Artificial Intelligence - Page 22

NVIDIA has officially unveiled its beefed-up GB300 "Blackwell Ultra" NVL72 AI server, with its new GB300 AI GPU featuring 50% more performance over GB200, and a larger pool of 288GB of HBM3E memory.

NVIDIA's new GB300 "Blackwell Ultra" AI GPUs will continue the AI domination led by GB200, with memory capacity increased through using new 12-Hi HBM3E memory stacks, and even more compute power for AI workloads. NVIDIA is also coupling GB300 with the latest Spectrum Ultra X800 Ethernet switches (512-Radix).

The new NVIDIA GB300 NVL72 rack-scale solution has 1.5x more performance than an NVIDIA GB200 NVL72 AI server, as well as increasing Blackwell revenue opportunities by 50x for AI factories compared to AI servers built on the Hopper GPU architecture.

NVIDIA's GTC conference has officially kicked off, and Phison, one of the world's leading companies in controllers for NAND flash memory chips, has unveiled what it's describing as an affordable path to supporting a 1 trillion parameter AI model.

The race to the first 1 trillion parameter AI model is nearing completion, at least according to Phison, which predicts we will see the first 1 trillion parameter AI model before 2026. Phison has briefly outlined the progression of AI models, with a 69 billion parameter model unveiled in 2023 with Llama 2, then a big jump to the 405 billion parameter Llama 3.1 model in 2024, and then the 671 billion parameter DeepSeek R3 model in 2025. Keeping on this same trajectory, Phison expects the world's first 1 trillion parameter model will be unveiled before the end of 2025.

With the increase in the size of these AI models, the hardware needed to train and support them will also need to be increased, which is why Phison is looking at ways to reduce operational costs, and the company has a solution. Phison expects the operational costs of a 1 trillion parameter model will be approximately $3 million in raw GPU power, but the company has outlined a plan to alleviate the cost down to just $100,000. How? A combination of SSDs, Phison's aiDAPTIV+ software, and an NVIDIA GH200 Superchip.

The world's first Large Language Model (LLM) training laptop is being showcased at GTC 2025, and it's a creation between Phison and Maingear.

Phison Electronics, a world leader in NAND flash technology, has teamed up with Maingear to bring its aiDAPTIV+ software stack to Maingear's ML-series laptop, creating the first AI laptop PC capable of Large Language Model Operations. The new laptop, which is being demoed at GTC 2025 with an RTX 4090 laptop GPU, will feature an NVIDIA RTX 5090 laptop GPU when it is released for sale. As for its capabilities, Phison has announced it will be able to train and inference LLMs that are up to 8 billion parameters, while also having support for fine-tuning, which enables users to inject their own data into the AI model.

Why does this laptop exist? Phison is setting out to reduce the cost of running AI models by offloading the processing, which is currently on high-cost GPU memory, to low-cost NAND flash storage, and by doing this the operational cost of the AI model shrinks, enabling larger models to be trained on cheaper devices. By utilizing the power of the company's aiDAPTIV+ software, users will now be able to customize an AI model with their own data, all within a device that can fit in your backpack.

Phison is expanding its aiDAPTIV+ software to the NVIDIA Jetson platform, enabling enhanced generative AI inference at the edge of robotics development.

Phison has created a specialized SSD for the NVIDIA Jetson platform and has integrated its aiDAPTIVE+ software to enable Fine-Tune training, which allows users to take an AI model and add their own data into the model. The idea behind this integration is to reduce the performance cost of running the AI by offloading some of the processing from the GPU VRAM over to the SSD.

Phison explained to me that with these upgrades to the Jetson platform they are able to provide up to 14x improvement on Time to First Token (TTFT), which is the time it takes for an AI to provide a response to a question. Moreover, these changes also increased the token length by 8x, which is the length of the response generated by the AI. Ultimately, Phison's improvements made to Jetson have unlocked the platform in a way that enables faster responses with a greater length, enhancing the quality of the AI chat.

Bots in competitive or cooperative multiplayer games have been a problem for years, with a new study from Echelon Insights stating that 59% of gamers have encountered third-party bots in games. And with that, 71% agree that bots make it less fun to play certain games, with 77% agreeing that it's essential to verify that you are a human before jumping into the action. And these bots are getting increasingly sophisticated in the age of AI gaming.

'Razer ID verified by World ID' is one solution to this problem, and it's set to be part of the upcoming multiplayer game TOKYO BEAST. The game is set in Tokyo in the year 2124, a time when humans and "advanced autonomous androids" coexist. It's a competitive title that includes collectible unique NFTs alongside in-game purchases via Razer.

'Razer ID verified by World ID' is powered by secure, single sign-on (SSO) technology designed to verify "real human gamers" to create a more authentic and immersive gaming experience. World ID, a company co-founded by Sam Altman and Alex Blania, is all about confirming someone's unique online human identity with a focus on privacy. It can tell that someone is human without knowing who they are.

With new AI chatbots released every week and new models seemingly by the day, it's an uncertain time to be a computer programmer - or consider entering the field. AI-assisted coding tools, from GitHub Copilot to ChatGPT-powered development, are capable of generating full applications with minimal human input. As automation advances, some question whether traditional programming skills will remain relevant.

Andrew Ng, the founder of DeepLearning AI and co-founder of Google Brain, Coursera, and an active Stanford professor, takes a different view. In a recent blog post, he argued that this is actually the best time to learn to code, dismissing claims that AI will make programming obsolete.

With the emergence of AI tools affecting almost every industry, few are more affected than the field of software development. 'Vibe coding' is one of the trends to emerge from that shift, and we'll go into what it is, and how it's affecting these industries.

To start at the beginning - in February 2025, OpenAI co-founder and former Tesla AI lead Andrej Karpathy coined the term on X. In a widely publicized post, he describes it as a method that essentially involves letting go of the intricacies of generated code - and essentially 'fully giving in' to the vibes. In other words, embracing an intuitive, instinctual approach to writing programs - and forgetting about understanding the code itself.

In practice, what this looks like is a conversational approach to software development. As Karpathy describes, it's giving the chatbot simple, almost dumb, 'vibe driven' prompts - and letting the AI generate beyond your comprehension. For example, a user might prompt an AI coding assistant with something as vague as "Make a flight simulator where you can buy an F-16" and let the AI generate the bulk of the game's code, or "Fix the UI so it looks cleaner, but keep the aesthetic fun." Instead of making manual adjustments, vibe coders just keep iterating with broad prompts like "Make the sidebar padding feel right" or "Tweak the animations so they pop more." The AI takes over, handling all the changes while the coder simply approves or rejects based on vibes alone.

NVIDIA is hosting its GPU Technology Conference (GTC) event next week, where we'll be introduced to its beefed-up GB300 "Blackwell Ultra" AI GPUs, new details on its next-gen Rubin R100 AI GPU which will feature next-gen HBM4 memory, new CPO technology, and more.

JP Morgan's outlook for NVIDIA GTC 2025 has Blackwell Ultra GB300 as the highlight of the show, with the new GB300 AI chip featuring a logic structure similar to B200 but with huge capacity increases on HBM and massive increases in power consumption.

NVIDIA's new GB300 Blackwell Ultra AI GPU will have up to 288GB of 12-Hi HBM3E memory, and a thermal design power (TDP) that has been increased up to 1.4kW. We are to expect a major performance boost from GB300 of up to 50% faster than GB200 in FP4 computing performance, and JP Morgan expects GB300 to start shipping in Q3 2025.

Apple launched its new M3 Ultra processor inside of its new Mac Studio earlier this week, packing a 32-core CPU and 80-core GPU and boy-oh-boy does it perform well with DeepSeek R1 models.

In a new video from YouTuber Dave2D, who has compared Apple's new M3 Ultra against the M2 Ultra, M4 Max, and M3 Air processors, the M3 Ultra absolutely destroys DeepSeek R1 performance. The YouTuber uses DeepSeek R1 and i ts 7B, 14B, and 32B parameter models first, which the M3 Ultra takes the crown (by far) in all tests.

Apple's new M3 Ultra (which is pretty much just 2 x M3 Max chips acting as one) has up to 512GB of unified RAM support, compared to the 128GB on M2 Ultra + M4 Max, and just 24GB on the M3 Air. This leads the Apple M3 Ultra to pump out 89.07 T/s on DeepSeek R1 with 7B parameters, compared to 54.22 T/s on the M2 Ultra, 53.18 T/s on the M4 Max, and just 17.7 T/s on the M3 Air.

In a breakthrough for archaeology, scientists are using artificial intelligence and Synthetic Aperture Radar (SAR) to uncover traces of 5,000-year-old civilizations in Dubai and medieval sites in Mongolia, revealing lost histories buried beneath desert sands.

As reported by Daily Galaxy, AI and SAR imaging have revealed 5,000-year-old settlements and pathways in Dubai, providing insight into how early civilizations thrived in harsh desert conditions. Similarly, in Mongolia, the same technology uncovered thousands of potential medieval sites along the historic Silk Road - one of the most significant trade networks in human history.

The most notable aspect of these developments is uncovering ancient settlements without excavation. Detecting entire lost civilizations beneath the desert without traditional digging is a game-changer for archaeology.

The newer breed of deep reasoning models - designed to 'think' before answering - are also more open to taking any route possible to solve a given problem, it seems, even if that means cheating.

Researchers submitted a paper to Cornell university entitled 'Demonstrating specification gaming in reasoning models' which tested AIs playing games of chess on Stockfish.

They found that the new models, such as ChatGPT o1-preview and DeepSeek-R1, would "often hack the benchmark by default" - meaning resorting to cheating of one kind or another.

NVIDIA is expected to unveil its new GB300 AI GPUs at GTC 2025 later this month, with the huge shift into fully liquid-cooled AI clusters on its way.

In a new report from UDN, we're learning that the energy consumption of NVIDIA's new GB300 AI GPU has "increased significantly" from GB200, and that the heat dissipation demand is greater, which has set off a "second cooling revolution". No longer will more water-cooling plates be introduced, but the use of water-cooling quick connectors will increase 4x with GB300 compared to GB200.

NVIDIA's new GB300 AI GPUs will be fully water-cooled, with Taiwan cooling giants Shuanghong and Qihong to make "huge profits" says UDN, but we'll also see water-cooling quick connector suppliers like Fushida and Shishuo will also be a large part of this purported "second cooling revolution".

NVIDIA is preparing 64,000 of its new GB200 AI GPUs for OpenAI and Oracle to use in their expansive new Stargate AI supercomputer.

In a new report from Bloomberg, we're learning that NVIDIA's super-fast GB200 AI GPUs will be supplied for the Stargate AI supercomputer by the end of 2026 according to "a person familiar with the matter". These new GB200 chips will be added to multiple halls of the AI data center in phases, with the initial rollout of 16,000 GB200 AI GPUs to be provided by the summer.

The new Stargate AI supercomputer project was announced by President Trump along with OpenAI, Japan's SoftBank Group Corp., and Oracle at the White House in January 2025. An OpenAI spokesperson said the AI startup is working with Oracle on the design and delivery of the new data center that is destined for Abilene, Texas, USA, and that Oracle is in charge of getting and operating the Stargate supercomputer built there.

Paramount Pictures appears to be experimenting with cheap AI-generated content based on its promotion for the film Novocaine, an action-thriller starring Jack Quaid set to release on Friday.

In a recent Instagram post, users sighted a reel promoting the film that blatantly utilizes AI scripting and narration - a content style that's lazy enough from spam accounts and content farms, let alone a major film production company.

As reported by WSJ, McDonald's recently announced their plans for a technological overhaul - with over 43,000 restaurants receiving new internet-connected kitchen equipment, AI-enabled drivethroughs, and powered AI tools for managers.

The move aims to provide 'better experiences for its customers and workers', citing issues such as broken machines and incorrect orders. McDonald's Chief Technology Officer Brian Rice has been insistent on the benefits of the new AI-powered tools, highlighting the stressful work environment, and the volume of orders for customers.

Apple's much-awaited upgrade for Siri is now indefinitely delayed until at least next year, according to a new report from Bloomberg's Apple insider Mark Gurman.

Apple unveiled Apple Intelligence at WWDC in June 2024, where it teased a new Siri that would be much smarter than the current version of the digital assistant. Siri's upgrade would enable users to interact with the digital assistant on a more personal level, as Siri would have deeper access to a device with cross-app integration, control over apps, and more knowledge about the user. Moreover, Siri would work similarly to ChatGPT and be able to engage with users through natural language, fulfill multiple requests, remember past sentences in conversations, etc.

In fact, that is exactly what Apple was planning for the iPhone 16 line-up, with the company running ads such as this one where it stated the new and improved Siri will have a richer language understanding. Apple has since removed a TV advertisement for the AI-infused Siri from its YouTube channel, deciding to private the video which featured Last of Us TV show star Bella Ramsey. Reports from Bloomberg's Mark Gurman are now stating Apple plans on rolling out the AI injection into Siri "in the coming year".

If you were already concerned about self-driving cars on the local highway - look out for the AI-powered MC20 Maseratis.

As part of an initiative between the Indy Autonomous Challenge (IAC) and Italian university Politecnico di Milano, scientists have put their hands together to break the autonomous land speed record. In a recent development, the team managed to push a modified Maserati MC20 Cielo, fitted with Lidar, camera, and sensor technology, to 197mph (or 318km/h ). A staggering pace, which for reference, is nearly as fast as a commercial airplane on takeoff or a Shinkansen Japanese Bullet Train.

The feat was achieved at the Space Florida Launch and Landing Facility (LLF) at the Kennedy Space Center, one of the longest runways in the world at 2.8 miles. While primarily used for government operations, the LLF served as the perfect testing ground to push the limits of self-driving vehicles.

OpenAI is reportedly gearing up to release specialized AI agents that can be hired for as much as $20,000 per month, according to a recent article by The Information.

The article states that OpenAI is planning on releasing several "agent" products that specialize in various professions. These AI agents will vary in price depending on their specialization. For example, an AI agent software engineer could cost approximately $2,000 per month, with another software AI developer costing as much as $10,000 per month. The publication states that the most expensive AI agent could cost as much as $20,000 per month and would come with knowledge proportionate to "Ph.D.-level research."

It's unclear when these AI agents will be released or if the aforementioned prices are set in stone, but SoftBank, an OpenAI investor, has committed to spending as much as $3 billion on OpenAI agent products in 2025. It appears OpenAI will need a $20,000 per month subscription, as the company paid as much as $5 billion in running costs for its products last year, among other expenses.

NVIDIA founder and CEO Jensen Huang along with industry visionaries will unveil "what's next in AI" at GTC 2025, the world's premier AI conference.

GTC 2025 will bring the brightest minds in AI to showcase breakthroughs happening now in physical AI, agentic AI and scientific discovery. The GPU Technology Conference (GTC) will see 25,000 attendees in person, and over 300,000 attendees virtually, where you can stream GTC 2025 right here when it kicks off March 17-21 in San Jose, California.

Jensen said: "AI is pushing the limits of what's possible -- turning yesterday's dreams into today's reality. GTC brings together the brightest scientists, engineers, developers and creators to imagine and build a better future. Come and be first to see the new advances in NVIDIA computing and breakthroughs in AI, robotics, science and the arts that will transform industries and society".

Amazon has announced Alexa+, its new generative AI-powered Alexa voice assistant that has beaten Apple to the punch.

Amazon's new Alexa+ is conversational, proactive, capable of reason and inference, and can learn from context and your life. It is a big leap into the true AI assistant world, where Alexa+ can plan a date, book a restaurant, text your loved ones, create a travel itinerary and add it to everyone's calendar, read your study guide, and give you a test on the answers, remember your favorite foods and movies, as well as finding your smart devices and take action. Phew.

It's not just boasting a host of new AI abilities, but Alexa+ will now be better than everything it used to do, no longer having to use precise phrasing or constantly having to say "Alexa". In the place of Skills, Amazon now has Experts, which is Alexa's various large language models (LLMs) that smartly tap into deliver on your requests however you phrase them -- this includes ordering food, turning on lights, through to things like who has been at your door, and what to cook for dinner.