• Will you be surprised when a robot shows up on your doorstep?
  • A CBDC for 1.4 billion people…
  • China’s robotaxis are nearing human efficiency…

Dear Reader,

Watching events unfold in China over the last few days stands in stark contrast to the Thanksgiving holiday weekend, and the excitement of the World Cup happening in Qatar.

The spark for what has now become widespread protest began in an apartment building in Urumqi, a city in the most northwest territory of China, Xinjiang.

It’s believed that more than 20 people burned to death after a fire broke out in the apartment building. Due to the draconian COVID-19 policies in use, people were literally locked inside their apartments and unable to flee.

Apartment Fire in Urumqi

Source: Twitter

Every fire is a tragedy, especially those that take lives. But this one was entirely avoidable, and all knew it. They also knew that it could have happened anywhere in China, where citizens are being physically locked into their homes for the purpose of quarantine.

After nearly three years of ineffective “zero COVID” policies, the people of China are finally fighting back. What happened in Urumqi has led to protests in Beijing, Shanghai, Wuhan, Guangzhou, Zhengzhou, and other locations.

These are not all peaceful protests, either. 

COVID-19 test booths are being toppled, gates are being torn down, and the citizens are fighting back. No uprising of this scale has taken place in mainland China since Tiananmen Square back in 1989. Hopefully, the end result won’t be the same.

The reality is that the people are angry. They’ve had enough. The human mind thirsts for freedom of movement and freedom of thought.

And after three years, even China knows the cold, hard facts. For anyone younger than 75 years of age, the chance of survival when infected by COVID-19 is more than 99.9%. For those in good health, it’s even higher. We now know that infection fatality rates are roughly on par with an influenza.

Yet China continues to pursue its zero-COVID policies. Every other country that tried has given up. Taiwan, Australia, and New Zealand have all thrown in the towel. These countries weren’t models of good policy; they were examples of exactly what not to do.

Their draconian COVID polices only delayed the inevitable. All saw the same massive spikes in COVID-19 infections and deaths as the virus became more infectious. Nothing stopped the spread – including the experimental COVID-19 drugs (incorrectly called “vaccines”).

And yet China appears to be tripling down on its policies. 

Below is an example of massive quarantine camps that continue to be built throughout China. This one can detain 48,000 people. Can you imagine being locked up in these tiny pods for weeks on end? Separated from friends and family. For no purpose at all.

The reality is that these policies aren’t about COVID, they’re about control.

A Quarantine Camp in China

Source: Twitter

Even coverage of the World Cup in Qatar has been carefully edited by China’s state broadcaster. Shots of the enthusiastic – and maskless – fans cheering for their teams are being edited and replaced with closeup shots of players, coaches, or other images.

“They” don’t want Chinese citizens to see that the world has moved on from COVID. For some, we’d rather just forget about it and move on. But for many others, we can’t let “them” forget, as we’re forced to deal with the aftermath of what was forced upon us. 

About the worst thing that can happen to the West from all this is more supply chain delays from protests and chaos. Already, Apple’s contract manufacturer for its iPhones – Foxconn – has had major disruption at its Zhengzhou manufacturing location that will result in a shortage of six million iPhones this year.

Who cares about the shortage of iPhones? I certainly don’t. I’m happy to wait. 

I’m much more concerned about the repressive working conditions at Foxconn forced upon employees due to these COVID-19 restrictions…. something that Apple would like us to think it cares deeply about.

I’m even more concerned about a repeat of the Tiananmen Square massacre. The outbreak of courage from the Chinese people right now is awesome, and a sight to behold. Many are publicly calling for President Xi Jinping to step down – something that was previously unthinkable.

Put anything under enough pressure for a long enough period of time, and cracks will form. It’s happening right now. And the world is watching.

An AI breakthrough in imitation learning…

The race this year between artificial intelligence (AI) giants like DeepMind, Nvidia, Tesla, Google AI, Meta, and OpenAI has been something to watch. Every few weeks has brought a major breakthrough with applications in this incredible technology.

This time, it’s a breakthrough in something called imitation learning from OpenAI. Imitation learning is exactly what it sounds like: an AI that learns by imitating.

OpenAI chose to develop an AI to play the massively successful game Minecraft. It was a good target, with more than 100 million users and seemingly endless amounts of game play available on YouTube.

Vast Collection of Minecraft Videos

Source: technologyreview.com

This is important, as videos are a useful tool for an AI to learn from – much in the same way that humans learn from watching videos on YouTube.

The team at OpenAI first hired humans to play and record 2,000 hours of Minecraft while annotating their keystrokes alongside the video recording. Then, the team used another AI – a neural network – to “watch” and annotate 70,000 hours of Minecraft video collected from YouTube.

Once the team had their annotated video training set, they developed their Minecraft AI, which is now the most advanced AI that can play Minecraft.

Minecraft is a unique challenge because, unlike most applications of AI that have a clear desired end goal, Minecraft is an open-ended game. This makes for a larger AI challenge.

But the implications of this kind of imitation learning are incredible and will become very relevant next year.

When I think of a wide range of general tasks where a robotic arm or dexterous humanoid robot could be useful, imitation learning makes perfect sense.

Whether at home or a workplace, imagine being able to show an AI-powered robot what task(s) need to be performed. An AI could learn by watching. Then it could supplement that learning by going to the internet and finding more learning material, and soon master any given task(s).

That’s what makes OpenAI’s announcement so incredible. It’s the beginning of robots being able to learn by watching, rather than through weeks of programming and testing. 

And with Tesla’s Optimus AI-powered robot under development – I’m predicting that we’ll see a pre-commercial model by the fourth quarter of 2023 – the timing for breakthroughs in this kind of AI couldn’t be better.

Imagine being able to show a robot how to go shopping at the supermarket to prepare goods for home delivery. Or training UPS, FedEx, and Amazon delivery robots how to take packages from trucks to consumers’ doorsteps.

It won’t be long before we see humanoid robots working around us and with us every day.

This CBDC pilot will likely lead to something more ominous…

The increase in announcements around central bank digital currencies (CBDCs) is clearly a sign of what’s coming. It’s also something that we should all be deeply concerned about.

This has been a big month in India, as it’s been testing its own CBDC – a digital rupee – for wholesale applications. 

It started off using the digital rupee for secondary transactions of government securities. Then, it teamed up with nine banks – eight of which are Indian – to test bank-to-bank transactions, as well as cross-border transactions.

But what comes next is telling. Next month, India will begin its retail pilot of its CBDC. 

The pilot will start off with the eight banks, and each bank will involve between 10,000 and 50,000 retail participants. And eventually, the “pilot” will involve all commercial banks in India.

As with most implementations of CBDCs, both users and merchants will download a digital wallet to store the digital rupees. Consumers will be able to “download” digital rupees much in the way that we get cash out of our accounts from an ATM today. No trip to an ATM required.

And while the technology is fantastic and convenient, the end result will sadly be a surveillance state where the government will have the ability to see and tax every transaction – and worse, control spending based on social credit scores that further perpetuate more irresponsible monetary policies.

Baidu is on the verge of scaling a very profitable business…

China’s equivalent of Google – Baidu – has been making even more progress on autonomous ride-hailing services than Google’s self-driving division, Waymo.

Baidu’s third-quarter earnings call revealed some impressive numbers from Baidu’s autonomous ride-hailing service, Apollo Go.

Apollo Go Vehicles in Beijing

Source: Visual China Group

In Beijing, Shanghai, and Guangzhou, Baidu’s autonomous taxis completed more than 15 rides a day, on average. This is a significant number, as it’s similar to a human driver working for a ride-hailing service on a regular shift.

In addition to the operational efficiency, Baidu’s autonomous rides are up by more than 300% year over year, and it’s completed more than 474,000 autonomous rides.

These are impressive numbers at scale. And Baidu now has approval to begin testing the service without safety drivers in the front seat. That’s the final step toward making a profitable ride-hailing service: removing the human driver.

But Baidu’s largest weakness is in its software. Baidu’s AI has been pre-programmed to work in geofenced areas, much in the way that Waymo has been designed to do. It’s remarkable how similar the two companies’ approaches have been.

This results in highly functional and safe driving within a predetermined service area. Generally, that makes sense for ride-hailing services that tend to be targeted at major metropolitan areas.

But this is clearly a shortcoming, as the software isn’t designed to navigate streets that it might not be familiar with. This is what Tesla’s full self-driving (FSD) software is designed to do. It doesn’t need to be pre-programmed with highly detailed maps and a pre-defined service area. 

That’s the grand challenge for autonomous driving, and Tesla is the only company that’s figured this out.


Jeff Brown
Editor, The Bleeding Edge