Artificial Intelligence News - Page 1

Foxconn has won exclusivity on manufacturing NVIDIA's new NVLink switches for its next-gen GB200 AI servers, a key component of GB200, which is known as the "magic weapon for improving computing power".

In a new report from UDN, we're learning that the order volume in 7x that of the GB200 AI server cabinets, where it's not just a new order, but the gross profit margin is also "much higher" than that of server assembly. It will become a powerful tool for Foxconn's gross profits and supplement its profits.

Foxconn has never commented on its orders or customer dynamics, but the industry has highlighted that the exclusive technology behind NVIDIA's new NVLink consists of two parts. First is the bridge technology, which connects the CPU and the AI chips together, while the other is the switch technology, which is the key to the interconnection between the GPUs.

NVIDIA couldn't be the company it is today without Taiwan, and CEO Jensen Huang knows that, and respects it, as well as the country itself.

During a recent interview, NVIDIA CEO Jensen Huang recently said that the company's $100 billion revenue is mostly manufactured in Taiwan, noting that AI has benefited Taiwan and that Taiwanese companies can thrive in the upcoming AI era.

In the full interview, Jensen said: said: "For over 25-years, we've been [working with companies in Taiwan]...Another reason is technology excellence. The ecosystem is unparalleled. Every aspect of building chips and building computers are located here. All the way down to the most minute, complex connectors that connect very high-speed cables are manufactured here".

NVIDIA announced the "most powerful AI chip on earth" by introducing its Blackwell-based GB200 AI superchip. Taiwan companies Foxconn and Quanta won major contracts for GB200 AI servers.

According to a new report by UDN, Foxconn will be handling 40% and Quanta will take 30% of GB200 AI servers, with shipments expected to begin in Q3 2024, driving a "big jump in overall revenue performance". NVIDIA's new DGX GB200 will also begin mass production in the second half of this year, with estimates of around 40,000 units in 2025.

Foxconn and Quanta will be the main suppliers of NVIDIA's upcoming NVL72 and NVL36 AI server cabinets, with UDN reporting that industry experts saying Foxconn has received orders for three categories of "DGX GB200" system cabinets (DGX NVL72, NLV32, and HGX B200). Foxconn said that GB200 series products will be shipped in rack form, with rough order estimates as high as 50,000 containers, with total revenue contribution to hit 2.5 trillion to 3 trillion yuan (around $344 billion to $413 billion USD or so).

NVIDIA CEO Jensen Huang made quite the famous trip to Taiwan this year, not just for Computex 2024, but to meet important tech partners in the country.

During his time, Jensen was interviewed by local media where he said that NVIDIA's annual growth exceeds $100 billion with various GPUs and AI GPUs, and that most of them are made in Taiwan. AI applications have benefitted Taiwan, and Taiwanese companies can thrive in the AI wave, reports CNA.

Jensen said that the AI era is a new beginning, something that NVIDIA has been preparing for a very long time now. Taiwan's technology industry is fully committed to the future of AI, with Jensen saying during a recent interview: "I am very happy to see this phenomenon".

Researchers have used machine learning techniques, the fundamental technology powering artificial intelligence, to analyze hundreds of wild elephants that were captured between 1986 and 2022.

The team published their findings in the peer-reviewed journal Nature Ecology and Evolution, which states that researchers looked at how each elephant "rumbles." These rumblings are believed to be the primary way elephants communicate with each other, and the study has broken the rumbles into three categories: greeting, caregiving, and contact calls.

While there are other prominent rumble categories, such as "let's go," the team found the aforementioned three types of rumble are most likely tied to names between elephants. Notably, the study found through the power of AI analysis that elephants appear to be attaching names to these rumbles, which is very different from imitation communication that is used by dolphins or parrots to communicate.

One of the most prolific tests a machine can undertake is called the Turing Test, created by Alan Turing in 1950 that's purpose was to test if a machine is capable of displaying a level of intelligence that can convince, or fool a human they are engaging with another human.

Researchers have put OpenAI's GPT-4 up against the Turing Test, and according to a new study the impressive large language model passed the test. The not-yet-peer-reviewed-study asked 500 people to engage with four different respondents. One of those respondents was a human, another was a 1960s-era AI called ELIZA, another was OpenAI's less sophisticated GPT-3.5, and finally GPT-4.

The paper states each conversation lasted approximately five minutes, and according to the study's results, humans found GPT-4 to be a human 54% of the time. These results lead researchers to claim GPT-4 has already passed the Turing Test. However, the study also indicated study participants believed the human respondent was a human 67% of the time, while ELIZA scored just 22%, and GPT-3.5 scored 50%.

After sustained criticism and widespread disdain, Microsoft will not roll out its controversial AI-powered 'screenshot everything you do on Windows 11, including capturing sensitive data' Recall feature as part of the Copilot+ PC launch. Instead, a revamped Recall feature will be available in the coming weeks as part of the Windows Insider Program (WIP).

Copilot+ PCs, which are basically laptops with CPUs equipped with at least a 40 TOPS NPU, are set to hit retail shelves this week. The Arm-based Snapdragon X Elite processor exclusively powers the first run, which, without Recall, will be limited to Microsoft's AI tools for photo and video editing, live captions, and Cocreator image generation. Plus, apps from third parties that require 40 TOPS of AI Performance.

This is for the best, as Recall has been under fire since its announcement on May 20. The controversial snapshot tool - designed to make it easier to search through your local PC usage to find documents, images, or anything else - was quickly discovered to be lacking security features. It would record things like passwords, banking information, and more.

Samsung is preparing to launch 3D packaging services for HBM within the next 12 months. The new technology will be introduced for HBM4, ready for the next-gen AI GPUs of the future, which will be released in 2025.

The company held its Samsung Foundry Forum 2024 in San Jose, California, and teased its new 3D packaging technology for HBM chips in a public event, with current-gen HBM memory chips packaged mostly with 2.5D technology.

NVIDIA is about to introduce its new Blackwell B100, B200, and GB200 AI chips which will use the latest HBM3E memory, but its next-gen Rubin GPU architecture was teased just weeks ago, with the next-gen Rubin R100 AI GPU to feature next-generation HBM4 memory. Rubin will enter mass production in Q4 2025, using TSMC's newer N3 node and new CoWoS-L advanced packaging.

McDonald's has just told its franchise operators on Thursday that it's removing AI order-taking technology from over 100 drive-thrus after a test period ends with an AI partnership with IBM.

A McDonald's spokesperson confirmed the news with Business Insider, saying that customers have been using the AI-powered drive-thru technology since 2021, after McDonald's entered a global partnership with IBM. Under the partnership, IBM acquired McD Tech Labs, something McDonald's created after taking control of AI speec company Apprente back in 2019.

The two companies deployed and developed the AI technology during the test period to "determined if an automated voice ordering solution could simplify operations for crew and create a faster, improved experience for our fans". But, it simply didn't work. McDonald's CEO Chris Kempczinski told CNBC in June 2021 that the voice recognition technology was accurate about 85% of the time, with human staff assisting in around 1-in-5 orders.

Artificial intelligence (AI) has joined the lexicon for the various now-viral applications of machine learning software. Now, Apple has arrived, unsurprisingly, fashionably late with its own implementation of AI that's slated to hit hundreds of millions of users.

Apple announced in iOS 18 that it will be introducing Apple Intelligence to its new line-up of devices while also outlining what older generation iPhones will get the new features. Simply put, to use Apple Intelligence, you will need an iPhone 15 Pro or iPhone 15 Pro Max, which are both under a year old at the current time of writing. As for Mac users, Apple states Apple Intelligence will work on Mac's using Apple Silicon, which is models 2020 and up.

So, why this gatekeeping of new AI features from older generations? This problem can be viewed in multiple ways. A common take would be to suggest Apple simply wants to increase its bottom line by enticing consumers with new features that are locked to the latest generation devices to sell more units. While that is certainly happening, there is also a very real hardware limitation for AI features, at least according to Apple and, separately, Microsoft with its Copilot+ PCs.