• The AI Pulse
  • Posts
  • 🤖 China’s ByteDance Exploits AI Chip Loophole

🤖 China’s ByteDance Exploits AI Chip Loophole

PLUS: AI’s $200 Billion Question?, OpenAI Dissects GPT-4’s Inner Workings

Author

James Roesser
June 10, 2024

Welcome back AI enthusiasts!

In today’s AI Report:

  • 🇨🇳China’s ByteDance Exploits AI Chip Loophole

  • 💵AI’s $200 Billion Question?

  • 🧫OpenAI Dissects GPT-4’s Inner Workings

  • 🛠5 Trending Tools

  • 💰Venture Capital Updates

  • 💼Who’s Hiring?

Read Time: 3 minutes

🗞RECENT NEWS

BYTEDANCE

🇨🇳China’s ByteDance Exploits AI Chip Loophole

Image Source: Division/CSIRO/Supercomputer Cluster

China’s internet giant and TikTok parent company, ByteDance, is renting Nvidia’s AI chips on U.S. soil to exploit a loophole in the Biden administration’s export restrictions.

Key Details:

  • Due to national security concerns, the U.S. government prohibits Nvidia from directly selling AI chips like the H100 Tensor Core GPUs to Chinese companies.

  • However, the restrictions don’t prevent Chinese companies from renting AI chips within the U.S.

  • ByteDance allegedly exploited this loophole by renting out Oracle servers equipped with H100 Tensor Core GPUs to train language models.

  • ByteDance reportedly had access to over 1,500 Nvidia H100 Tensor Core GPUs last month through the Oracle partnership.

  • Other Chinese internet giants, such as Alibaba and Tencent, are exploring renting from U.S. providers or building U.S.-based data centers.

Why It’s Important:

  • China doesn’t possess the correct infrastructure or knowledge to produce AI chips. So, without stockpiled Nvidia AI chips, they can’t pursue training AI models and must halt Large Language Model (LLM) projects.

  • Taiwan Semiconductor Manufacturing Company Limited (TSMC) builds Nvidia’s Graphics Processing Units (GPUs). The restrictions could fuel China’s desire to invade Taiwan.

SEQUOIA CAPITAL

💵AI’s $200 Billion Question?

Image Source: Canva AI Image Generator

David Cahn, a partner at venture firm Sequoia Capital, outlined the glaring short-term mismatch between investments in GenAI and the technology’s revenue.

Key Details:

  • Nvidia’s stock price has more than doubled this year, increasing the company’s total market value to $2.593 trillion and signaling a continuing AI boom.

  • The technology sector has an insatiable demand for GPUs to fuel AI model training. Investors base their decisions on Nvidia’s success, leading to rapid investment in unclear applications.

  • Big Tech will likely commit upwards of $200 billion in CapEx for AI infrastructure this year.

  • CapEx refers to the funds a company uses to purchase, improve, or maintain long-term assets essential for its operations.

  • Microsoft expects to generate over $10 billion in AI revenue from products like Microsoft Copilot.

  • Assuming Apple, Alibaba, Bytedance, Google, Meta, OpenAI, Oracle, Tesla, and X also achieve generous gains from AI, “there’s a $125 billion hole that needs to be filled for each year of CapEx.”

  • So, how much of this CapEx is linked to current demand, and how much is projected based on future demand?

  • Also, what will these Big Tech companies use all this AI infrastructure for?

  • Big Tech needs to make sure they’re building AI infrastructure consumers actually want, not just guessing what might be popular later.

Why It’s Important:

  • Many venture firms believe the capital investment required to build and deploy GenAI has surpassed the revenue it generates in the short term.

  • GPU capacity is being overbuilt without a clear revenue stream to repay the upfront capital investment.

  • Perplexity CEO Aravind Srinivas believes AI will significantly improve existing business models by allowing companies to complete tasks with limited resources at dramatically lower costs, leading to indirect revenue.

  • AI investment is a long-term play. AI’s potential applications through AI-powered tools, autonomous vehicles, personalized education, drug discovery, and robotics will transform industries.

AI RESEARCH

🧫OpenAI Dissects GPT-4’s Inner Workings

OpenAI published a paper detailing a method for reverse engineering concepts within GPT-4 to extract human interruptible features that help explain the neural activity within language models.

We don’t really understand the inner workings of Neural Networks (NNs): an AI framework designed to mimic the human brain.

NNs aren’t designed directly. Instead, test engineers and research scientists design the algorithms that train them.

OpenAI’s “Scaling and Evaluating Sparse Autoencoders” method outlines a technique to identify patterns representing specific concepts within GPT-4.

An autoencoder is designed to efficiently compress (i.e., encode) input data to its essential features and then reconstruct (i.e., decode) the original input from this compressed representation.

Essentially, OpenAI is leveraging autoencoders to crack open the “Black Box” of language models to see what elements contribute to their outputs.

🛠TRENDING TOOLS

🗺MapsGPT finds exciting places near you.

🦾Replicate runs and fine-tunes open-source AI models.

🌮RecipeGenerator creates uniquely delicious recipes for you.

📂Riffo streamlines file organization with smart renaming and batch processing.

⛓ChainGPT is a personal expert in all cryptocurrency and blockchain-related topics.

🔮Browse our always Up-To-Date AI Tools Database.

💰VENTURE CAPITAL UPDATES

  • Greptile secures a $4M Seed Round to build an AI-fueled code base expert.

  • Bem closes a $3.7M Series C to automate unstructured data conversion for engineers.

  • Tektonic AI raises a $10M Seed Fund to build GenAI agents for automating business operations.

💼WHO’S HIRING?

  • WSP (Los Angeles, CA): Civil Engineering Intern, Fall 2024

  • Cohere (San Francisco, CA): Machine Learning (ML) Intern/Co-Op, Fall 2024

  • Adobe (New York, NY): Digital Academy Intern, Digital Strategy Analyst, Fall 2024

  • KPMG (San Francisco, CA): Audit Intern, Summer 2026

  • Meta (San Francisco, CA): Software Engineer, Computer Vision, Technical Leadership

🤖PROMPT OF THE DAY

MARKET RESEARCH

🛍Product Updates

Provide me with five features I could add to [Product] in [Industry] for [Target Audience] to enhance the value of my business.

Product = [Insert Here]

Industry = [Insert Here]

Target Audience = [Insert Here]

📒FINAL NOTE

If you found this useful, follow us on Twitter or provide honest feedback below. It helps us improve our content.

How was today’s newsletter?

❤️TAIP Review of the Day

“Keep up the engaging work!”

-Krishaan (⭐️⭐️⭐️⭐️⭐️Nailed it!)
REFER & EARN

🎉Your Friends Learn, You Earn!

You currently have 0 referrals, only 1 away from receiving 🎓3 Simple Steps to Turn ChatGPT Into an Instant Expert.

Refer 9 friends to enter 🎰June’s $200 Gift Card Giveaway.