Russia, South Korea, Turkey, Hong Kong, and India. We also included countries like Iran, Zimbabwe, and Tunisia, where it was difficult to collect data. We studied 5,684 globally popular apps, each with over 1 million installs, from the top 22 app categories, including Books and Reference, Education, Medical, and News and Magazines.

Our study showed high amounts of geoblocking, with 3,672 of 5,684 globally popular apps blocked in at least one of our 26 countries. Blocking by developers was significantly higher than takedowns requested by governments in all our countries and app categories. We found that Iran and Tunisia have the highest blocking rates, with apps like Microsoft Office, Adobe Reader, Flipboard, and Google Books all unavailable for download.

Continue Reading

Biz & IT

Russia plans “massive cyberattacks” on critical infrastructure, Ukraine warns

Published

on

gwengoat | Getty Images

The Ukrainian government on Monday warned that the Kremlin is planning to carry out “massive cyberattacks” targeting power grids and other critical infrastructure in Ukraine and in the territories of its allies.

“By the cyberattacks, the enemy will try to increase the effect of missile strikes on electricity supply facilities, primarily in the eastern and southern regions of Ukraine,” an advisory warned. “The occupying command is convinced that this will slow down the offensive operations of the Ukrainian Defence Forces.”

Monday’s advisory alluded to two cyberattacks the Russian government carried out—first in 2015 and then almost exactly one year later—that deliberately left Ukrainians without power during one of the coldest months of the year. The attacks were seen as a proof-of-concept and test ground of sorts for disrupting Ukraine’s power supply.

The first attack repurposed a known piece of malware, called BlackEnergy, created by Kremlin-backed hackers. The attackers used this new BlackEnergy3 malware to break into the corporate networks of Ukrainian power companies and then further encroach into the supervisory control and data acquisition systems the companies used to generate and transmit electricity. The hack allowed the attackers to use legitimate functionality commonly found in power distribution and transmission to trigger a failure that caused more than 225,000 people to go without power for more than six hours.

The 2016 attack was more sophisticated. It used a new piece of malware written from scratch specifically designed for hacking electric grid systems. The new malware—which goes by the names Industroyer and Crash Override—was notable for its mastery of the arcane industrial processes used by Ukraine’s grid operators. Industroyer natively communicated with those systems to instruct them to de-energize and then re-energize substation lines.

“The experience of cyberattacks on Ukraine’s energy systems in 2015 and 2016 will be used when conducting operations,” the Ukrainian government said on Monday.

Monday’s advisory comes two weeks after Ukrainian forces recaptured vast swaths of territory in Kharkiv and other cities that had been under Russian control for months. Russian President Vladimir Putin last week called for the mobilization of 300,000 Russian citizens to bolster the country’s military invasion of Ukraine.

The move, which was the first time since World War II that Russia has done so, has prompted protests and a diaspora of mostly male Russians fleeing the country. A pivot to increased reliance on hacking by the country’s military could be seen as a way to achieve objectives without further straining the ongoing personnel shortage.

It’s hard to assess the chances of a successful hacking campaign against Ukraine’s power grids. Earlier this year, Ukraine’s CERT-UA said it successfully detected a new strain of Industroyer inside the network of a regional Ukrainian energy firm. Industroyer2 reportedly was able to temporarily switch off power to nine electrical substations but was stopped before a major blackout could be triggered.

“We don’t have any direct knowledge or data to make an assessment on Ukraine’s capability to defend its grid, but we do know that CERT-UA stopped the deployment of INDUSTROYER.V2 malware that targeted Ukraine’s electric substations earlier this year,” Chris Sistrunk, technical manager of Mandiant Industrial Control Systems Consulting, wrote in an email. “Based on that, and what we know about the Ukrainian people’s overall resolve, it’s increasingly clear that one of the reasons cyberattacks in Ukraine have been dampened is because its defenders are very aggressive and very good at confronting Russian actors.”

But researchers from Mandiant and elsewhere also note that Sandworm, the name for the Kremlin-backed group behind the power grid hacks, is among the most elite hacking groups in the world. They are known for stealth, persistence, and remaining hidden inside targeted organizations for months or even years before surfacing.

Besides an attack on electrical grids, Monday’s advisory also warned of other forms of disruptions the country expected Russia to ramp up.

“The Kremlin also intends to increase the intensity of DDoS attacks on the critical infrastructure of Ukraine’s closest allies, primarily Poland and the Baltic states,” the advisory stated. Since February, researchers have said pro-Russian threat actors have been behind a steady stream of distributed denial-of-service attacks targeting Ukraine and its allies.

Continue Reading

Trending