AI & Robotics News and Discussions

[Yuli Ban]

Amazon Patented a New Delivery System That Could Have Your Block Crawling With Robots

Amazon’s taking over the world. Or, at least, the US. Even as Jeff Bezos steps down as CEO and hands the reins over to former AWS head Andy Jassy (and Bezos himself gets ready to go to space), the e-commerce giant is projected to account for more than 40 percent of all online sales in the country by the end of this year.
Not surprisingly, the company is continuously exploring new ways to streamline its operations and cut costs. Last week a patent application that Amazon filed earlier this year became public. It’s for a delivery system that, if it comes to pass, would bring a new technology to neighborhoods where Amazon is popular (which, as implied above, is most of them).
What if, instead of the delivery people in Amazon uniforms you see frantically dashing from one home to the next, their trucks double-parked with the blinkers on, you saw this: an Amazon truck pulls up near your home and, most likely, still double-parks and puts the blinkers on (I mean, some things don’t change, and traffic congestion in cities isn’t going away anytime soon). But rather than the driver’s door opening and a person jumping out to start getting boxes where they need to be, the rear of the truck opens and a ramp is lowered to the pavement.
Down that ramp roll several smaller vehicles, maybe similar to the Kiwi delivery robots seen ferrying food to students in Berkeley a couple years ago. These small vehicles disperse, each one making its way to a different house or building within a one-to-two-block radius. They roll up to peoples’ doorsteps, deposit a package, then faithfully return to the mothership.

[Yuli Ban]

Advanced New Artificial Intelligence Software Can Compute Protein Structures in 10 Minutes

Accurate protein structure prediction now accessible to all.

Scientists have waited months for access to highly accurate protein structure prediction since DeepMind presented remarkable progress in this area at the 2020 Critical Assessment of Structure Prediction, or CASP14, conference. The wait is now over.

Researchers at the Institute for Protein Design at the University of Washington School of Medicine in Seattle have largely recreated the performance achieved by DeepMind on this important task. These results were published online by the journal Science on July 15, 2021.

Unlike DeepMind, the UW Medicine team’s method, which they dubbed RoseTTAFold, is freely available. Scientists from around the world are now using it to build protein models to accelerate their own research. Since July, the program has been downloaded from GitHub by over 140 independent research teams.

Proteins consist of strings of amino acids that fold up into intricate microscopic shapes. These unique shapes in turn give rise to nearly every chemical process inside living organisms. By better understanding protein shapes, scientists can speed up the development of new treatments for cancer, COVID-19, and thousands of other health disorders.

[Yuli Ban]

AI Is Learning To Understand How Vegetables Taste

With the global demand for food escalating, vertical farms are becoming a critical component of agriculture's future. They use robotics, machine learning and artificial intelligence (AI) to automate farming and perfect the growing of greens and vegetables. With steady growth, the vertical farming market was had an estimated value of $4.4 billion in 2019 and is expected to reach $15.7 billion by 2025.

[raklian]

[Yuli Ban]

AI Models to Help Photovoltaic Systems Find Their Place

With the looming threat of climate change, it is high time we embrace renewable energy sources on a larger scale. Photovoltaic systems, which generate electricity from the nearly limitless supply of sunlight energy, are one of the most promising ways of generating clean energy. However, integrating photovoltaic systems into existing power grids is not a straightforward process. Because the power output of photovoltaic systems depends heavily on environmental conditions, power plant and grid managers need estimations of how much power will be injected by photovoltaic systems so as to plan optimal generation and maintenance schedules, among other important operational aspects.

In line with modern trends, if something needs predicting, you can safely bet that artificial intelligence will make an appearance. To date, there are many algorithms that can estimate the power produced by photovoltaic systems several hours ahead by learning from previous data and analyzing current variables. One of them, called adaptive neuro-fuzzy inference system (ANFIS), has been widely applied for forecasting the performance of complex renewable energy systems. Since its inception, many researchers have combined ANFIS with a variety of machine learning algorithms to improve its performance even further.

[Yuli Ban]

Google parent Alphabet launches Intrinsic: a new company to build software for industrial robots

Google’s parent-company Alphabet has a birth to announce: a new company called Intrinsic which will focus on building software for industrial robots. The subsidiary will be one of Alphabet’s “other bets” — relatively speculative firms focusing on new technology like Waymo (self-driving cars), Wing (delivery drones), and Verily (healthcare and biotech).

Details on what exactly Intrinsic is building or who its customers will be are unclear. A blog post from the company’s new CEO, Wendy Tan White, discusses Intrinsic’s ambitions in broad terms, saying it will “unlock the creative and economic potential of industrial robotics for millions more businesses, entrepreneurs, and developers” by creating software that will make industrial robots “easier to use, less costly and more flexible.”

Robotics have been an obsession at Google for years, but the company’s efforts have been unfocused and have yet to produce any commercial hits.

[Yuli Ban]

[weatheriscool]

Nimble robotic arms that perform delicate surgery may be one step closer to reality
https://techxplore.com/news/2021-07-nim ... rgery.html
by Peter Ramjug, Northeastern University

Researchers at Northeastern are working to eliminate the stiff, herky-jerky motions in robotic arms to make them graceful and deft enough to gently pick up an egg or sturdy enough to stack dinner plates. The findings could one day allow doctors to remotely perform surgery on a distant battlefield or help bomb disposal experts safely remove an explosive device.

A video demonstration of a university project involving a researcher wearing a C-shaped gripping claw attached to his right hand while a nearby robotic arm mimicked his exact movements showed the promise of hydraulic technology designed to be low friction.

The researcher in the video lowered and raised his arm, swept it left and right, and bent it at the wrist, smooth actions that were copied in tandem by the robotic arm. What was not readily apparent was how the human operator was

able to feel the same forces as the mechanical arm when it closed on an object, allowing the user to get a sense of textured surfaces.

[Yuli Ban]

Bipedal robot developed at Oregon State makes history by learning to run, completing 5K

Cassie the robot, invented at Oregon State University and produced by OSU spinout company Agility Robotics, has made history by traversing 5 kilometers, completing the route in just over 53 minutes.

Cassie was developed under the direction of robotics professor Jonathan Hurst with a 16-month, $1 million grant from the Advanced Research Projects Agency of the U.S. Department of Defense.

Since Cassie’s introduction in 2017, OSU students funded by the National Science Foundation have been exploring machine learning options for the robot.

Cassie, the first bipedal robot to use machine learning to control a running gait on outdoor terrain, completed the 5K on Oregon State’s campus untethered and on a single battery charge.

[wjfox]

With post-pandemic AI, we’ve now stepped into the Age of Acceleration

July 25, 2021 6:21 AM

As the IBM Watson experience shows, the path to AI success is fraught with challenges. Yet overall, it has been a very good year for AI and the companies developing it. So much so that Alphabet CEO Sundar Pichai, in a recent podcast recorded by BBC, says: “I view [AI] as a very profound enabling technology. If you think about fire or electricity or the internet, it is like that, but I think even more profound.”

That profound impact is becoming more pronounced as AI is showing up in more industries, ranging from semiconductor design to software development to voiceovers, farming, distribution, music creation, and even classical sculpting. In all instances, AI is augmenting and possibly replacing human activities while dramatically speeding up development of the final product. In biology, determining the structure of just one protein can take years of laboratory work, but new AI released to the public by the University of Washington can reduce this time to as little as 10 minutes. In the sculpture example, a replica of “Psyche Revived by Cupid’s Kiss,” produced by ABB2, an industrial robot developed by ABB Robotics, required just over 11 days to produce, while the original by 18th-century sculptor Canova required roughly five years. And due to the pandemic, demand for industrial robots has surged in the last year across many industries.

Read more: https://venturebeat.com/2021/07/25/with ... eleration/

[Yuli Ban]

[Yuli Ban]

[Yuli Ban]

The Future of Deep Learning Is Photonic

Think of the many tasks to which computers are being applied that in the not-so-distant past required human intuition. Computers routinely identify objects in images, transcribe speech, translate between languages, diagnose medical conditions, play complex games, and drive cars.

The technique that has empowered these stunning developments is called deep learning, a term that refers to mathematical models known as artificial neural networks. Deep learning is a subfield of machine learning, a branch of computer science based on fitting complex models to data.

While machine learning has been around a long time, deep learning has taken on a life of its own lately. The reason for that has mostly to do with the increasing amounts of computing power that have become widely available—along with the burgeoning quantities of data that can be easily harvested and used to train neural networks.

The amount of computing power at people's fingertips started growing in leaps and bounds at the turn of the millennium, when graphical processing units (GPUs) began to be harnessed for nongraphical calculations, a trend that has become increasingly pervasive over the past decade. But the computing demands of deep learning have been rising even faster. This dynamic has spurred engineers to develop electronic hardware accelerators specifically targeted to deep learning, Google's Tensor Processing Unit (TPU) being a prime example.

Here, I will describe a very different approach to this problem—using optical processors to carry out neural-network calculations with photons instead of electrons. To understand how optics can serve here, you need to know a little bit about how computers currently carry out neural-network calculations. So bear with me as I outline what goes on under the hood.

[Yuli Ban]

[Yuli Ban]

OSU Dynamic Robotics Laboratory's research team, led by Agility Robotics’ Jonathan Hurst, combined expertise from biomechanics and robot controls with new machine learning tools to accomplish something new: train a bipedal robot to run a full 5K on a single battery charge.

[Yuli Ban]

[Yuli Ban]

Kitchen robot in Riga cooks up new future for fast food

A pasta order comes in and the robotic arm springs into action at the Roboeatz eatery in Riga. After five minutes of gyrations, a piping hot plate is ready.

The Riga cafe, located under a crumbling concrete bridge, is designed in such a way that customers can observe the robotic arm at work.

It also has a seating area, although most customers prefer take away since vaccination certificates are required to be able to eat indoors in Latvia.

A Roboeatz app allows customers to order and pay for their dish before picking it up at the cafe.

[Yuli Ban]

A talking robot is serving restaurant customers in California because the manager can't find enough workers during the labor shortage

A restaurant in Stockton, California, is using a robot helper to serve customers as it struggles to hire workers amid the US labor shortage.

"We are struggling to find people to come in and work, just like every other business right now," Ana Ortiz, general manager of the Sugar Mediterranean Bistro, told NBC News.

"I don't have enough employees to be running around food and serving tables," she added.

The robot was created by Richtech Robotics, and has been given the name Matradee by the company. It's capable of opening kitchen doors, allowing deliveries to go from the kitchen to the table, Richtech's website states.

The Matradee also uses LiDAR, among other technology, to maneuver and detect its surroundings up to 20 feet.

This helps it to carry up to four trays of food and dishes to hungry customers. "So, let's say I'm at table two, I'm taking the order for table two while the robot is running the food for me to table seven. I load up the robot with dirty dishes, and it takes it right back to the dishwasher," Ortiz told NBC News.

[Yuli Ban]

[2107.12544] Human-Level Reinforcement Learning through Theory-Based Modeling, Exploration, and Planning

Reinforcement learning (RL) studies how an agent comes to achieve reward in an environment through interactions over time. Recent advances in machine RL have surpassed human expertise at the world's oldest board games and many classic video games, but they require vast quantities of experience to learn successfully -- none of today's algorithms account for the human ability to learn so many different tasks, so quickly. Here we propose a new approach to this challenge based on a particularly strong form of model-based RL which we call Theory-Based Reinforcement Learning, because it uses human-like intuitive theories -- rich, abstract, causal models of physical objects, intentional agents, and their interactions -- to explore and model an environment, and plan effectively to achieve task goals. We instantiate the approach in a video game playing agent called EMPA (the Exploring, Modeling, and Planning Agent), which performs Bayesian inference to learn probabilistic generative models expressed as programs for a game-engine simulator, and runs internal simulations over these models to support efficient object-based, relational exploration and heuristic planning. EMPA closely matches human learning efficiency on a suite of 90 challenging Atari-style video games, learning new games in just minutes of game play and generalizing robustly to new game situations and new levels. The model also captures fine-grained structure in people's exploration trajectories and learning dynamics. Its design and behavior suggest a way forward for building more general human-like AI systems.

[Yuli Ban]

Australian court rules an AI can be considered an inventor on patent filings

An Australian Court has decided that an artificial intelligence can be recognised as an inventor in a patent submission.

In a case brought by Stephen Thaler, who has filed and lost similar cases in other jurisdictions, Australia's Federal Court last month heard and decided that the nation's Commissioner of Patents erred when deciding that an AI can't be considered an inventor.

Justice Beach reached that conclusion because nothing in Australia law says the applicant for a patent must be human.

As Beach's judgement puts it: "… in my view an artificial intelligence system can be an inventor for the purposes of the Act.

"First, an inventor is an agent noun; an agent can be a person or thing that invents. Second, so to hold reflects the reality in terms of many otherwise patentable inventions where it cannot sensibly be said that a human is the inventor. Third, nothing in the Act dictates the contrary conclusion."

The Justice also worried that the Commissioner of Patents' logic in rejecting Thaler's patent submissions was faulty.

"On the Commissioner's logic, if you had a patentable invention but no human inventor, you could not apply for a patent," the judgement states. "Nothing in the Act justifies such a result."