Two years since its inception, Google‘s Project Green Light is turning heads and making waves in the world of sustainable urban planning. Revealed at the company’s Sustainability ‘23 event, this initiative uses artificial intelligence (AI) to transform ordinary traffic signals into pollution-fighting tools.

Google’s Project Green Light has been deployed in Rio, Jakarta, Manchester & more

So how does this work? Simply put, Google deploys machine learning to examine Maps data, gauging traffic congestion and vehicle wait times at stoplights. This data then trains AI systems to adjust the traffic light timings, aiming to minimize both idle time and the start-and-stop nature of city driving. This isn’t just tech for tech’s sake. Google is after a lofty goal: aiding its partners in cutting down one gigaton of carbon emissions by the end of this decade.

Initially a pilot in four Israeli intersections, the program’s success—a reduction in fuel use and intersection delays by up to 20 percent—led to its expansion to 12 global cities. Locations now include Rio de Janeiro, Manchester, and Jakarta, and Google has plans to broaden its reach in 2024.

Yael Maguire, Google’s VP of Geo Sustainability, said the system is unique in its scalability and cost-effectiveness, suggesting a potential 30% reduction in traffic stops. What’s more, city engineers can start seeing the impact within weeks of implementation. Results from Manchester show emission levels and air quality improving by as much as 18%.

Not to overlook the role of Google Maps here. By optimizing routes, the service has already averted around 2.4 million metric tons of carbon emissions. That’s comparable to sidelining half a million petrol cars for an entire year. Pretty impressive, right? Let us know if you think so too, in the comments below!

RELATED:

• End of the Road: Family Accuses Google Maps of Leading to Husband’s Death
• You can Now Tag your Favorite Locations on Google Maps Using Emojis
• Download the best GCam APK for OnePlus Nord CE 3 5G

(Via)

The post Google’s Project Green Light is Successfully Solving Traffic Issues, Thanks to AI appeared first on Gizmochina.

The tech industry is no alien to Meta and Qualcomm. Both these massive tech organizations have a strong hold over different aspects of the industry. As per the latest reports, Qualcomm and Meta have joined forces to optimize the execution of Meta’s Llama 2 large language models directly on-device, eliminating the need for heavy reliance on cloud services. By running generative AI models like Llama 2 on smartphones, PCs, VR/AR headsets, and vehicles, developers can offer users more private, reliable, and personalized experiences while saving on cloud costs.

On-device AI can function seamlessly even in airplane mode

The partnership aims to make on-device Llama 2-based AI implementations available, empowering developers to create innovative AI applications. Customers, partners, and developers will be able to harness the power of these AI experiences to build intelligent virtual assistants, productivity tools, content creation platforms, and entertainment applications. One significant advantage of on-device AI is its capability to function in areas with no internet connectivity or even in airplane mode.

Durga Malladi, senior vice president, and general manager of technology, planning, and edge solutions businesses at Qualcomm Technologies, expressed support for Meta’s approach to open and responsible AI. Malladi emphasized the company’s commitment to driving innovation and facilitating AI development for developers of all sizes, envisioning generative AI to run seamlessly on both the cloud and edge devices such as smartphones, laptops, vehicles, and IoT devices.

The collaboration between Meta and Qualcomm Technologies builds upon their longstanding history of driving technology innovation to deliver premium device experiences. This partnership includes joint research and product engineering efforts aimed at supporting the Llama ecosystem. Qualcomm Technologies’ leadership in on-device AI uniquely positions it to support Llama, given its extensive presence at the edge, powering billions of smartphones, vehicles, XR headsets, glasses, PCs, and IoT devices with industry-leading AI hardware and software solutions.

Scheduled to be available from 2024 onwards, Qualcomm Technologies plans to offer Llama 2-based AI implementations on devices powered by Snapdragon. Developers can get started today by optimizing their applications for on-device AI using the Qualcomm® AI Stack, a dedicated set of tools enabling more efficient AI processing on Snapdragon. This breakthrough technology makes on-device AI possible even in small, thin, and light devices. The future of technology seems promising, and full of surprises. We are here for it all!

RELATED:

• Fairphone 5 emerges on Geekbench with 8GB RAM and a Qualcomm Snapdragon SoC
• Qualcomm launches Snapdragon 4 Gen 2; this phone could be first to feature it
• AYA NEO KUN teased as the brand’s best-ever Windows handheld

(Source)

The post Qualcomm and Meta Collaborate to Bring On-Device AI with Llama 2 Models appeared first on Gizmochina.

In a surprising breakthrough, Microsoft has unveiled its latest language model, Phi-1, with 1.3 billion parameters. Contrary to the conventional belief that larger models perform better, Microsoft’s approach focuses on the quality of the training data. Phi-1, trained on a meticulously curated “textbook-level” dataset, has outperformed GPT-3.5 with 100 billion parameters.

The training time for Microsoft’s model was merely 4 days

Microsoft’s Phi-1 language model, built on the Transformer architecture, has garnered attention for its impressive performance. The team behind Phi-1 placed emphasis on the quality of training data, a departure from the prevailing trend of increasing model stack size. Using a high-quality dataset consisting of “textbook-level” content sourced from the internet, the Microsoft team processed the information using GPT-3.5. With the assistance of 8 Nvidia A100 GPUs, the training process was completed in just four days.

According to Microsoft, the focus on enhancing training data quality, rather than escalating parameter count, has yielded promising results. In comparative tests, Phi-1 achieved an accuracy score of 50.6%, surpassing GPT-3.5’s performance of 47% with a staggering 175 billion parameters.

Microsoft plans to open source Phi-1 on HuggingFace, further strengthening the accessibility and collaborative potential of this language model. This isn’t the first instance of Microsoft developing a smaller language model; previously, they introduced Orca, a 13 billion parameter model trained on synthetic data using GPT-4. Even Orca has proven to outperform ChatGPT. The research paper on Phi-1 has been published on arXiv, providing detailed insights into its architecture and training methodology. For those interested in exploring the technical aspects, the paper offers a comprehensive overview of Phi-1’s development.

Microsoft’s Phi-1 language model challenges the notion that increased stack size is essential for improved performance. By focusing on high-quality training data, Phi-1 has showcased remarkable accuracy, surpassing even larger models. The open sourcing of Phi-1 further demonstrates Microsoft’s commitment to advancing the field of natural language processing.

RELATED:

• The US Congress Implements New Rules for ChatGPT Usage in Offices
• ChatGPT to Metaverse: Huawei Chairman predicts 10x computing power increase by 2030
• Best VR Games 2023: The Ultimate List of Virtual Adventures and Experiences

(Via)

The post Microsoft’s Latest Small Language Model Outperforms ChatGPT with a Fraction of Parameters appeared first on Gizmochina.

Artifact, a personalized news reader developed by Instagram co-founders, Kevin Systrom and Mike Krieger, has now launched and is open to the public without any sign-up required. The app originally required a phone number and an invitation to try it out, which attracted initial demand but kept it out of reach for many potential users. The company received approximately 160,000 sign-ups while on the waitlist, and users outside the U.S. were unable to try Artifact, as a U.S. phone number was necessary for sign-up. With the removal of the waitlist and phone number requirement, users will now have instant access to Artifact upon its first launch. The app’s first social feature has been introduced, and feedback controls have been added to improve the news reading experience. 

Artifact’s goal is to provide tools to allow readers to personalize and expand their feed. Users can manage their interests by blocking or pausing publishers, or selecting and unselecting general interest categories. In addition, a new feature allows users to upload their contacts to see a signal that a particular article is popular within their network. This feature is different from Twitter‘s Top Articles feature, which shows articles popular with the people you follow, as Artifact’s feature is more privacy-focused. It does not tell you who read it or how many of them read it, so it keeps privacy. 

Systrom stated that the underlying technology requires a certain amount of data and a number of people using it to offer the best experience. The company believes the app is now ready for a larger audience. The newly added stats feature shows users the categories they have read, recent articles they have read within those categories, and the publishers they have been reading the most. Very few news publishing apps follow this approach. The reading is grouped more narrowly by specific topics rather than general categories. The broader goal is to expand the social experience to include a way to discuss news articles within Artifact itself. 

Artifact is a U.S.-based version of China‘s Toutiao or Japan‘s SmartNews. Apps like these utilize algorithms and machine learning to form a curated set of news articles for each user based on how they engage with the app’s content. Systrom claims that while Artifact is similar to these other personalized news readers, the “devil is in the details”. The machine learning used in Artifact is based on the transformer, which was invented in 2017 at Google. It’s interesting to note that the ‘T’ in ChatGPT stands for Transformer as well. 

Systrom compares the emergence of Artifact with the founding of Instagram when there were already other image-sharing apps available. There has to be something that distinguishes new technology from existing technology in order to create a niche in the market. In this case, that would be machine learning and other similar algorithms.

RELATED:

• Meta to Start Charging Users for Blue Checkmark on Facebook & Instagram
• Samsung Collabs With Instagram, TikTok and Snapchat
• Best Chromebooks For Students To Improve Productivity

The post Artifact, Instagram Co-Founder’s App Provides AI-Powered News & is Open To The Public appeared first on Gizmochina.

Google has done an appreciable job with the amalgamation of Pixel 6 and machine learning. Each feature on the smartphone is enhanced and furnished with the help of AI. On top of that, the Google Tensor chipset further improvises some features and the performance of the overall smartphone. Among the exciting Pixel 6 features, Live Translate is another useful feature that is currently exclusive for the Pixel 6 series. However, a little workaround can help you get in a feature on your old Pixel smartphone. Let’s see how in the guide below

Must See: Some Google Pixel 6 units get Camera-based Heart Rate & Respiratory Tracking

The Pixel 6 exclusive feature – Live Translate, is now easily available for old Pixels. Unlike what Google said at the Pixel 6 launch event, the Live Translate feature works flawlessly on Pixel 4a without Google Tensor. Magic Eraser was also acclaimed to be optimized by Google Tensor on Pixel 6, but it also worked on previous Pixel devices without any problem.

You can easily install the ASI – Android System Intelligence ported from Pixel 6 to grab the feature without any additional setting. As revealed by Twitter user @abhinav_937, the feature runs without any flaw on a non-rooted device. He tested it on the Pixel 4a by installing the latest Pixel 6 ASI apk.

What is Live Translate on Pixel 6?

Google Translation is already a productivity tool, having millions of daily users communicating across languages. It’s been widely used across the globe via different platforms including Twitter, YouTube, and other important services to ease communications. The search giant is actively working to improve its communication and translation services with the advent of new techniques and ways.

In such a step, Google introduced Live Translate with Pixel 6 series to translate the text across different third-party apps. Using the Google Tensor AI prowess, the app offers automatic translation of text messages on Google Messages, WhatsApp, and another conversation app in your native or default language. It works two-way, as Google Translate even translates your reply into the sender’s native language. Live Translate works in sync with Gboard while typing the reply and further enhances the functionality.

It is a full overhaul of Google Translate capabilities in a more integrated manner. Live Translate feature can even transcribe the audio in a video to your native language in real-time. The offline capability of Live Translate is an additional advantage that boosts its overall usability.

Steps to get Live Translate on Old Pixels

It is quite easy to enable the Live Translate feature on your old Pixel smartphone. Precisely follows the steps below:

1. Download the Pixel 6 latest ASI ( Android System Intelligence) from here.
2. Install the APK on the old Pixel device without any additional settings.
3. Now go to Settings > System > Live Translate will be available.
4. Download the preferred language, and see it working in your favorite messaging app. Note. It is working on messaging app only.
5. Enjoy!!

Currently, the feature is working, but we’re not sure about the future as Google may switch it off on old Pixels via the backend. If you found it working apart from the ASI version, then do let us know via the comments below.

Related: Some Pixel 6 features can be enabled on older Pixels through this Google Camera update

The post How to get Pixel 6 Live Translate feature on Old Pixels appeared first on Gizmochina.

Apple, the known smartphone maker, has bought an Edge computing based startup for a massive sum of 200 million US Dollars. A new report had surfaced after a close source to the smartphone OEM shared the information to a reporting agency called GeekWire.

Xnor.ai is an Edge computing based startup that had branched out from the nonprofit Allen Institure for AI back in 2017. Apple has confirmed the transaction of the large sum to acquire the firm. The Cupertino based giant also mentioned that it acquires smaller businesses every now and then but generally doesn’t incline in sharing the purpose or plans of the buyout.

Initially, Xnor.ai focused on machine learning with efficient algorithms being developed that could run on various low tier hardware, or IoT devices (Internet of Things). This includes electronics like a security camera which consumer low amount of power to function and had a one-track streamlined function. However, with Xnor.ai’s algorithms, these gadgets could accomplish complex tasks like object recognition; which is a task that generally requires powerful processing from cloud based computer via an internet connection.

Editor’s Pick: TSMC 10nm chip shipments contributed to just 1% of wafer revenue, 7nm at 35% in Q4 2019

According to the source, the amount of 200 million US Dollars for the acquisition is just an estimate so the numbers are not concrete. Although, the numbers and investment are still ranging in tens of million which surpass any of its previous investments raised by a large margin. Xnor.ai is also expected to move to the Seattle offices of Apple which, as stated by the report.

The acquisition for an Edge computing based firm by Apple is forward thinking move from the company. Other industry giants like Google, Microsoft, and even Amazon are heavily investing in this field. With Xnor.ai, Apple plans on making its products work more independently from its cloud computing services to possibly include functions like facial recognition, natural language processing and augmented reality as well.

The implications of the field also extend towards security and privacy, since edge computing places an emphasis on local processing. Thus, Apple can make its devices more secure and even improve its imaging algorithms with machine learning. Notably, iPhones are a loose example of edge computing with local storage of biometric data (facial recognition or fingerprint scanners) and its processing.

UP NEXT: Can Redmi Note 8 survive the harsh sub-zero temperatures in Antarctica? We’ll find out soon!

(Source,Via)

The post Apple products may get Smarter as the company buys an Edge Computing Startup for $200 million appeared first on Gizmochina.