To Be or Not To Be

Can't read or see images? View this email in a browser

Chipmaker AMD is in a dilemma. While they want to be fully compliant with the US export controls, they also don’t want to lose their grip on the Chinese market. Tough call. Meanwhile, in an earnings call on Tuesday, AMD’s CEO Lisa Su expressed her concern for the company.  Since last year, the US has been imposing back-to-back restrictions on semiconductor companies trying to prevent them from exporting high-performing GPUs to China.  The ban on chip sales to Chinese customers prohibits the offering of advanced chips with high performance (at least 300 teraops) and fast interconnect speed (around 600 gigabytes per second). Nvidia's A100 chip is one such example, surpassing the performance requirement with 600 teraops and meeting the necessary interconnect speed. In this geo-political tussle between the US and China, NVIDIA, Intel and AMD have lost one of the biggest chip markets in the world. China imports about 53.7% of the world’s supply of chips worth around $240 billion.  Affected by the brutal losses, these chipmakers are trying to circumvent the system by tweaking the product to win back the Chinese market.  In response to the export regulations, Nvidia made adjustments to its A100 chip and introduced the A800 version in November last year to comply with legal export to China. Subsequently, in March 2023, the company unveiled another version for China, the H800 chip, which is a modified version of the H100 chip. Chinese tech giants like Alibaba Group Holding Ltd, Baidu Inc, and Tencent Holdings Ltd have already adopted this chip for their cloud computing units. In July this year, Intel made a strategic move by launching the Gaudi2 HL-225B chip, specifically designed for the Chinese market. To comply with US trade restrictions, Intel slightly reduced the interconnect bandwidth by around 17%. As of now, AMD's MI300s surpass the performance limits imposed by the export controls in October last year, preventing the sale of certain advanced chips to China. While Nvidia and Intel have already introduced customised chips for the Chinese market, AMD is yet to follow suit. It remains to be seen if AMD will develop chips specifically tailored for China in the future. Read the full story here. Open Source AI for Good NASA, Hugging Face, and IBM have come together to create Earth's largest open-source foundation model for climate scientists and researchers. The collaboration aims to efficiently utilize data from NASA's Earth Science Missions to build AI models that will benefit humanity. The model, available on Hugging Face's platform, uses NASA's satellite data and IBM's watsonx.ai. IBM and NASA are also working with Clark University to adapt the model for various climate and Earth science applications. By leveraging open-source technologies, this partnership aims to accelerate advancements in climate change solutions.  Open-source software plays a significant role in various sectors, and the Call For Code initiative by IBM also focuses on sustainability issues. Embracing open-source solutions can lead to diverse perspectives and more impactful solutions for a better planet. Read the full story here. Vulnerable Microsoft  In March 2023, a cybersecurity engineer at Tenable found a critical issue with the Microsoft Azure platform that allowed unauthorized access to cross-tenant applications and sensitive data. The breach was so severe that it exposed authentication secrets to a bank. Although Microsoft acknowledged the problem, it took them three months to fully resolve the issue.  Furthermore, the fix was only implemented for new applications, leaving existing users vulnerable. This lack of transparency and slow response drew criticism from Amit Yoran, Tenable's chairman and CEO. The rise of generative AI capabilities through Microsoft's AI offerings adds new cybersecurity challenges, and the effectiveness of their cybersecurity measures is being closely scrutinized in the face of evolving threats. Read the full story here. A Balancing Act Large language models are struggling to strike the right balance between bias and variance, also known as the bias-variance tradeoff. Bias refers to oversimplifying data patterns, leading to underfitting, while variance is the sensitivity to changes in the training data, causing overfitting.  Finding the right balance is crucial for optimal LLM performance. However, it's challenging because bias and variance are inversely connected, making it hard to minimize both. These models, like GPT, with billions of parameters, are susceptible to the tradeoff. Underfitting can be tackled by including diverse examples, and overfitting by employing techniques like regularization and data augmentation. Continuously refining model architectures and using diverse datasets can also help alleviate bias and enhance LLM performance. Read the full story here. TAUSIF ALAM & AMIT RAJA NAIK Friday, Aug 4, 2023 | Was this email forwarded to you? Sign up here DOWNLOAD OUR MOBILE APP Stay Connected info@analyticsindiamag.com © 2023 Analytics India Magazine

This email was sent by support@analyticsindiamag.com to alexvarboffin.abbb@blogger.com Not interested? Unsubscribe | Manage Preference | Update profile Analytics India Magazine | 280, 2nd floor, 5th Main, 15 A cross, Sector 6, HSR layout Bengaluru, Karnataka 560102