A Chip Off the Old Stock

PLUS: No Chip Off the Old Block

June 28, 2023

The Hitchhiker’s Guide to AI

Welcome, Wednesday warriors of the AI realm! We're halfway through the week, and if you're feeling like a bot running on low battery, don't worry, our AI espresso shot is here to recharge you. Like a trusty algorithm, we're here to help you optimize your week. Now, let's dive back into the ocean of AI, ready to uncover more pearls of wisdom. Remember, even the most complex algorithms began as a simple line of code.

In today’s Tracker:

  • 🧪 Research highlights: A New Frontier in Machine Learning

  • 🚨 Industry news: Nvidia Leads Decline Amid AI Chip Sale Concerns

  • ⚖️ Policy and regulation: US Ponders Stricter AI Export Rules to China

  • 🌐 AI and society: Reshaping the AI Carbon Footprint

  • 🧰 Tool of the day: Mailyr: Effortless Emails

💻A New Frontier in Machine Learning

A recent study by the Los Alamos National Laboratory revealed a significant theoretical advance in quantum machine learning, specifically concerning the application of overparametrization in quantum neural networks (QNNs).

The researchers demonstrated that increasing the parameters in a QNN beyond a critical threshold significantly improves its performance and training simplicity.

This process, known as overparametrization, is a well-established technique in classical machine learning, but its implications for quantum models remained obscure until now.

The primary contribution of this study is developing a theoretical framework that predicts the critical number of parameters required for a quantum machine learning model to become overparametrized, resulting in improved training efficiency and performance.

The authors used the analogy of a hiker investigating a landscape to illustrate how increasing the number of parameters in a QNN provides more exploration directions, thereby avoiding false peaks (suboptimal solutions) and leading to the proper peak (optimal solution).

The team anticipates that their findings will be instrumental in leveraging machine learning to comprehend better quantum data, such as the classification of various phases of matter, a task that classical computers find challenging.

 📈 Nvidia Leads Decline Amid AI Chip Sale Concerns

In response to reports that the United States could close loopholes in the sale of powerful AI processors to China, tech stocks declined, with Nvidia Corp. leading the decline.

Nvidia, which derives roughly one-fifth of its revenue from China, declined as much as 3.2% in after-hours trading in New York, while its competitor, Advanced Micro Devices Inc., fell about 3%.

The two companies are market leaders for processors integral to creating generative AI models such as ChatGPT.

The news also affected trading in China, where various AI-related equities were sold, resulting in a 10% decline for key hardware suppliers Inspur Electronic Information Industry Co. and Unisplendour Corp.

Potential restrictions could substantially reduce sales in the largest semiconductor market in the world, causing a ripple effect throughout the technology sector.

The primary focus is the effect on Nvidia and other tech equities because the United States plans to tighten restrictions on AI chip exports to China.

🇺🇸 The US Ponders Stricter AI Export Rules to China

The United States is considering imposing additional restrictions on the export of artificial intelligence processors to China, which could affect the operations of U.S. chipmakers like Nvidia, Micron, and AMD.

Shares of Nvidia and AMD fell more than 2% and 1.5%, respectively, as a result of this upcoming event, which is expected to take effect as early as July. Notably, Nvidia had previously complied with US officials' requests to cease exporting two of its top AI processors to China. Subsequently, it introduced a new chip, the A800, designed specifically to comply with export control regulations.

🏜 Reshaping the AI Landscape From Carbon Footprints to Green Milestones

The AI industry is presently experiencing a global paradigm shift, as enterprise adoption has doubled since 2017 and compute volumes have risen exponentially, resulting in significant environmental implications due to increased electricity consumption and data storage requirements. Consequently, a more sustainable approach to AI is imperative.

Fortunately, AI can also be a part of the solution, with the potential to promote sustainability across multiple industries.

Combining real-time data collection with AI can help identify operational areas for large-scale emission reduction.

Examples include optimizing building efficiency factors such as HVAC systems and implementing carbon-aware computing by shifting duties in accordance with the availability of renewable energy sources.

AI can also aid in managing the developing data storage problem by identifying and removing unnecessary data.

In addition, a more responsible design of AI projects can be attained by emphasizing data quality over quantity, considering the level of accuracy truly required for a particular use case, leveraging domain-specific models, balancing hardware and software, and utilizing open-source solutions and libraries of optimizations.

Despite requiring effort, forethought, and compromise, these measures can result in significant advancements in reducing carbon emissions and mitigating the effects of climate change.

📤 Mailyr: Your AI-Powered Wordsmith for Effortless Emails

Mailyr removes the tension from email composition. Type a few sentences, and AI will handle the rest.

Snippets

Did you miss our last newsletter? Check it out here!

He who conquers himself is the mightiest warrior. — Confucius

  • 📨 Pipl.ai - Send cold emails at scale with ease.

  • ✍🏽 Wordhero - Create original content in seconds.

  • ⌨️ Smartwriter - Creates personalized cold emails/Linkedin messages.

If you liked this edition, please share it with a friend or two! You could get some exclusive content in your inbox or even some merch very soon👇🏾

And that's a wrap!

Did we miss anything? Or want to say hey? Hit reply - I'd love to hear from you! You can

And if you haven’t already, sign up to get this in your inbox daily.