Artificial Intelligence - Page 20

Through the power of artificial intelligence, two people who cannot understand each other's languages can now communicate fluidly. Well, almost.

Meta has rolled out a new update to its Ray-Ban smart glasses that adds a live translator feature where a wearer can simply say, "Hey Meta, start live translation," and the feature will be enabled. Once enabled, the conversation can begin, and the wearer of the glasses will hear the translated conversation in their preferred language. As always, there are some caveats. Currently, only four languages are supported: Spanish, English, French, and Italian.

Another caveat is that for live translation to work in the ideal setting, both people should be wearing a pair of Meta's Ray-Ban smart glasses. However, the feature does work if one person is wearing the glasses and the other is using the Meta app to receive the translation. What's nice is that users can download each language, making the translation feature available offline. This means you won't need an active internet connection to take advantage of live translation.

The AI chatbot landscape is extremely competitive with big tech companies fighting to provide a platform with the best user experience that combines fluid information interactions, a slick user-interface, and additional features such as image generation, etc.

While AI is relatively new, usage figures such as those obtained through Google's antitrust lawsuit showcase just how much the technology has been adopted by the masses. For example, Google's Gemini had just 9 million daily active users in October 2024, but according to recent data, that number has jumped to 35 million daily active users as of last month.

As for monthly active users, the same data indicates Google's Gemini pulls in as many as 350 million monthly active users, which is quite a lot, but not as many as competing AI tools such as OpenAI's ChatGPT, which has roughly 600 million monthly active users as of March. ChatGPT's astronomical numbers put it at the table with the likes of Meta AI, which company CEO Mark Zuckerberg said in September last year had obtained as many as 500 million monthly users.

The JEDEC Solid State Technology Association has published its highly anticipated High Bandwidth Memory (HBM) DRAM standard: HBM4.

HBM4 has been designed as an "evolutionary" step beyond the previous HBM3 standard, with the new "JESD270-4 HBM4" to further enhance data processing rates while maintaining essential features like higher bandwidth, power efficiency, and increased capacity per die and/or stack.

The advancements coming with HBM4 are vital for applications that requires efficient handling of large datasets and complex calculations, including generative artificial intelligence (AI), high-performance computing (HPC), high-end graphics cards, and servers. HBM4 includes multiple improvements over the HBM3 standard, including:

The US government has just banned NVIDIA and its H20 AI GPUs from entering China, with the company taking a $5.5 billion hit in Q1 2025 earnings.

NVIDIA has warehouses full of its H20 AI GPU and were preparing mass shipments to China, but President Trump's latest US export restrictions are now making waste to those $5.5 billion worth of AI chips. H20 was designed to offer a precise amount of AI workload performance that skirted around US export controls, and now those chips are sitting around doing nothing and NVIDIA is taking a $5.5B hit this quarter.

The kicker is that NVIDIA custom-tuned the H20 AI GPU after the increasing US export regulations, so this isn't just another card but a tailor-made chip for Chinese companies like Alibaba and TikTok parent company ByteDance, to enjoy enough AI power without hitting those restrictions.

Huawei just unveiled its new Ascend 920 AI chip, the successor to its China-exclusive Ascend 910C -- which is being used in a new "CloudMatrix" AI server -- within hours of new US export restrictions on NVIDIA's H20 AI GPU entering the country.

In a new report from DigiTimes, we're learning that Huawei's new Ascend 920 is a direct competitor to the NVIDIA H20 AI GPU, with a focus on the supply chain being entirely in-house with this new chip. Huawei's new Ascend 920 will be made on SMIC's 6nm process, which is an upgrade over the Ascend 910C which was fabbed on 7nm at SMIC.

We are expecting to see 900 TFLOPs of compute performance, 4TB/sec of memory bandwidth through its HBM, and a chiplet design that is similar to what we've seen with the Ascend 910C. This should all lead into around 30-40% more performance from the new Ascend 920 over 910C, meaning the Ascend 920 is a better AI chip than the now even-more-restricted H20 AI GPU from NVIDIA.

China is now using Huawei's Ascend 910C AI chips in a huge cluster called "CloudMatrix" and it reportedly has more performance than NVIDIA's leading GB200 NVL72 AI server, but CloudMatrix uses far more power.

In a new report from SemiAnalysis, we're learning about Huawei's rack scale architecture and just how the Ascend 910C will power China's new CloudMatrix 384 AI cluster, which has performance that rivals NVIDIA's most powerful AI server in the GB200 NVL72 AI server rack.

Huawei's new CloudMatrix 384 "CM384" AI cluster is powered by 384 Huawei Ascend 910C AI chips connected in an "all-to-all topology" configuration. Huawei is outweighing the architectural flaws of its AI chips by using 5x more of them than NVIDIA uses with GB200 inside of NVL72 servers. This is why the company doesn't care about the costs, performance inefficiency, scalability ratios, and more.

Once again, the more hidden costs of the AI revolution are under the spotlight, and in this case, specifically it's the topic of how much simply being polite to ChatGPT might run to.

What's being talked about here is, for example, saying thank you to the chatbot, an extra 'query' that ChatGPT processes and responds to, and how much that actually costs in terms of the electricity used therein.

That question was posed by a denizen of X, Tomie, and Sam Altman, the CEO of OpenAI, actually responded to it directly as you can see above (hat tip to TechRadar for picking up on this).

NVIDIA has just announced it will be making AI supercomputers in the US, with the new American-made supercomputers to feature NVIDIA Blackwell AI GPUs.

The company announced the news on its official website, saying it has commissioned more than a million squire feet of manufacturing space to both build and test NVIDIA Blackwell AI GPUs in Arizona, and AI supercomputers in Texas. NVIDIA Blackwell AI chips have started production at TSMC's semiconductor facilities in Phoenix, Arizona.

Jensen Huang, founder and CEO of NVIDIA, said: "The engines of the world's AI infrastructure are being built in the United States for the first time. Adding American manufacturing helps us better meet the incredible and growing demand for AI chips and supercomputers, strengthens our supply chain and boosts our resiliency".

Alphabet has committed to spending $75 billion on building out its data center capacity, even with all of the hoopla surrounding US tariffs.

Google CEO Sundar Pichai said that the $75B investment would buy the chips and build the servers needed to increase Alphabet (Google) and its core offerings, including search, as well as continued support in the development of AI services including its Gemini AI model. Pichai said: "The opportunity with AI is as big as it gets" as a surprise appearance was made at Google's annual conference for its cloud computing unit.

When asked about the potential for US tariffs to increase the cost of building out its data centers, Google Cloud infrastructure unit vice president and general manager, Sachin Gupta, said that the cost of importing hardware might increase but customer demand continued to necessitate the increased investment, he told Reuters: "We're all processing what's happening with tariffs".

China has released another video mocking overweight Americans working in factories amid the ongoing trade war between the US and China.

The second AI-generated video circulated on China's social media and has now made its way over to Western social media platforms, where it has garnered millions of views. The first video, titled "Make America Great Again," was a clear nod to President Trump's campaign slogan and depicted overweight people working in factories building clothes and smartphones.

The timing of the first AI-generated video came only days after President Trump announced heavy tariffs on China, and now a second video has begun circulating showing similar overweight Americans working in factories for prominent American companies such as Nike, Apple, NVIDIA, and Tesla. The video shows Americans building Nike shoes and eating fast food with a play on Trump's slogan on the wall, "Make America Strong Again". The video then cuts to Americans working on Tesla vehicles, smartphones, and then NVIDIA chips.

The US government has just reversed its plans to restrict exports of NVIDIA's H20 AI GPU to China, after CEO Jensen Huang had dinner at Mar-a-Lago with President Trump and his administration.

In a new report from NPR that cites two sources, the newly-tweaked US export controls on the H20 AI GPU had been in the works for a few months now, and that the change in plans kicked off after NVIDIA promised the Trump administration that it would be investing around $200 billion into new US investments in AI data centers.

NVIDIA CEO Jensen Huang recently attended a dinner at Mar-a-Lago, with the Trump administration reversing its restrictions on the H20 AI GPU entering China. Not only that, but the Trump administration also had planned to roll out additional restrictions, but they too have been put on hold according to "two sources with knowledge of the plan who were not authorized to speak publicly" adds NPR.

Samsung Ballie is described as a "home AI companion robot" that will arrive in the United States and Korea this summer. The company confirmed this week that Google Cloud's generative AI technology will power the AI side of the cute robot. This will allow owners to communicate with Ballie using natural conversational language for all their companion robot needs.

Suppose you're wondering what Ballie is about outside of rolling around and looking like a cute robot from a Star Wars or Pixar film. In that case, it's essentially a range of generative AI tools and functionality paired with an inbuilt projector. Plus, it can do things like greet guests, adjust smart home things like lighting and cooling, and set reminders for appointments and things. A robot butler that doesn't serve food or do any cleaning.

Samsung's description of how Ballie will use cloud-based AI is vague - though it provides some examples. Yongjae Kim, Executive Vice President of the Visual Display Business at Samsung Electronics, states that Ballie "moves with users, anticipates their needs, and interacts in more dynamic and meaningful ways than ever before," thanks to generative AI.

Amid President Trump's recently announced trade tariffs, an artificial intelligence-generated video of Americans working in factories has gone viral on Chinese social media. The video has now made its way to Western social media platforms, where it has garnered quite a lot of attention.

The video has been copied and reshared on various social media platforms and shows a satirical take on what American people look like working in factories. The video was titled "Make America Great Again," a clear nod to President Trump's campaign slogan. It depicts overweight people working in what appears to be sweatshops, making various items of clothing and smartphones. The AI-generated people in the video also look miserable.

The timing of the AI-generated video mocking Americans comes only days after President Trump announced heavy tariffs on China, along with numerous other countries. Trump's goal of the tariffs is to make better deals with its trade partners and ultimately bring manufacturing back to the United States. It appears China is attempting to show Americans what the future holds for them with this satirical AI generated video.

You aren't in Kansas anymore, you're in Las Vegas - or you might be in August, watching an AI-revamped version of The Wizard of Oz that'll likely blow you away hurricane-style.

The iconic movie has been jazzed up with generative AI for screening on the 16K monster LED canvas of the Las Vegas Sphere, and we're betting it'll make quite an impact when the film starts showing on August 28 (as Neowin spotted, and FOX5 Las Vegas reports above).

Google tells us:

Microsoft is celebrating its 50th anniversary, a significant milestone for a tech company still at the forefront of the industry. Like other software and hardware giants, Microsoft is pivoting toward AI for its cloud-based computing business, Windows, and devices like Xbox gaming consoles.

As part of its Windows strategy, all of the software's AI features fall under the Copilot banner. Copilot+ PCs are mobile devices that can use AI tools locally using powerful NPU hardware. From image generation to chatbots, summarizing documents, and transcribing audio live, Microsoft's AI endgame for Windows is to create a Copilot companion for everyone - one that will get to know you and with whom Windows users will form a relationship.

"We're really trying to land this idea that everybody is going to have their own personalized AI companion," said Mustafa Suleyman in an interview with The Associated Press. Of course, there's no timeline for this, but AI is quickly becoming a part of Microsoft's present and near future.

Chinese AI companies have reportedly placed orders for a hefty $16 billion worth of NVIDIA's new H20 AI GPUs, according to a new report from The Information.

In a world where US sanctions are meant to be stopping China from getting their hands on some of the most advanced AI chips made, Chinese tech companies like Baidu and TikTok parent company ByteDance, are speeding up their data center building and model training to keep up with the West.

This shouldn't come as a surprise at all, with recent stories -- in the links below -- had us reporting that NVIDIA could experience more Chinese GPU export restrictions from the Trump administration, as well as Chinese AI firm DeepSeek using 50,000 NVIDIA H100 AI GPUs, even with US restrictions in place. In 2024, we reported that the US government was asking NVIDIA how its advanced AI chips were getting into the hands of Chinese companies, which skirted US sanctions.

SK hynix is the leader in AI memory with its next-gen HBM4 and even next-gen-er HBM4E memory is being promised for later this year, with mass production of its new 12-layer HBM4 promised at the same time.

Choi Jun-yong, Vice President of the HBM (High Bandwidth Memory) Business Planning at SK hynix, recently said: "We will further solidify our HBM leadership by not only mass producing (6th generation HBM) HBM4 12-layer this year, but also timely supplying (7th generation) HBM4E".

In an interview with VP Choi that day in the SK hynix newsroom, the executive said: "In parallel with the development of new HBM, we will provide optimal solutions to various customer needs through custom HBM that meets the specialized needs of customers".

In an unconventional deployment of AI tools, a defendant in the New York Courts has attempted to use an artificial intelligence avatar to represent himself in a legal dispute.

The defendant in question, 74-year-old Jerome Dewald, appeared in the New York Supreme Court on March 26 regarding an employment matter. During the hearing, Dewald attempted to submit footage of an AI-generated avatar, which appeared in front of the judges and uttered a few words before being quickly shut down by the judge.

SK hynix has started its HBM expansion project, with its "M10F" semiconductor fab re-purposing now in operation, meaning it can make more HBM memory chips for the AI industry later this year.

In a new report from ETnews, we're learning that SK hynix has completed the re-purposing of its semiconductor fab known as "M10F" located in Icheon, South Korea. This semiconductor fab has had its existing DRAM packaging lines converted to an HBM packaging line.

An industry insider said: "They introduced and replaced some packaging equipment and modified the required materials to suit the new packing specifications. They obtained the necessary permits from the local fire department at the end of last month and have commenced mass production".

Vietnamese shop Nguyencongpc have been building custom AI servers based on NVIDIA GeForce RTX 5090 graphics cards, with 7 x RTX 5090 cards used with 28 power cables in total, and 4000W of power consumption for the custom AI servers.

If you thought cryptocurrency mining back a few years ago caused issues with GPU shortages, it seems we have to fear custom AI servers using consumer-focused GeForce RTX 50 series gaming GPUs creating shortages for PC gamers. At the time, NVIDIA gimped crypto mining performance, but crypto miners continued to find ways around that.

However, NVIDIA's new GeForce RTX 5090 graphics card ships with a huge 32GB of ultra-fast GDDR7 memory, with large pools of memory favored by AI workloads, so a Vietnamese shop has used 7 of them in a custom AI server for various AI workloads, and 7 x 32GB of RAM is a big deal for AI users.