The Underground Network Sneaking Nvidia Chips Into China

Pivot 5: 5 stories. 5 minutes a day. 5 days a week.

1. The Underground Network Sneaking Nvidia Chips Into China

A network of buyers, sellers, and couriers is bypassing U.S. export controls on Nvidia's AI processors, despite U.S. restrictions. Over 70 distributors advertise these chips online, with many promising delivery in weeks and some even selling entire servers.

A 26-year-old Chinese student smuggled six Nvidia AI processors from Singapore to China last fall, claiming the activity is part of a broader operation to evade U.S. government restrictions. Nvidia's CUDA framework is popular in China's AI space, with many verified sellers offering preorders and deliveries within weeks.

Read the full story here

2. This HR company tried to treat AI bots like people — it didn’t go over well

Lattice, an AI company, announced in a blog post that it would be the first to provide official employee records for digital workers. The digital workers would be securely onboarded, trained, and assigned goals, performance metrics, and systems access. However, after a backlash, Lattice announced it would not further pursue digital workers in its product.

Some HR professionals and employees have expressed concerns about the integration of AI workers into human management processes. Lattice's attempt to respond to AI bots has backfired, especially among those who care most about it.

Read the full story here

3. MUSCLE: A Model Update Strategy for Compatible LLM Evolution

LLMs are frequently updated to improve performance, often focusing on increasing overall performance metrics without ensuring compatibility with previous models. This can lead to user dissatisfaction as users build mental models of the machine learning model they interact with.

The study aims to provide seamless model updates by providing evaluation metrics for compatibility with prior models, and proposing a training strategy to minimize inconsistencies in model updates. This can reduce negative flips by up to 40% from Llama 1 to Llama 2.

Read the full story here 

4. The AI-focused COPIED Act would make removing digital watermarks illegal

A bipartisan group of senators has introduced the Content Origin Protection and Integrity from Edited and Deepfaked Media Act (COPIED Act), aimed at ensuring the authenticity and detection of artificial intelligence-generated content. The bill would direct the National Institute of Standards and Technology to create guidelines to prove content origin and detect synthetic content, such as watermarking.

It also requires AI tools for creative or journalistic content to allow users to attach information about their origin and prohibit it from being removed. The bill also allows content owners to sue companies they believe used their materials without permission or tampered with authentication markers.

Read the full story here

5. Inside Google's augmented reality reset

Google's AI boom is perfectly timed for an AR revival, as the company consolidates efforts to move faster and spark partnerships for head-worn devices. The company has made several false starts in Extended Reality, but knows these wearables are a logical home for the AI features it is developing.

Google teams are heavily focused on Android XR, a new software that Google hopes hardware partners will want to build devices on. Google and Samsung partnered on a headset powered by Google's new software, but leaders have continually pushed it back, with the launch delayed to Q1 of next year. Tech giants are now setting the stage for an arms race as they build AR platforms and lure partners onto them.

Read the full story here

Advertise with Pivot 5 to reach influential minds & elevate your brand

Get your brand in front of 65,000+ businesses and professionals who rely on Pivot 5 for daily AI updates. Book future ad spots here.