• Unitree Robotics IPO
  • OpenAI Finally Does Something
  • J.D. Power US Tech Experience Index

Unitree Robotics IPO
I first learned of Unitree several years ago when I bought the Unitree PUMP weight training system that works like resistance bands in a box that can be programmatically altered for lower or higher resistance comparable to increases of poundage as you'd find with traditional weights. It's not a robot, but it showed enough promise for me to buy a basic system, and I still use it to this day. I hate that it requires an app to be used and I have to charge the boxes regularly to be able to make use of the resistance, but that feels like everything in the world lately. If it doesn't have an app, most likely with a subscription, does it even really exist?

But Unitree has moved far beyond simple weight training boxes and now they are a leading manufacturer of robots. They became the first humanoid robot maker to go public on a stock exchange this week as they debuted in China on the Shanghai Stock Exchange. Shares have surged 600% since launch. Unitree says they will invest the proceeds into new manufacturing facilities and increased research and development.

Robotics in China is a booming market and it's a key area where the United States has fallen embarrassingly behind. Unitree sells multiple humanoid and robotic dog models that can dance, box, walk, and with their on-board computers, they offer an embodied AI existence for a variety of LLMs and agents. Humanoid robots are used in China predominantly in manufacturing and warehousing, but robots are increasingly appearing in restaurants, hotels, and even hospitals where they can greet visitors and help process their requests. There is also a growing industry for companion robots that can be especially helpful for the elderly. The Chinese government is actively fostering the development and deployment of these robots as they acknowledge their aging population and the special needs they present.

We really don't have much in the United States in terms of a robotic presence, for better or for worse. And the FCC's recent decision to ban foreign made robots won't make us suddenly sprout dozens of new companies to fill the shoes of the Chinese manufacturers. The manufacturing skills and supply chain still live in and favor China. If there are benefits to be gained from the robotics industry, we are going to need to wait years before we can operate at a level where China already is. And if embodied AI is the next level of AI innovation, we may end up missing that boat.

OpenAI Finally Does Something
Recently, I've written quite a bit about the apparent negligent behavior of certain frontier AI companies when it comes to cybersecurity and testing safety. In a blog post from August 18, OpenAI announced that they are temporarily slowing their pace to allow more time for strengthening their testing protocols and environments. OpenAI specifically mentioned that they are working to "establish more evidence of alignment" before moving forward with more testing. According to OpenAI, strengthening their safeguards includes enhanced monitoring, alignment, and security measures. And improving their research environments includes stronger workload isolation (i.e. sandboxing), better network isolation, and conducting continuous security testing. OpenAI says they have removed potentially vulnerable services and standing privileges. This seems like an admission that they employed vulnerable services and managed privileges improperly, but ultimately, recent events exposed this, so there was no real reason to pretend otherwise. Perhaps most importantly, OpenAI says they are stepping up their chain-of-thought monitoring, which ostensibly means they will be paying closer attention to what the models are "thinking" as they work to solve problems. They have initiated a goal of responding to "concerning activity" within 30 minutes of it being observed.

OpenAI says the changes they are making will require 20% more computing power. How this additional expense may impact their profitability and IPO readiness is yet to be seen. They also indicate that they will publish a report of their learnings in "the coming weeks". Many people, including me, think it is preposterous and unacceptable that OpenAI didn't already have more safeguards and a better security framework in place before the recent escape exploits, but we can perhaps be hopeful that they will be more diligent in the future.

J.D. Power US Tech Experience Index
I was not aware that J.D. Power polled customers about their experience with technology in their cars, but it does make sense and gives me hope that someone is paying attention to what people actually want vs. what companies want to implement and attach subscriptions to. And it turns out, for most people, simplicity wins over complexity.

Some quick takeaways:

  • People like smart ignition and digital keys
  • Augmenting existing systems proves valuable, such as putting cameras in rearview mirrors or adding blind spot cameras
  • People aren't using passenger-specific screens
  • People prefer assistive driving features over hands-off automated driving

Some of this reflects my own personal experiences. I don't mind the smart cruise control in my car in theory; the idea of keeping my car in the lane and holding a steady following distance even when the speed of other cars around me changes can take a small mental load off my mind, but I hate that the system brakes because it cannot "see" that traffic is slowing ahead of me so that it could simply let off the gas pedal like most humans would. It feels like an attempt to extract more maintenance expense from me in the future when I need to replace my brakes sooner than I otherwise would have needed. I also like proximity unlocking and/or being able to unlock my car without having to specifically insert a key in the door. And having lived with push-button ignition for many years means I feel somewhat like a cavewoman when driving a car that requires turning the key in the ignition. When it comes to cars, humanity has always had full control of them (again for better or for worse) and there is a certain resistance we tend to have to releasing control to something that doesn't do things exactly the same way we would. And while having large screens is very helpful for navigation, they remain an excellent source of distraction and in cars like mine, not making them touchscreens in an effort to make them less distracting doesn't help. Where we once fiddled with the radio and the climate control buttons, now we fiddle with Google Maps, Spotify, YouTube, and countless other diversions designed to keep us entertained on our travels. Much like we are slowly learning to put our phones down to stop doom scrolling, we should probably learn to put our hands back on the wheel and spend some time just enjoying the ride and paying attention to where we are going, physically and metaphysically.

Here's hoping that car manufacturers pay more attention to what consumers really want and give us systems and features that help and not hinder.