Artificial Intelligence - Page 34

Intel has officially launched its Gaudi 3 AI accelerator, with the new AI chip coming in slower than NVIDIA's dominant H100 and its new HBM3E-fueled H200 AI GPUs, meaning Intel is aiming its Gaudi 3 by pushing that it's cheaper, and has a lower total cost of ownership (TCO).

Inside, the new Intel Gaudi 3 AI accelerator features two chiplets with 64 tensor processor cores (TPCs, 256x256 MAC structure with FP32 accumulators), eight matrix multiplication engines (MMEs, 256-bit wide vector processor), and 96MB of on-die SRAM cache with a 19.2TB/s bandwidth.

Gaudi 3 also features 24 x 200GbE networking interfaces and 14 media engines, with the media engines capable of handling H.265, H.264, and VP9 to support vision processing. Intel's new Gaudi 3 AI accelerator features 128GB of HBM2E memory with up to 3.67TB/sec of memory bandwidth.

Apple has released its new iPhone 16 family of handsets, with the iPhone 16, iPhone 16 Pro, and iPhone 16 Pro Max available across the planet right now... but... where are all those AI features that hte company promised with Apple Intelligence?

The on-device AI functionality that I've even seen advertised on TV, where they've got Bella Ramsey from The Last of Us looking literally useless without her iPhone and Apple Intelligence. The Hollywood actor... acts... as if she hasn't read the script given to her, so she gets a summary of it from Apple Intelligence, like that's meant to make you buy a multi-thousand-dollar new iPhone 16.

Anyway, now there's analyst Mark Gurman talking about the useless AI features in Apple's new iPhone 16 handsets and that some of them won't be fully-baked until 2025, and some of them are radically behind competing AI on the market. Gurman said that Apple has been a "little disingenuous with its marketing" as Apple claimed the iPhone 16 was the first model "built from the ground up for Apple Intelligence".

A modder has created what he calls "The Ultimate Cheating Device" with a regular TI-84 calculator hacked to run ChatGPT, which is perfect for students who want to ninja an AI-powered calculator into the class.

YouTuber ChromaLock uploaded the video that is embedded above, with some hardware modifications applied and the use of some open-source software modified for the TI-84 that he made, allowing the calculator to run ChatGPT. The modder has uploaded the software required onto GitHub, under the TI-32 repository, which is described as "a mod for the TI-84 Plus Silver Edition and TI-84 Plus C Silver Edition calculators to give them Internet access and add other features, like test mode breakout and camera support".

A microcontroller small enough to fit inside of the TI-84 shell with all of its components is the hardest step, after that the software modification is applied. All of the required TI-84 software features require the use of a link port to connect to bulky external devices, so if students were doing that it would be too obvious... but hardware mods + ChatGPT installed? Game changer.

Microsoft has just signed a 20-year deal with Constellation Energy, the owner of the Three Mile Island nuclear plant in Pennsylvania, and once revived, will provide clean energy to Microsoft and its AI data center and cloud computing needs for 20 years.

Three Mile Island has two nuclear reactors: the first with a capacity of 906 MW was shut down back in 1979 after the "Three Mile Island nuclear incident" that I'm sure you've heard of (if not, you should read into it). The other, has a capacity of 819 MW and was closed in 2019 over economic issues, but will now be restarted thanks to the deal with Microsoft.

Constellation Energy will invest $1.6 billion in restarting the Three Mile Island nuclear reactor, a process that had been in development since early 2023 when the company looked at the feasibility of bringing the nuclear reactor back online. After it decided to go ahead and restart the nuclear reactor, it began talking with potential buyers... with Microsoft showing immediate interest, and now the deal is inked (and for 20 years).

TikTok parent company ByteDance is developing not one but two new AI GPUs that are reportedly being made on TSMC 5nm process node, and will enter mass production in 2026.

The information is coming from The Information, where according to their sources ByteDance will reduce its reliance on NVIDIA for its AI hardware, all the while staying in the lines of US export regulations. ByteDance's new AI GPUs are in the design phase, with one of them being for AI training and the other for AI inference.

ByteDance's new AI GPUs are said to be made on TSMC 4N/5N process nodes, which is similar to the 4NP process node that TSMC uses to make NVIDIA's new Blackwell AI GPUs. The TikTok parent company reportedly spent over $2 billion buying over 200,000+ NVIDIA H20 AI GPUs (which cost around $10,000 per H20 AI GPU) this year alone, with many of the AI GPUs not yet delivered by NVIDIA.

At the recent Goldman Sachs Communacopia and Technology Conference, NVIDIA CEO Jensen Huang was asked about exciting use cases for AI and responded with a nod to DLSS and other RTX technologies. "In our company, we use it for computer graphics," Jensen replied. "We can't do computer graphics anymore without artificial intelligence."

NVIDIA DLSS, or Deep Learning Super Sampling, is a crucial part of modern PC gaming on GeForce graphics cards, where the AI-powered upscaler boosts performance by generating new pixels that hit a target resolution and frame rate. With the arrival of the GeForce RTX 40 Series, NVIDIA expanded this to add Frame Generation, where AI generates entire frames. Throw in the impressive AI-powered Ray Reconstruction, and it's the reason why games like Black Myth: Wukong is playable with all settings maxed out.

"We compute one pixel, we infer the other 32. I mean, it's incredible," Jensen continues. "And so we hallucinate, if you will, the other 32, and it looks temporally stable, it looks photorealistic, and the image quality is incredible, the performance is incredible, the amount of energy we save - computing one pixel takes a lot of energy. That's computation."

Oracle CTO Larry Ellison recently had dinner with SpaceX and Tesla CEO Elon Musk, as well as NVIDIA CEO Jensen Huang, with Ellison saying both he and Elon were "begging" Jensen for GPUs. Check it out:

Ellison said: "I went to dinner with Elon Musk, Jensen Huang, and I would describe the dinner as Oracle and me and Elon begging Jensen for GPUs. Please take our money... no no, take more of it. You're not taking eough. We need you to take more of our money, please".

It was recently revealed that Oracle would be spending over $100 billion on 2000+ data centers in the future, with 130,000+ of NVIDIA's new Blackwell AI GPUs powering its new AI supercluster, with the company talking about requiring 3 nuclear power plants just to power the supercluster.

NVIDIA CEO Jensen Huang has spoken to CNBC after meeting with the Biden administration and tech executives at the White House about the future of AI development and what it will take to produce it.

Huang joined other major tech executives at the White House who are also assisting in the development of AI, with reports confirming the presence of OpenAI's CEO Sam Altman, Anthropic CEO Dario Amodei, Google President Ruth Porat, Amazon's cloud chief Matt Harman, and Microsoft President Brad Smith. On the other side of the fence were government officials from various agencies, such as Commerce Secretary Gina Raimondo, National Security Advisor Jake Sullivan, and Energy Secretary Gennifer Granholm.

What was the meeting about? Huang explained that the conference focused on how the US government could assist in the development of data centers across the US, particularly in the form of implementing initiatives that ultimately would lead the US to maintain its lead in the global AI race. Companies require a significant amount of energy to power these massive data centers, which was also a topic of the meeting as the Energy Department will help data center owners find clean and reliable power sources along with new resources in the form of loans, grants, and tax credits.

Meta Platforms is reportedly putting the "final touches" on one of its new AI supercomputers, powered with over 100,000+ NVIDIA H100 AI GPUs.

In a new report from The Information, the new AI supercomputer from Meta will feature 100,000+ NVIDIA H100 AI GPUs and will be located "somewhere" in the US. The new supercomputing cluster will train the next version of Meta's Llama model: Llama 4. Meta's new 100K+ NVIDIA H100 AI supercomputer cluster will be fully completed by October or November, says The Information.

Meta's new AI supercomputer with its 100,000+ NVIDIA H100 AI GPUs reportedly costed over $2 billion for the H100 AI GPU chips alone, which means Mark Zuckerberg is signing some fat cheques to NVIDIA. Speaking of which, NVIDIA CEO Jensen Huang recently spoke with Meta CEO Mark Zuckerberg, where Jensen said Meta now has 600,000+ NVIDIA H100 AI GPUs to which Zuck replied saying that Meta were "good customers for NVIDIA".

NVIDIA CEO Jensen Huang has said that there is a mad scramble for companies to get their hands on the limited supply of Blackwell AI GPUs, and that is frustrating some companies, while raising tensions with others.

Jensen Huang was recently speaking to the audience at the Goldman Sachs Group Inc. technology conference in San Francisco, where he said: "The demand on it is so great, and everyone wants to be first and everyone wants to be most. We probably have more emotional customers today. Deservedly so. It's tense. We're trying to do the best we can".

He continued, adding that TSMC's "agility and their capability to respond to our needs is just incredible" said Jensen, adding: "And so we use them because they're great, but if necessary, of course, we can always bring up others".