Let me use a baseball expression for a second. You’re probably familiar with the idea of the “three true outcomes” if you watch the sport. They are a home run, a strikeout, and a walk, exactly. The commonality among the three is that, in most cases, the defence is not in charge of them.
Of course, there are ambiguities, as there always is when trying to define anything as an absolute. The concept’s worth has also long been debated in the never-ending cold war of baseball analysis. Everything is alright since my main goal in using the word here is to co-opt it.
The three real outcomes for a robotics startup would be, roughly speaking:
- Becoming public
- Getting bought
- Dying
Similar to the world of baseball, there is a lot of room for manoeuvring in this situation. Particularly in robotics, DARPA funds may be used indefinitely to support a perfectly successful business. It’s possible to execute any combination of the aforementioned three, unlike baseball.
But the main point I want to address here is: What’s the best result for a robotics startup? Of course, nobody wants option number three. However, it is a very real — and frustrating — possibility, much like being out at bat. And as we’ve seen, even a massive infusion of venture capital money cannot completely preclude startup failure, particularly in the case of robotics where the entrance barrier is so very high. Furthermore, given the overall market tendencies, a little market correction in the robotics sector is warranted.
Even during the (now-defunct) heyday of SPACs, IPOing has been a very uncommon result for robotics firms. Some scheduled SPACs were temporarily suspended due to the status of the market in an effort to ride more positive trends. Sincere, I believe that result number two is very realistic and often the optimum one for many businesses. Long runways and a lot of resources, which a large firm can provide, are needed for robotics.
However, fit is when you start to run into issues. Conversations when the prospective acquirer and acquiree have radically divergent ideas must occur often, in my opinion. Of fact, these poor fits do occur sometimes. The corporation may not have understood the market fit or the resources needed to sustain a robotics company, or maybe they just had drastically divergent ideas about what their robots could and couldn’t achieve. For every Amazon that purchases Kiva, there are several Googles who purchase Boston Dynamics
The subsequent purchase of the latter company by Hyundai raised certain concerns. Although I will admit that this week’s introduction of the Boston Dynamics AI Institute is a fascinating — and exciting — twist to this tale, a vehicle firm isn’t the most obvious match for what Boston Dynamics does. The corporation has always placed a strong emphasis on research, and the $400 million investment in the new centre gives it a lot more runway and resources. That is far more than Ford just spent in its own facility at the University of Michigan.
The institution will be led by Marc Raibert, the creator and former CEO of BD, which is the most exciting aspect of all. In a press release related to the development, he said, “Our aim is to construct future generations of sophisticated robots and intelligent machines that are smarter, more agile, perceptive and safer than anything that exists now.” We will be able to create robots that are simpler to use, more productive, capable of performing a wider variety of tasks, and safer working with humans thanks to the Institute’s unique structure, which combines top talent focused on fundamental solutions with sustained funding and excellent technical support.
It’s worth checking in on Google’s efforts in the sector after the company botched the purchase (and a number of others at the same time, under Andy Rubin’s direction). Alphabet X grads have generally been the focus of my coverage of the area. The most well-known (so far) is the drone delivery service Wing, but Intrinsic, a robotics software company, is beginning to do some fascinating stuff.
The lab’s Smarty Pants, a promising soft robotic exoskeleton, received some column space from us last year. The lab also provided a sneak peek at Project Mineral in March, an autonomous rover designed to gather agricultural data. It aims to phenotype plants specifically. The business claims:
Most scientist’s phenotype plants nowadays by meticulously walking across fields while noting various plant characteristics with a notebook, pen, and ruler. But consider attempting to estimate the number of beans contained in a bean pod, the length of the leaves, or the number of blooming flowers. Now picture doing it by hand each week for hundreds of plants in the sweltering summertime. The bottleneck in phenotyping is this.
Mineral has been providing the Alliance’s researchers with tools to let them carry out further studies and identify more agricultural characteristics in order to assist them with this difficulty. Since last year, Mineral’s rovers, which the local team has dubbed “Don Roverto,” have been leisurely rolling through the test fields outside Future Seeds, photographing every bean plant and using machine learning to identify traits like leaf count, leaf area, leaf colour, flower count, plant count, and pod dimensions. The rover constantly does this for each plant in the field and keeps track of each one’s location so that it may return a week later and provide an update on the health of each plant.
A little envious that Haje was able to visit Google’s internal robotics team this week. He wrote about the encounter, which included collaboration with another alumni of X. He clarifies:
Speed and accuracy are one thing, but Google’s robotic laboratories are really working on the interface between robots and human language. It is achieving some remarkable advancements in its capacity to interpret natural language as it may be used by humans. You may ask a person, “When you have a minute, could you fetch me a drink from the counter?” It’s a relatively simple request. That phrase, however, condenses a great deal of information and comprehension into what seems to be a simple query to a computer. Let’s break that down: “When you have a minute” can be merely a figure of speech or it might be a request for the robot to complete what it is doing. Could you fetch me a drink? The “proper” response can only be “yes,” if a robot is being too literal. It can get a drink, and it affirms that it can. But you, the user, didn’t specifically instruct the robot to do that task. If we’re being very picky, you didn’t specifically request that the robot bring you a drink.
Overall, I believe there is a case to be made for building robots and AI companies inside, however few businesses have the resources of an Alphabet/Google. We still have a long way to go before we can understand how such endeavours could genuinely pay off, even with Google’s resources (time, money, and patience).
Xiaomi’s attempts, however, are very dubious. The company’s robotics efforts so far resemble Samsung more. I don’t have many reasons to think its work is currently more than show, save from some success with robotic vacuums. This includes the Spot-like CyberDog from a year ago and the brand-new humanoid robot CyberOne, which made its premiere with several phones. It’s easy to understand why the robot is being likened to Tesla’s as-yet-unseen efforts from the perspective of design. Additionally, it provides a more…realistic notion of what to expect from a robot with two legs.
I’ll leave you with some fundraising news from an intriguing business before I go for the week. YC-backed Mobot just closed a $12.5 million Series A round. The business makes robots that developers can use to test their applications for bugs.
In order to automate testing for mobile applications, there are solutions created by businesses like Applitools, Test.ai, and others that use current emulation testing frameworks. Unfortunately, software-based, mimicked testing often allows numerous flaws to slip through the cracks since it is not a true representation of testing on actual hardware, according to Eden Full Goh, the company’s founder, in an interview with TechCrunch. At this time, Mobot does not see itself as a rival to emulators or an alternative to automated testing. Instead, we want to replace the manual quality assurance that everyone must still do now and will inevitably need to do more of in the next five to ten years as device fragmentation increases.
In the meanwhile, I got an exclusive from CleanRobotics, the Colorado-based company that created the robotic garbage can TrashBot that sorts rubbish. To scale up a robot intended to enhance recycling sorting at the source, the business obtained a $4.5 million Series A round of funding.
According to CEO Charles Yhap, “Recycling guidelines are unclear, and customers are often so perplexed that their recycling accuracy is below chance, resulting in severely polluted recyclables that no one is purchasing.” Our technology increases the amount of garbage that is diverted from landfills, increasing the amount of recyclable materials.
