By Jeff Brown, Editor, The Bleeding Edge After Tesla’s October 10th We, Robot event, all of the major players in AI-powered robotics have been pushing out news and updates about their progress. Despite a range of negative comments from the media about Tesla’s Optimus humanoid robot – along the lines of… Elon Musk’s Optimus bot stole the show at Tesla’s robotaxi unveil – but the AI was all smoke and mirrors… – Fortune Tesla’s big ‘We, Robot’ event criticized for ‘parlor tricks’ and vague timelines for robots, Cybercab, Robovan… – VentureBeat Tesla Optimus humanoid robot needed human help at AI event… – Axios … investor interest hasn’t been dampened. Quite the opposite. It’s through the roof. More Than Just a “Point Release” All of the criticisms about Optimus not being fully autonomous yet completely miss the point. After all, Teslas have been driving themselves for years. It all started with simpler tasks – like driving on a highway, staying in a single lane, and maintaining a safe distance from any vehicle in front. In the beginning, progress was slow but steady. With each software release of Tesla’s Autopilot – and eventually full self-driving software (FSD) – we saw material improvements in the technology. But then everything changed in May 2023. That’s when Tesla released a version of its FSD that was completely rearchitected – it now uses neural networks to control the car. That was the inflection point. That’s when we saw the hockey stick up and to the right, showing us exponential growth. I was enjoying my time “out in the wild” as it was happening, writing daily about the incredible discoveries I was making at the outer limits of tech. And because I tested my own Tesla with each and every software release, I could experience the change. I wrote about it my e-letter at the time, Outer Limits – AGI – Money is No Object. Here’s the importance of what was done: Just think… [Each] Tesla has 8 mounted cameras, including the cabin cam. These cameras don’t just “stare” at asphalt all day, logging mileage. It’s estimated that each camera has a depth-of-field of up to 250 meters and 360 degrees of visibility. [Teslas] use their seven exterior cameras to “perceive” millions of objects every day, training the deep neural networks on the data. And they have been doing this for years. Using a neural network has enabled Tesla to create something like a synthetic visual cortex. And with the improved computational power, Tesla’s AI was able to ingest high-resolution video – not just images – from cameras located around the car, process that information in a similar way as a human brain, and infer the correct course of action in a way that optimizes safety and achieves the end goal of getting to a destination. Release 11.4.1 was the beginning, the turning point, for what Tesla’s full self-driving (FSD) software has become today. This is what the mainstream critics are all missing. Tesla’s FSD isn’t just for its EVs… After all, what’s a self-driving car other than an autonomous robot on wheels? And it’s not about what these humanoid robots are capable of today. It’s about the rate of technological advancement that matters. It has gone exponential now. We’ve already seen this unfold with Tesla’s FSD software. Musk recently said that version 13 (v13), which will be fully autonomous, will be out before Thanksgiving in the U.S. Level 5 autonomy – something that hasn’t been achieved yet. I can’t wait to enjoy the rides. And yes, I’m going to be very thankful. Smart Money Turns to Robots The point today is that this exponential progress in autonomous vehicles translates directly to what’s happening in the humanoid robot industry. Many institutional investors understand this now. They took the time to have Teslas drive them around. They witnessed the progress of humanoid robots over the last two years. They now use large language models (LLMs) daily for their work. And they want investment exposure. They want to deploy capital into promising robotics companies… Robotics companies raised an impressive $3.7 billion just in September this year. In fact, $14.6 billion has already been invested from January through September. Despite all the media criticisms, the truth is that any legitimate robotics team will get funding now. It’s a multitrillion-dollar opportunity and a greenfield opportunity – open for the taking. This is why robotics companies are stepping up their efforts. They don’t want to get beaten by Tesla, and they also want to position themselves for further funding to move faster. At this stage, it has become a sprint. Those knowledgeable in the space can sense that full autonomy is within reach. And everyone knows that the utility of fully autonomous robots is nearly endless. The cost of the robots will be reasonable – in the tens of thousands of dollars – and that will be amortized over a number of years. And the operational costs will drive adoption – that is the cost of electricity to fuel them and any maintenance costs. That’s it. Boston Dynamics and its owner, Hyundai Motor Group, understand this. Catching Up With Atlas Hyundai, a massive South Korean industrial conglomerate that includes the production of cars and trucks, acquired 80% of Boston Dynamics in 2021 from Softbank, valuing the business at $1.1 billion at the time. The company just released some footage of its new-and-improved, all-electric Atlas humanoid robot. It’s performing a task in what appears to be an automotive factory. Atlas Moving Engine Covers | Source: Boston Dynamics If I had to guess, I’d say that this work is being done at a Hyundai Motor location. If we look at the bottom of the supplier container on the left, it says “Return to SCA.” SCA is an automotive manufacturer that specializes in light-duty trucks. The company was acquired by Fox Factory (FOXF) in 2020. Automotive suppliers like SCA ship car parts in containers to factories where cars/trucks are assembled. Those parts have to be moved to mobile containers (or “dollies”) and be brought to the correct location in the assembly line. In this case, SCA supplied the engine covers, which is what Atlas is moving from the supplier container to the mobile container, ultimately bound for assembly. This is exactly the kind of time-consuming, physically tiring, monotonous task that is well-suited for humanoid robots. Ironically, this is a very similar demo to what Figure AI showed this July, seen below. Figure AI Assembling Car Parts | Source: Figure AI The Boston Dynamics Atlas demo is not insignificant. Atlas’ human “input,” its instruction set, is simply a list of bin locations to take the engine covers from and bring to. Atlas is programmed with a machine learning vision model to “see” the individual bins. It also can understand – in real time – the position of the object it is carrying and can manipulate it to achieve its instructed task. We can see this in action in the short clip below. Atlas Autonomous Course Correction | Source: Boston Dynamics Atlas accidentally catches the top of the engine cover on the bin, which stops Atlas from completing its task. Atlas then autonomously pulls the engine cover back, realigns, and completes the task correctly. Atlas Takes a Jab at Optimus I did get a chuckle out of the “Fully Autonomous” notation in the lower right part of the Atlas video. It is, in part, a jab at Tesla… because of the criticisms that Tesla received for having some functions at the We, Robot event supported by human teleoperators, suggesting that Optimus is not fully autonomous. These kinds of industry jabs are pretty normal. And they’re also silly. Optimus was wandering and mingling in crowds in an unstructured environment at We, Robot – in the same way that a Tesla EV now does daily while out driving. And the goal of the event, let’s remember, wasn’t to show full autonomy for the robots. It was to show a vision of the future. Not to mention that the event was about the reveal of the Cybercab and Robovan – Optimus had no major announcements at the event. The reality is that Tesla’s Optimus overtook Boston Dynamics in less than two years, despite Boston Dynamics having nearly a three-decade lead. Boston Dynamics was founded in 1992, and I started tracking the company closely in the late 2000s. This business opportunity will be measured in billions of units. It may have taken institutional investors a while to understand how quickly the technology is improving, and the size of the market. But they see it now. And it's right around the corner. Regards,
Jeff | 