TikTok Threats

Estimated Reading Time: 4 minutes

TL;DR

This week, we cover two trends. First, TikTok’s growing influence provides a direct lever for the Chinese Communist Party (CCP) to influence American society. Regulating TikTok is a crucially important task that has yet to gain traction in Washington. On a more cheerful note, we also review some interesting uses of AI for chip design emerging from Nvidia. We end with a new job posting on our job board from Modern Electron.

The Dangerous and Growing Influence of TikTok

__ @NaithanJones

TikTok is eating passive screen time every single streaming company is going to have to compete against this. It has network effects, where they don’t even have networks, content is endless and cheap and ever expanding, it’s free, and it requires less commitment to consume

The growth of TikTok usage among the American public is a concerning trend that continues to slip under the threat radar. Social networks hold immense powers; Facebook and Twitter have reshaped America’s political landscape over the last decade. TikTok looks to be developing similar power, but is not an American company (TikTok is a subsidiary of ByteDance). The CCP has started using company committees to exert increasing control over China’s large tech companies (cite). For example, it would be extremely hard to detect if the CCP were to issue a directive to ByteDance to alter the TikTok algorithm to favor one US political candidate over another or to worsen political dysfunction within the US. The CCP has not been shy about tightening the screws on its tech giants (see Jack Ma). Allowing the CCP to gain a direct lever to influence American politics is a dangerous risk to our democracy.

AI at Nvidia Shows Results in Chip Design

Nvidia has made considerable progress in using AI models to accelerate its internal chip design efforts. In a recent talk, Bill Dally, Nvidia’s chief scientist, explained how deep learning methods have been used to model the complex physics of GPUs. Tasks such modeling voltage drops, power estimation, parasitic effects, chip placements and more can be performed using graph neural networks. Graph neural networks and reinforcement learning are emerging as foundational technologies for AI-driven chip design.

Bill Dally, Nvidia’s chief scientist explains how convolutional networks have been used to map voltage drops on chips. Deep models are able to achieve high accuracy with inference time in seconds as opposed to hours long classical simulation runs (source).

A higher resolution image of the MAVIREC architecture used to model voltage drops (https://arxiv.org/abs/2012.10597).

GRANITE uses graph neural networks to perform accurate power estimation. Netlists (a description of the connectivity structure of a chip) are converted into graph topologies used to structure the graph neural network (source).

Nvidia’s ParaGraph architecture converts a circuit schematic to a graph neural network topology (source).

Bill Dally explains how chip routing congestion can be modeled with graph networks (source)

Reinforcement learning can be used to guide placement of transistors when designing a new standard cell (source).

Interesting Links and Pictures

• https://www.math.columbia.edu/~woit/wordpress/?p=12775: A post on Peter Woit’s blog is critical of a review article covering Mochizuki’s proposed proof of the ABC conjecture without mentioning Scholze’s critique.

Job Board

These postings are for companies we personally find interesting (we are not currently accepting paid ads for posting). If you would like your company’s jobs to be featured, send us a brief blurb for us to review.

• Modern Electron is a sustainable heat and power technology company that helps people save money, reduce carbon emissions, and increase resiliency during power outages. We've raised $70M of venture financing, including backing from Bill Gates.  Our first technology transforms boilers’ and furnaces’ heat into power for use onsite, reducing wasted energy. The company’s next-generation technology decarbonizes further by converting gas to clean hydrogen for use in heating buildings and industry without CO2 emissions.  Heat is 50% of final energy demand, so this is critical to help us meet our climate goals.  Link to Open Roles.

Feedback and Comments

Please feel free to email me directly (bharath@deepforestsci.com) with your feedback and comments! 

About

Deep Into the Forest is a newsletter by Deep Forest Sciences, Inc. We’re a deep tech R&D company building an AI-powered scientific discovery engine. Deep Forest Sciences leads the development of the open source DeepChem ecosystem. Partner with us to apply our foundational AI technologies to hard real-world problems. Get in touch with us at partnerships@deepforestsci.com! 

Credits

Author: Bharath Ramsundar, Ph.D.

Editor: Sandya Subramanian, Ph.D.