Google is piloting "Learn About," an innovative AI learning tool using the LearnLM AI model. Unlike standard chatbots, "Learn About" offers an interactive approach, incorporating quizzes, lists, and contextual information for a deeper understanding. Currently in limited release, the platform emphasizes reliable educational sources, setting it apart from general AI tools.

Save 30% on the Google Nest WiFi Pro at Amazon.

Google Gemini Live may let you talk to it about your uploaded files

Google Fonts - Different emoticons are shown when no fonts are found.

/via Thomas Park

Google Maps - When crossing state lines, you’re welcomed to the new state.

/via Thomas Park

Google Forms – Can tell from the language in your question that you will want checkboxes, and changes the question type automatically.

/via Jack Hallahan

Google live sports reporting - There’s a moving green line underneath the ‘Live’ label to indicate and reassure the user that the page is active and reporting the live scores. 

/via @sophieriches08

Google Calendar – When you select one day on Google Calendar it recommends dinner, when you select a week, it recommends a vacation

Google Docs - When you rename an Untitled Document in Google Docs it takes the first line of the document as a suggestion for a title.  Text is highlighted so one stroke of Delete key and it’s gone.

Google – When you search for “kerning”, every instance of the word has a large amount of kerning.

Google Docs - The capitalization options in the ‘Format’ menu are demonstrated in the capitalization names.

/via Eyal Shahar

España es el primer país democrático en el que Google cierra su servicio de noticias.

Nadie le preguntó al pueblo cubano qué responder a Google ante su oferta de regalarnos la instalación gratuita de internet en Cuba.

Google wants to solve Pixel phones' overheating and battery issues. But not without removing some serious features to save money.

Gmail добавил интеграцию с Google Calendar в свою боковую панель Gemini, которая уже работает с Drive, Docs, Sheets и Slides. Это позволяет пользователям Gmail быстро получать ответы на вопросы о календаре и создавать события в режиме диалога.

Nvidia, Oracle, Google, Dell and 13 other companies reported how long it takes their computers to train the key neural networks in use today. Among those results were the first glimpse of Nvidia’s next generation GPU, the B200, and Google’s upcoming accelerator, called Trillium. The B200 posted a…

SEPM union that represents magazine staff in France alleges that the search giant was going to start testing a scheme under which it would remove from search results some articles from media with which it was clashing over rights for use of online news.

Google Chrome for iOS has rolled out several new features that aim to make finding information and completing tasks easier. The web browser now lets iPhone users add text queries to visual searches via Google Lens for more helpful and relevant results. It also introduces new ways to save space on iPhone, find better deals while shopping online, and view maps of addresses quickly.

If you happen to break or damage the screen of your Google Pixel 9 Pro Fold, know that its replacement display won’t be cheap.

The post The Google Pixel 9 Pro Fold’s replacement display will not come cheap appeared first on Phandroid.

The new features will be available on Pixel phones first, and will arrive for more Android devices soon.

The post Google Unveils Upcoming Online Protection Tools for Android Users appeared first on Phandroid.

Nvidia, Oracle, Google, Dell and 13 other companies reported how long it takes their computers to train the key neural networks in use today. Among those results were the first glimpse of Nvidia’s next generation GPU, the B200, and Google’s upcoming accelerator, called Trillium. The B200 posted a doubling of performance on some tests versus today’s workhorse Nvidia chip, the H100. And Trillium delivered nearly a four-fold boost over the chip Google tested in 2023.

The benchmark tests, called MLPerf v4.1, consist of six tasks: recommendation, the pre-training of the large language models (LLM) GPT-3 and BERT-large, the fine tuning of the Llama 2 70B large language model, object detection, graph node classification, and image generation.

Training GPT-3 is such a mammoth task that it’d be impractical to do the whole thing just to deliver a benchmark. Instead, the test is to train it to a point that experts have determined means it is likely to reach the goal if you kept going. For Llama 2 70B, the goal is not to train the LLM from scratch, but to take an already trained model and fine-tune it so it’s specialized in a particular expertise—in this case, government documents. Graph node classification is a type of machine learning used in fraud detection and drug discovery.

As what’s important in AI has evolved, mostly toward using generative AI, the set of tests has changed. This latest version of MLPerf marks a complete changeover in what’s being tested since the benchmark effort began. “At this point all of the original benchmarks have been phased out,” says David Kanter, who leads the benchmark effort at MLCommons. In the previous round it was taking mere seconds to perform some of the benchmarks.

Performance of the best machine learning systems on various benchmarks has outpaced what would be expected if gains were solely from Moore’s Law [blue line]. Solid line represent current benchmarks. Dashed lines represent benchmarks that have now been retired, because they are no longer industrially relevant.MLCommons

According to MLPerf’s calculations, AI training on the new suite of benchmarks is improving at about twice the rate one would expect from Moore’s Law. As the years have gone on, results have plateaued more quickly than they did at the start of MLPerf’s reign. Kanter attributes this mostly to the fact that companies have figured out how to do the benchmark tests on very large systems. Over time, Nvidia, Google, and others have developed software and network technology that allows for near linear scaling—doubling the processors cuts training time roughly in half.

First Nvidia Blackwell training results

This round marked the first training tests for Nvidia’s next GPU architecture, called Blackwell. For the GPT-3 training and LLM fine-tuning, the Blackwell (B200) roughly doubled the performance of the H100 on a per-GPU basis. The gains were a little less robust but still substantial for recommender systems and image generation—64 percent and 62 percent, respectively.

The Blackwell architecture, embodied in the Nvidia B200 GPU, continues an ongoing trend toward using less and less precise numbers to speed up AI. For certain parts of transformer neural networks such as ChatGPT, Llama2, and Stable Diffusion, the Nvidia H100 and H200 use 8-bit floating point numbers. The B200 brings that down to just 4 bits.

Google debuts 6th gen hardware

Google showed the first results for its 6^th generation of TPU, called Trillium—which it unveiled only last month—and a second round of results for its 5^th generation variant, the Cloud TPU v5p. In the 2023 edition, the search giant entered a different variant of the 5^th generation TPU, v5e, designed more for efficiency than performance. Versus the latter, Trillium delivers as much as a 3.8-fold performance boost on the GPT-3 training task.

But versus everyone’s arch-rival Nvidia, things weren’t as rosy. A system made up of 6,144 TPU v5ps reached the GPT-3 training checkpoint in 11.77 minutes, placing a distant second to an 11,616-Nvidia H100 system, which accomplished the task in about 3.44 minutes. That top TPU system was only about 25 seconds faster than an H100 computer half its size.

A Dell Technologies computer fine-tuned the Llama 2 70B large language model using about 75 cents worth of electricity.

In the closest head-to-head comparison between v5p and Trillium, with each system made up of 2048 TPUs, the upcoming Trillium shaved a solid 2 minutes off of the GPT-3 training time, nearly an 8 percent improvement on v5p’s 29.6 minutes. Another difference between the Trillium and v5p entries is that Trillium is paired with AMD Epyc CPUs instead of the v5p’s Intel Xeons.

Google also trained the image generator, Stable Diffusion, with the Cloud TPU v5p. At 2.6 billion parameters, Stable Diffusion is a light enough lift that MLPerf contestants are asked to train it to convergence instead of just to a checkpoint, as with GPT-3. A 1024 TPU system ranked second, finishing the job in 2 minutes 26 seconds, about a minute behind the same size system made up of Nvidia H100s.

Training power is still opaque

The steep energy cost of training neural networks has long been a source of concern. MLPerf is only beginning to measure this. Dell Technologies was the sole entrant in the energy category, with an eight-server system containing 64 Nvidia H100 GPUs and 16 Intel Xeon Platinum CPUs. The only measurement made was in the LLM fine-tuning task (Llama2 70B). The system consumed 16.4 megajoules during its 5-minute run, for an average power of 5.4 kilowatts. That means about 75 cents of electricity at the average cost in the United States.

While it doesn’t say much on its own, the result does potentially provide a ballpark for the power consumption of similar systems. Oracle, for example, reported a close performance result—4 minutes 45 seconds—using the same number and types of CPUs and GPUs.

Google is conducting a “test” that will omit results from EU-based news publishers. The company says the time-limited trial will only affect a small portion of users in nine EU countries and will help “assess how results from EU news publishers impact the search experience for our users and traffic to publishers.” But given the fragile state of the news media — and the company’s history of threatening to pull its services in the face of news-related regulations — it’s tempting to view it as the equivalent of a mob boss conducting a “little test” to see how the corner laundromat fares without its protection.

Google describes the experiment (via The Verge) as a “small, time-limited test” to omit EU results from search, Google News and the personalized Discover feed. It will only affect one percent of users in Belgium, Croatia, Denmark, France, Greece, Italy, the Netherlands, Poland and Spain. Those users will still see results from other websites, including non-EU news publications.

The company says news results will reappear as usual once the test concludes. (It didn’t list a specific timeframe.) Google stresses that the experiment won’t impact the publisher payments it makes under the European Copyright Directive (EUCD), under which the company has inked deals with over 4,000 EU publishers.

Google does have a history of using the potential withdrawal of its visibility as a negotiating stick in similar situations. In some cases, the tactic has helped it draw concessions.

Last year, Google pulled its news links from Canada in response to Bill C-18 (the Online News Act), which required tech companies to negotiate compensation with online publishers for linked content. After months of negotiations, Google said Canada had addressed its concerns and given it a path to an exemption. Canada said it granted one to Google last month, with the company agreeing to pay $100 million annually to news organizations.

In April of this year, Google briefly removed links to California news outlets in response to the proposed California Journalism Preservation Act (CJPA), which would require Google to pay news publishers in exchange for continuing to link to their websites. Although the bill’s fate is still up in the air, Google struck a deal with state lawmakers this summer, committing tens of millions of dollars to a fund supporting local news.

In 2021, the company threatened to remove its entire search engine from Australia in response to a then-proposed law requiring tech companies to share royalties with news publishers. The nation’s then-Prime Minister stood firm. “Let me be clear. Australia makes our rules for things you can do in Australia,” Scott Morrison said. After the bill was passed and enacted, Google struck deals with Australian media companies to license content.

Google says it hopes the data analysis tools it provides publishers will help them use the EU test to “understand traffic patterns.”

London’s National Gallery is celebrating 200 years with newly digitised artworks, curated stories and an interactive AI experience connecting art and music on Google Art…

An overview of Google’s work at APEC and the public-private partnerships expanding AI’s benefits to more people.

The startup has helped over 50,000 people in rural India in the past two years by offering them data annotation jobs

The Real McCoy of email verification is Brand Indicators for Message Identification (BIMI).

1. Your Google Pixel Phone's Newest Android 15 Beta Update Arrived  Droid Life
2. Google rolls out Android 15 QPR2 Beta 1 update  FoneArena
3. Google is preparing to bring back a beloved customization feature from Android 11  Android Authority
4. Android 15 QPR2: Do Not Disturb gets big upgrade with Modes  9to5Google
5. Android 15 QPR2 brings the Pixel 9's new Linux kernel to all Tensor-powered phones and tablets  Android Police

1. How this ‘pop-up’ on Windows laptops and PCs may again lead to Google vs Microsoft tussle  The Times of India
2. Microsoft Edge is trying to forcefully get your Chrome tabs again  The Verge
3. Microsoft Edge has a sneaky new pop-up that tries to copy your Google Chrome data  9to5Google
4. Microsoft is sneakily trying to import tabs from other browsers into Edge  Ghacks
5. Microsoft is, once again, trying to force users into using Edge  MSN

"Our technology goes beyond the sensor components of a traditional eye-tracking system, actually giving the user the power to control their device through eye-interaction," digitallook reported, quoting the company.

WhatsApp will launch video calling service on its app, a move that will help the popular messaging platform compete with the likes of Skype, Apple's FaceTime and the recently launched Google Duo.

Tech giant Google has rolled out a new scanning app — Google PhotoScan — that allows users to scan and digitise physical photographs.

Google announced the expansion of its artificial intelligence (AI) flood forecasting system on Monday. With this expansion, the Mountain View-based tech giant will now cover 100 countries and offer...