Lifeboat Foundation

NVIDIA’s Blackwell GB200 AI servers are anticipated to see major traction, reaching 2 million units shipped in 2025 & utilizing new packaging tech.

NVIDIA To Overcome CoWoS Supply Chain Bottlenecks By Shifting To The Newer “PFLO” Standard, 420K Units Shipping This Year With Up To 2 Million Anticipated For 2025

The success of NVIDIA’s Hopper AI products last year not only uplifted the company’s economics to new heights but also revealed massive flaws in the supply chain. Due to these flaws, the products became victims of long order backlogs. The main culprits at that time were HBM & CoWoS supply, which was in a much inferior position to what it is today. Despite seeing massive upgrades, NVIDIA has decided to resolve CoWoS issues with its latest Blackwell product, as the firm is rumored to have switched to a newer packaging technology by 2025–2026.

So dl what you can…

WASHINGTON, May 22 (Reuters) — The House Foreign Affairs Committee on Wednesday voted overwhelmingly to advance a bill that would make it easier for the Biden administration to restrict the export of artificial intelligence systems, citing concerns China could exploit them to bolster its military capabilities.

The bill, sponsored by House Republicans Michael McCaul and John Molenaar and Democrats Raja Krishnamoorthi and Susan Wild, also would give the Commerce Department express authority to bar Americans from working with foreigners to develop AI systems that pose risks to U.S. national security.

Continue reading “US lawmakers advance bill to make it easier to curb exports of AI models” »

At a minimum, systems will need 16GB of RAM and 256GB of storage, to accommodate both the memory requirements and the on-disk storage requirements needed for things like large language models (LLMs; even so-called “small language models” like Microsoft’s Phi-3, still use several billion parameters). Microsoft says that all of the Snapdragon X Plus and Elite-powered PCs being announced today will come with the Copilot+ features pre-installed, and that they’ll begin shipping on June 18th.

But the biggest new requirement, and the blocker for virtually every Windows PC in use today, will be for an integrated neural processing unit, or NPU. Microsoft requires an NPU with performance rated at 40 trillion operations per second (TOPS), a high-level performance figure that Microsoft, Qualcomm, Apple, and others use for NPU performance comparisons. Right now, that requirement can only be met by a single chip in the Windows PC ecosystem, one that isn’t even quite available yet: Qualcomm’s Snapdragon X Elite and X Plus, launching in the new Surface and a number of PCs from the likes of Dell, Lenovo, HP, Asus, Acer, and other major PC OEMs in the next couple of months. All of those chips have NPUs capable of 45 TOPS, just a shade more than Microsoft’s minimum requirement.

“Recall uses Copilot+ PC advanced processing capabilities to take images of your active screen every few seconds,” Microsoft says on its website. “The snapshots are encrypted and saved on your PC’s hard drive. You can use Recall to locate the content you have viewed on your PC using search or on a timeline bar that allows you to scroll through your snapshots.”

By performing a Recall action, users can access a snapshot from a specific time period, providing context for the event or moment they are searching for. It also allows users to search through teleconference meetings they’ve participated in and videos watched using an AI-powered feature that transcribes and translates speech.

At first glance, the Recall feature seems like it may set the stage for potential gross violations of user privacy. Despite reassurances from Microsoft, that impression persists for second and third glances as well. For example, someone with access to your Windows account could potentially use Recall to see everything you’ve been doing recently on your PC, which might extend beyond the embarrassing implications of pornography viewing and actually threaten the lives of journalists or perceived enemies of the state.

Yann LeCun argues current AI methods are flawed as he pushes for ‘world modelling’ vision for superintelligence.

An international research team including Los Alamos National Laboratory and Tel Aviv University has developed a unique, mechanical metamaterial that, like a computer following instructions, can remember the order of actions performed on it. Named Chaco, after the archaeological site in northern New Mexico, the new metamaterial offers a route to applications in memory storage, robotics, and even mechanical computing.

OpenAI has landed itself in hot water for pushing out an update to ChatGPT that features a virtual assistant with an uncanny vocal resemblance to Scarlett Johansson — and it could be staring down the barrel of a compelling lawsuit.

Almost instantly, comparisons to the movie “Her” abounded, in which the actress plays a chatbot named Samantha that falls in love with a lonely man. Had OpenAI just aped her role — and her voice? Officially, it said no. Then, Johansson dropped a bombshell: leadership at the AI startup had in fact asked permission to use her voice last year. She said no, and they did it anyway.

The codes to replicate the simulations of the paper: Available at: https://arxiv.org/abs/2405.12832 and also: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4835325

For now, we just added the codes to…

In this paper, we introduce Wav-KAN, an innovative neural network architecture that leverages the Wavelet Kolmogorov-Arnold Networks (Wav-KAN) framework to enhance interpretability and performance. Traditional multilayer perceptrons (MLPs) and even recent advancements like Spl-KAN \cite{kan} face challenges related to interpretability, training speed, robustness, computational efficiency, and performance. Wav-KAN addresses these limitations by incorporating wavelet functions into the Kolmogorov-Arnold network structure, enabling the network to capture both high-frequency and low-frequency components of the input data efficiently. Wavelet-based approximations employ orthogonal or semi-orthogonal basis and also maintains a balance between accurately representing the underlying data structure and avoiding overfitting to the noise. Analogous to how water conforms to the shape of its container, Wav-KAN adapts to the data structure, resulting in enhanced accuracy, faster training speeds, and increased robustness compared to Spl-KAN and MLPs. Our results highlight the potential of Wav-KAN as a powerful tool for developing interpretable and high-performance neural networks, with applications spanning various fields. This work sets the stage for further exploration and implementation of Wav-KAN in frameworks such as PyTorch, TensorFlow, and also it makes wavelet in KAN in wide-spread usage like nowadays activation functions like ReLU, sigmoid in universal approximation theory (UAT).

editor of[Possible Minds: Twenty-Five Ways of Looking at AI], talked about the future of artificial intelligence with two contributors to the book, David Chalmers and Daniel Dennett.