Wednesday☕️

Time Is Money

Sep 4, 2024 • 5 min read

Trending:

Researchers Find Brain Memory System
OpenAI‘s New Custom AI Chips

Yesterday’s U.S. stock market:

Yesterday’s commodity market:

Yesterday’s crypto market:

Researchers Find Brain Memory System:

Researchers from New York University, in collaboration with colleagues from institutions such as McGill University in Canada and the University Hospital of Münster in Germany, have identified a critical interaction between two molecules, PKMzeta and KIBRA, that is essential for the stabilization and maintenance of long-term memories.
PKMzeta is an enzyme known for its role in strengthening synaptic connections between neurons, which are crucial for storing memories. KIBRA, a molecule associated with memory performance, anchors PKMzeta at specific synapses, ensuring that these connections remain stable over time. This interaction is often referred to as "memory glue" due to its role in keeping memories intact despite the brain's ongoing molecular turnover. This "memory glue" is particularly significant because it explains how memories can last for many years, even though the proteins involved in their maintenance are continually being replaced.
Typically, brain proteins are replaced within days, but the persistent bond between PKMzeta and KIBRA allows the synaptic connections necessary for memory retention to remain intact over time. Disruption of this bond has been shown to weaken long-term memories, underscoring the critical role this interaction plays in preserving our experiences. By understanding how PKMzeta and KIBRA work together to maintain memory stability, scientists can explore new therapeutic approaches for memory-related disorders, such as Alzheimer's disease.
Targeting this interaction could potentially help preserve or even enhance memory function in individuals affected by such conditions. As research progresses, a deeper understanding of this "memory glue" could pave the way for breakthroughs in both the treatment of memory-related disorders and the development of technologies that enhance memory storage and retention capabilities.

OpenAI‘s New Custom AI Chips:

OpenAI has decided to develop its own AI chips in collaboration with Taiwan Semiconductor Manufacturing Company (TSMC) to reduce its reliance on third-party hardware, particularly Nvidia’s AI servers. By securing production slots for TSMC’s upcoming 1.6 nm A16 process node, OpenAI aims to improve performance, power efficiency, and chip density—essential elements for running large-scale AI models. This move is designed to optimize computational resources for applications like OpenAI’s text-to-video service, Sora.
The A16 node, which is expected to enter mass production by 2026, incorporates advanced technologies such as Gate-All-Around (GAAFET) nanosheet transistors and backside power delivery, known as Super Power Rail. These innovations are intended to boost power efficiency and chip performance at smaller scales, which is essential for meeting the growing computational needs of AI models. Originally, OpenAI considered using TSMC’s N5 process node but shifted its focus to A16 for its long-term benefits. The company may also collaborate with Broadcom and Marvell for chip design.
This development aligns with a growing trend in the industry, where companies like Google and Amazon are also moving towards designing custom chips to enhance AI training and computational capabilities. Custom chips allow these companies to address the limitations of general-purpose hardware, as they can be fine-tuned to optimize machine learning and AI-specific tasks. By creating hardware tailored to their software, firms aim to increase efficiency, reduce costs, and improve scalability as the demand for AI grows.
For OpenAI, developing custom chips offers significant advantages in overcoming current hardware limitations. Custom chips can be optimized for the parallel processing and data-intensive tasks that AI workloads require, leading to better power efficiency and lower operational costs. By aligning hardware more closely with the needs of its AI models, OpenAI can accelerate innovation, improve performance, and scale its infrastructure more effectively. This reflects the broader industry shift towards specialized hardware as AI demands continue to grow.

Statistic:

Largest assets by market capitalization: