They’re programmed to work hard and play hard – TechCrunch


Industrial robotics are large and heavy, and in some cases, legitimately dangerous. They are also extremely difficult to train, especially if you plan to implement them for tasks outside of your specifically designed intentions.

There’s a great opportunity for the right AI / software company to step in and help make bulky systems intended for things like car manufacturing easier to program and more versatile. Honestly, there is probably enough room to support several companies in the category as robots become an increasingly essential part of the way we do business.

This week we saw a couple of important news from companies that operate in that space. On Tuesday, Covariant announced an $ 80 million raise, a quick follow-up to the $ 40 million Series B it announced in May 2020.

Image credits: Covariant

I spoke to President, Chief Scientist, and Co-founder (and recurring guest of TC Sessions: Robotics) Pieter Abbeel for the piece, which you can check here. Also, I chose the UC Berkeley professor’s brain for a long time over some broader robotic trends.

We have seen a marked increase in investment activity in robotics and automation since the beginning of the pandemic. Do you anticipate that this interest will continue?

It will not only hold. It will continue to accelerate on a dramatic scale. The demand is not new, but the pandemic has certainly increased the demand for rugged and robust robotics. COVID-19 sped up a timeline that was already in motion. Other factors contributing to the momentum include the rise of e-commerce replacing in-store purchases along with Amazon’s push for efficiency. Consumers’ expectations for fast delivery have risen across the board and delivering on that promise often starts with warehouse automation.

As someone with experience in both an educational setting and a startup, how has universities’ approach to incubating businesses evolved? What else can and should be done to encourage entrepreneurship?

With AI, the transition from research to practice has been exceptionally fast. An idea could be published today and many companies could implement it in their systems the next day. This trend has positioned AI researchers in a unique position to create new apps (compare it to, say, Airbnb, Uber, food delivery companies, etc., which were not enabled by research advancements, but by everyone. those with a smartphone, allowing a new model). doing business).

Structurally, a clear change in many universities is the introduction of artificial intelligence in many programs. A great example is the course “The Business of AI,” which I co-teach at Haas Business School in Berkeley, and which provides business students with a solid understanding of the role of artificial intelligence today, as well as trends and trends. what the future might bring. .

To foster greater entrepreneurship in the US, leadership needs to consider how many international students are also top AI researchers. A faster visa / green card process for entrepreneurs would have a very high impact.

Do you plan to continue teaching as Covariant grows?

Yes. I see a very strong synergy between being at the forefront of academic AI research at Berkeley and being at the forefront of industrial R&D bringing AI robotics to the real world as Covariant’s chief scientist. The culture that our CEO Peter Chen has fostered at Covariant is also very much aligned with this; Curiosity and lifelong learning are core values ​​at Covariant.

How actively does your team consider biases in your AI work?

Bias in AI systems is of course a broader industry problem and is on the minds of our team members. To this day, bias in artificial intelligence systems does not play a direct role in our current robotic storage efforts. However, quality assurance in general is fundamental to everything we do, and quality assurance is not a one-axis thing, we have to consider the quality and coverage of various data sources and the performance in SKUs, warehouses, customers, etc. In that sense, there are actually a lot of technical parallels.

It appears that most of the activity on the industrial robotics front is happening on the software / AI side. Are robotics manufacturers continuing to evolve their hardware as software improves?

In fact, even though we are primarily focused on software / AI, we work with incredible partners to deliver fully functional robotic systems. In doing so, we also see continual improvement in hardware. Most visible in a short period of time are the continuous changes in the tools at the end of the arm. Additionally, we see exciting multi-year roadmap ideas in robotic arm form factors that require more R&D and design effort to bring to market.

Image credits: Research by Gramazio Kohler, ETH Zurich

The other big news of the week is the intrinsic disclosure, Alphabet’s latest robotics game. Or, I guess I should say, the most recently announced robotics game. The Alphabet X spinout has apparently been in the works for about five years. It follows a rather uneven robotics track record for Alphabet / Google that involved a brief ownership of Boston Dynamics. But the company’s offering seems to be much more in line with what Google highlights.

Here’s Intrinsic CEO Wendy Tan-White, who most recently served as Vice President of Alphabet’s Moonshots:

For the past several years, our team has been exploring how to give industrial robots the ability to automatically detect, learn, and adjust as they complete tasks, so that they work in a wider range of configurations and applications. Working collaboratively with teams at Alphabet and with our partners in real-world manufacturing environments, we have been testing software that uses techniques such as automated perception, deep learning, reinforcement learning, motion planning, simulation, and force control.

Image credits: Agility

Closing the summary of the week with a couple of ‘athletic bots’. First is Cassie’s return, Biped Robot from Oregon State University. Cassie was a bit in the background to Agility’s OSU-derived delivery robot Digit, but the school is still doing cool things with the rig. A research team helped teach the robot to run, using a deep reinforcement learning algorithm.

In fact, Cassie managed to run 5 kilometers in 53 minutes. Not great by human standards, but extremely robust for a robot that uses a single battery, particularly when taking into account the 6.5 minutes of troubleshooting an overheated computer and a poorly maneuvered turn.

Outside of the Olympians and t-shirt sellers, Toyota may have been the most disappointed in the initial decision to delay the Summer Olympics. The auto giant clearly envisioned the Tokyo games as an ideal opportunity to showcase its technology to the world.

Now that the games started, the company’s basketball robot CUE is back in a big way. After debuting in 2018, CUE again sank triples at halftime in the USA-France game.


feedproxy.google.com

Leave a Reply

Your email address will not be published. Required fields are marked *