Artificial Intelligence - Page 55

Sam Altman, the CEO of OpenAI, the company behind many popular and powerful AI-powered tools such as ChatGPT, Sora and the underlying technology called GPT, was asked during an interview if he was worried about losing control of AGI once its created.

Altman sat down for an interview with Lex Fridman on the "Lex Fridman Podcast," where he was asked if he worries about "losing control of the AGI itself", as there are many people out there, including numerous security researchers concerned about the creation of a super-intelligent AI system and the implications that has on the planet through existential risk. Fridman prefaces the question by saying that "losing control" would not be because of "state actors, not because of security concerns, but because of the AI itself".

The OpenAI CEO responded promptly by saying, "That is not my top worry as I currently see things." Adding, "there have been times I worried about that more. There may be times again in the future where that's my top worry. That's not my top worry right now." Fridman followed up by asking "What's your intuition about that not being your top worry," adding "do you think we could be surprised?".

Sam Altman, the CEO of OpenAI, the company behind GPT-4, ChatGPT, Sora, and many other industry-leading AI technologies, has sat down for an interview with Lex Fridman to discuss multiple topics regarding artificial intelligence and the impressive creations being made at OpenAI.

Lex Fridman asked Altman if he trusts himself with the power of leading a company that could potentially create the first Artificial General Intelligence (AGI), an AI system that is capable of human-level and beyond intelligence. Altman responded honourably to the question, saying that he believes that its important that "I nor any other one person have total control over OpenAI, or over AGI, and I think you want a robust governance system."

The OpenAI CEO further explained that he "continues to not want super voting control over OpenAI" and "I continue to think that no company should be making these decisions and that we really need governments to put rules of the road in place."

YouTube has taken to its blog to announce a new rule for its platform designed to protect users from being deceived by creators who are trying to pass off their content as authentic when it was created using the assistance of artificial intelligence-powered tools.

With the unveiling of OpenAI's Sora tool, which is designed to create photorealistic video content from user text prompts, and the popularity of ChatGPT, it isn't a reach to say that AI-powered tools such as Sora will be popular when they are fully developed and released.

Some of the video examples provided by Sora creator OpenAI featured tell-tale signs of AI-powered creation, but the vast majority at a glance, or even viewed by the untrained eye, wouldn't raise alarm bells for synthetic creation, further blurring the line between what's real and what's fake on the internet.

NVIDIA has unveiled its next-gen Blackwell AI GPU, which features 208 billion transistors, 192GB of super-fast HBM3E memory, a cutting-edge custom TSMC 4NP process, and groundbreaking new networking that sees NVIDIA NVLink deliver 1.8TB/s bidirectional throughput per GPU.

The company also announced its new NVIDIA GB200 Grace Blackwell Superchip, which connects two NVIDIA B200 Tensor Core GPUs to the NVIDIA Grace CPU "over a 900GB/s ultra-low-power NVLink chip-to-chip interconnect." With serious AI horsepower, NVIDIA's powerful next-gen AI hardware arrives during the AI boom, so, unsurprisingly, it announced its long list of Blackwell partners.

NVIDIA writes, "Among the many organizations expected to adopt Blackwell are Amazon Web Services, Dell Technologies, Google, Meta, Microsoft, OpenAI, Oracle, Tesla, and xAI." Even though most, if not all, of these companies are investing in creating their own AI chips and hardware for generative AI-which is used in everything from scientific research to medicine and chatbots-Blackwell will drive many of the advances over the next few years.

It's been a massive day for AI news, with NVIDIA's premiere GTC 2024 AI conference underway. Earlier today, the company unveiled its next-gen Blackwell AI GPU - an unprecedented AI monster with 208 billion transistors and 192GB of super-fast HBM3E memory.

NVIDIA's first multi-GPU die is built on a cutting-edge TSMC 4NP process. It allows organizations to run real-time generative AI on "trillion-parameter large language models" at up to 25X less cost and energy consumption than its Hopper predecessor. Throw in the latest version of NVIDIA NVLink that delivers 1.8TB/s bidirectional throughput per GPU, and it's no wonder everyone is signing up to join the Blackwell family - including Microsoft.

Alongside the big Blackwell reveal, NVIDIA and Microsoft have announced that Microsoft Azure will adopt NVIDIA Grace Blackwell Superchips to help accelerate customer and first-party (Microsoft's own) AI offerings, Including Microsoft 365, aka the online versions of Microsoft Word, PowerPoint, Excel, and OneNote. NVIDIA GPUs and NVIDIA Triton Inference Server will help power Microsoft Copilot for Microsoft 365.

Localization is a big thing in gaming, where a game's dialogue and presentation are recorded in various languages to reach a global audience. Like how Netflix dubs all of its original programming into multiple languages, localization for games is a complex process that requires accurately translating dialogue and text, maintaining dramatic or comedic tones, and then re-recording dialogue with voice actors from different countries.

Where games differ from movies or television drama is that all performances are digital and completely malleable, which is why you've got a situation where Sony's PlayStation-exclusive Ghost of Tsushima (which is coming soon to PC) offers both English and native Japanese language options with full lip-syncing and correct facial animation.

Although it's possible to do this for dozens of languages, Ghost of Tsushima limits full lip-synching to two language options due to the task's complexity and the animation involved. This is where generative AI, specifically NVIDIA's Audio2Face technology, will step in.

At GDC 2024, NVIDIA presented a new tech demo for a theoretical detective game, Covert Protocol, created in collaboration with Inworld. If you recall NVIDIA's recent AI collabs, which created cyberpunk-style tech demos featuring AI avatars you can interact with, this takes it all one step further. It presents an old-school adventure game where you're talking to characters to solve a mystery in a brand-new way.

In Covert Protocol, you play a private detective and explore a realistic environment, speaking to 'digital humans' as you piece together critical and key information. The game is powered by the Inworld AI Engine, which fully uses NVIDIA ACE (the technology we've been following over the past year) services to ensure that no two playthroughs are the same.

"This level of AI-driven interactivity and player agency opens up new possibilities for emergent gameplay," NVIDIA writes. "Players must think on their feet and adapt their strategies in real-time to navigate the intricacies of the game world."

NVIDIA isn't just changing up the GPU and AI GPU game with its Blackwell AI GPU chips, announcing Earth-2 today during the GPU Technology Conference (GTC).

NVIDIA announced its new Earth-2 climate digital twin cloud platform so people could simulate and visualize weather and climate at scales never seen before. Earth-2's new cloud APIs are available on NVIDIA DGX Cloud, allowing virtually anyone to create AI-powered emulations to make interactive, high-resolution simulations ranging from the global atmosphere to localized cloud cover all the way through to typhoons and mega-storms.

The new Earth-2 APIs offer AI models that use new NVIDIA generative AI model technology called CorrDiff, using state-of-the-art diffusion modeling, capable of generating 12.5x higher resolution images than current numerical models that are 1000x faster and 3000x more energy efficient.

GIGABYTE is showing off its next-gen compact GPU cluster scalable unit: a new rack with GIGABYTE G593-SD2 servers, which have NVIDIA HGX H100 8-GPU designs and Intel 5th Gen Xeon Scalable processors inside.

The company has said it will support NVIDIA's new Blackwell GPU that succeeds Hopper, with enterprise servers "ready for the market according to NVIDIA's production schedule". The new NVIDIA B200 Tensor Core GPU for generative AI and accelerated computing will have "significant benefits," says GIGABYTE, especially in LLM inference workloads.

GIGABYTE will have products for HGX baseboards, Superchips, and PCIe cards with more details to be provided "later this year," adds the company.

NVIDIA has just revealed its new Blackwell B200 GPU, with its new DGX B200 systems ready for the future of AI supercomputing platforms for AI model training, fine-tuning, and inference.

The new NVIDIA DGX B200 is a sixth-generation system that's air-cooled in a traditional rack-mounted DGX design used worldwide. Inside, the new Blackwell GPU architecture powers the DGX B200 system using 8 x NVIDIA Blackwell GPUs and 2 x Intel 5th Gen Xeon CPUs.

Each DGX B200 system features up to 144 petaFLOPS of AI performance, an insane 1.4TB/sec of GPU memory (HBM3E) with a bonkers 64TB/sec of memory bandwidth, driving 15x faster real-time inference for trillion-parameter models over the previous-gen Hopper GPU architecture.

NVIDIA has finally announced its new Blackwell GPU, DGX system, and Superchip platforms all powered by Blackwell B200 AI GPU and Grace CPU.

The new NVIDIA GB200 Grace Blackwell Superchip is a processor for trillion-parameter-scale generative AI, with 40 petaFLOPS of AI performance, a whopping 864GB of ultra-fast HBM3E memory with an even more incredible 16TB/sec of memory bandwidth.

Each of the new GB200 Grace Blackwell Superchips will feature 2 x B200 AI GPUs and a single Grace CPU with 72 Arm-based Neoverse V2 cores. Alongside the 864GB HBM3E memory pool, 16TB/sec memory bandwidth is joined by a super-fast 3.6TB/sec NVLink connection.

NVIDIA has just revealed its next-gen Blackwell GPU with a few new announcements: B100, B200, and GH200 Superchip, and they're all mega-exciting.

The new NVIDIA B200 AI GPU features a whopping 208 billion transistors made on TSMC's new N4P process node. It also has 192GB of ultra-fast HBM3E memory with 8TB/sec of memory bandwidth. NVIDIA is not using a single GPU die here, but a multi-GPU die with a small line between the dies differentiating the two dies, a first for NVIDIA.

The two chips think they're a single chip, with 10TB/sec of bandwidth between the GPU dies, which have no idea they're separate. The two B100 GPU dies think they're a single chip, with no memory locality issues and no cache issues... it just thinks it's a single GPU and does its (AI) thing at blistering speeds, which is thanks to NV-HBI (NVIDIA High Bandwidth Interface).

ZOTAC has just announced its expanded GPU Server Series systems. The first of the series is the Enterprise lineup, which offers companies affordable, high-performance computing solutions for countless applications, including AI.

These new ZOTAC systems are aimed at core-to-core inferencing, data visualization, model training, HPC modeling, simulation, and AI workloads. ZOTAC's new family of GPU servers comes in varying form factors and configurations, with the Tower Workstation and Rack Mount Servers both offered in either AMD EPYC or Intel Xeon CPU form.

There's support for a huge 10 x GPUs, with a modular design that makes it easier to get in and configure the system, with a high space-to-performance ratio, industry-standard features like redundant power supplies and various cooling options, ZOTAC has your back with its new GPU Server Series.

NVIDIA will unveil its next-generation Blackwell GPU architecture at GTC 2024... tomorrow, if you can believe it, detailing its new B100 AI GPU and giving us a tease at the beefed-up B200 AI GPU expected in 2025.

In a new post on X by "XpeaGPU," we hear that NVIDIA's new B100 is truly a monster: 2 x GPU dies on the latest TSMC CoWoS-L (Chip-on-Wafer-on-Substrate-L) 2.5D packaging technology, which allows companies to design and manufacture larger chips. NVIDIA's next-gen B100 will have up to 192GB of ultra-fast HBM3E memory on 8-Hi stacks, while the beefed-up B200 AI GPU will feature a huge 288GB of HBM3E memory.

NVIDIA's current H100 AI GPU ships with 80GB or 141GB of HBM3 memory, while its competitor, the AMD Instinct MI300X, ships with 192GB of HMB3 memory. The release of the B100 AI GPU will see NVIDIA match AMD for the amount of HBM memory, but NVIDIA's new B100 will have the new ultra-fast HBM3E memory and will be the first GPU with HBM3E to market.

NVIDIA and Microsoft-backed humanoid robot Figure has published a new video of what it calls "Figure 01", a humanoid robot powered by OpenAI technology designed to simulate speech.

Figure, which raised $675 million in Series B funding at a $2.6 billion valuation received funding from companies such as Microsoft, OpenAI's Startup fund, NVIDIA, Amazon founder Jeff Bezos and much more. The goal of the company is to develop "next-generation AI models for humanoid robots" and judging by the latest video posted by the company they are well along the way of achieving that.

The new video shows an engineer chatting with Figure 01, with the engineer asking the humanoid robot, "Can I have something to eat?" to which the robot responded, "Sure thing," and then proceeded to hand over a red apple. Figure 01 was then asked why it "did what it just did" while it was picking up trash from a table. The robot explained that it gave the engineer the red apple as it was the "only edible item I could provide you with from the table."

Dr. Lisa Su, AMD's CEO, delivered the keynote at SXSW the other day with her speech focused on the future of AI. "AI is the most important technology to come on the scene in the last 50 years," Lisa Su said. Adding, "Companies that learn how to leverage AI are going to win over companies that are not."

With that, AMD is all in on AI, leveraging AI to help design better chips and software. Lisa Su added that AI is a productivity tool within AMD's walls. With 2024 ushering in the era of the AI PC, there are already plenty of options out there (both mobile and desktop) powered by AMD Ryzen CPUs with inbuilt NPU hardware and Radeon RX GPUs with dedicated AI hardware.

Like other big players, Lisa Su and AMD advocate for an open-source ecosystem for AI because "no one company has all the answers" when it comes to building the AI future; "it takes a village."

Sony has announced that PS5 system software version 24.02-09.00.00 contains controller improvements.

PlayStation 5 owners received a notification on Wednesday to download the latest version of their console, which makes the DualSense and DualSense Edge controller speakers louder when used to produce in-game sounds and voice chat. Additionally, Sony writes on its website that details the update version 24.02-09.00.00 will also enhance the microphone through a "new AI machine-learning mode" that reduces the sound of background noise caused by button presses, thus improving the voice chat experience.

The update has also added some more customizability to the PS5 power indicator, which can be found in Settings > System > Beep and Light > Brightness. From there, users can select between dim, medium, and bright, which is the default setting. Other new features within the update are pointers and emoji reactions to the Share Screen feature, which viewers of your shared screen can send to the host. Notably, emojis and pointers can be switched off by the host within the Share Screen settings.

NVIDIA and AMD's best AI GPUs use HBM3 memory, but with the introduction of the H200 AI GPU, NVIDIA will be the first to market with an HBM3E-based AI GPU.

HBM3E will be found inside of NVIDIA's upcoming H200 and next-gen B100 AI GPUs, with TrendForce noting that the supply bottleneck through advanced CoWoS packaging technology and the long production cycle of HBM extend the timeline from wafer initiation to the final production past 6 months.

NVIDIA's current H100 AI GPU uses HBM3 memory primarily supplied by SK hynix, which has caused stock worldwide issues due to the crazy-high demand for AI GPUs. Samsung's entry into NVIDIA's supply chain with its new HBM3 memory in late 2023, were "initially minor, signifies its breakthrough in this segment," reports TrendForce.

A new report commissioned by the US State Department warns the exponential development of artificial intelligence may pose a significant risk to national security and even humanity.

The new report titled "An Action Plan to Increase the Safety and Security of Advanced AI" recommends the US government move "quickly and decisively" with implementing measures that hinder the rise of artificial intelligence-powered systems being developed, even to the point of potentially limiting compute power used to train such models. The report goes on to say that if these hindering measures aren't implemented, there is a chance of AI or Artificial General Intelligence (AGI) being an "extinction-level threat to the human species."

The US State Department report involved more than 200 experts in the field, which included officials from companies that are big players in the AI game, such as OpenAI, Meta, Google, Google DeepMind, and government workers. The report goes on to recommend the US government implement limitations on how much compute power any given party developing AI is able to have at one time while also requiring AI companies to request permission from the US government to train any new AI model.

It was only last month that OpenAI revealed its upcoming text-to-video generator platform named Sora, and the general reaction to the new AI-powered tool was impressive yet concerning.

The upcoming AI-powered tool works exactly the same way as OpenAI's extremely popular ChatGPT, but instead of the chatbot responding to user prompts with text its capable of producing high-quality video, even to the point of photorealism. OpenAI took to its YouTube channel to share the above video showcasing Sora's capabilities and at first glance it appears some of the examples shown are shot with a real-life camera.

However, upon closer inspection of the examples tell-tale signs of AI-generated content begin to stand out, such as physics-based movements like people walking, hand movements, and more. OpenAI is currently "red-teaming" Sora to iron out these issues before its released to the public, which means people are pushing the AI model to its brink to bring these vulnerabilities to light so they can be fixed.