AI & Robotics News and Discussions

[Vakanai]

Okay, I will admit it - DALL-E 2 both impresses and scares me. AI doing text and voice thingies is something I'm long used to and expect. It doing art thingies is mind blowing and wow.

(Yes, thingies, I'm not a technical wordy type person, please don't judge)

[Nero]

This is insane levels of performance gain for such little compute. I actually did not believe it.

To put this in perspective this is akin to a video game on your PS4 running on the original XBOX from 2001 at the SAME level of performance bear minimum. The fact we can make models three orders of magnitude more efficient in less than two years is astounding even if it is limited in application.

Seriously this is insane. Utterly insane.

[Yuli Ban]

[Yuli Ban]

starspawn0

Look at what systems like DALL-E2, PaLM, Chinchilla, Flamingo, and Gato can already do. Hard to imagine what 10 more years of progress like that will look like. 10 years ago Alexnet wasn't even on the radar -- it didn't compete in the Imagenet competition until late September 2012, and before then there were just some isolated successes from deep neural nets -- nothing spectacular. The transformation over the next 10 years will probably be even greater.

....

Given that Apple and Google are now entering the AR / VR race with Facebook, there'll be a need for applying DALL-E2-like synthetic media to generate VR content. It would be something to behold to walk around in a 3D painting. I suppose getting enough training data to pull this off will be the main issue.

There should also be big advances in video synthesis. Current lack of progress might just be due to insufficient scale; or it could be that people are making some false assumptions about what's holding it back. I'm sure they'll find the right mix of things to make it work really well soon enough.

Of course there will also be conversational AI systems that can do talk deeply about basically anything, and do things on your behalf. They won't be immediately released to the public, but you'll read about them.

We might see some big advances in robotics, due to large companies like Google training armies of robots, collecting the training data, and rolling it all into a Gato-like model. They could hire hundreds of people, also, to drive bots to complete tasks using a controller.

[Tadasuke]

>
Autonomous cargo ship completes 500 mile voyage, avoiding hundreds of collisions

https://electrek.co/2022/05/13/autonomo ... ollisions/

"The “world’s first” autonomous commercial cargo ship has successfully completed a near-500 mile voyage in the congested waters of Tokyo Bay, traveling without human intervention for 99% of the trip. The 750 gross-ton vessel was powered by Orca AI, whose software helped the ship avoid hundreds of collisions autonomously.

Orca AI is a developer of safety software platforms designed specifically for maritime vessels. Founded in 2018 by two naval technology experts, the company combines sensors with existing safety systems onboard to help improve the safety and navigation of ships on crowded waterways.

Headquartered in Israel, Orca AI looks to bridge sea-bound ships to the shore with 24/7 insights to ensure shipping companies keep their cargo safe and efficient at all times, all while providing the technology to bring autonomous cargo ships to reality."

By 2040, a significant % of cargo ships will be autonomous, sailing on autopilot. Not all of them though. Gradually, less and less crew will be needed.

[caltrek]

Twisted Soft Robots Navigate Mazes Without Human or Computer Guidance
May 23, 2022

Introduction:

(EurekAlert) Researchers from North Carolina State University and the University of Pennsylvania have developed soft robots that are capable of navigating complex environments, such as mazes, without input from humans or computer software.

“These soft robots demonstrate a concept called ‘physical intelligence,’ meaning that structural design and smart materials are what allow the soft robot to navigate various situations, as opposed to computational intelligence,” says Jie Yin, corresponding author of a paper on the work and an associate professor of mechanical and aerospace engineering at NC State.

The soft robots are made of liquid crystal elastomers in the shape of a twisted ribbon, resembling translucent rotini. When you place the ribbon on a surface that is at least 55 degrees Celsius (131 degrees Fahrenheit), which is hotter than the ambient air, the portion of the ribbon touching the surface contracts, while the portion of the ribbon exposed to the air does not. This induces a rolling motion in the ribbon. And the warmer the surface, the faster it rolls. Video of the ribbon-like soft robots can be found at

“This has been done before with smooth-sided rods, but that shape has a drawback – when it encounters an object, it simply spins in place,” says Yin. “The soft robot we’ve made in a twisted ribbon shape is capable of negotiating these obstacles with no human or computer intervention whatsoever.”

The ribbon robot does this in two ways. First, if one end of the ribbon encounters an object, the ribbon rotates slightly to get around the obstacle. Second, if the central part of the robot encounters an object, it “snaps.” The snap is a rapid release of stored deformation energy that causes the ribbon to jump slightly and reorient itself before landing. The ribbon may need to snap more than once before finding an orientation that allows is to negotiate the obstacle, but ultimately it always finds a clear path forward.

Read further here: https://www.eurekalert.org/news-releases/953139

[weatheriscool]

Computer scientists suggest research integrity could be at risk due to AI generated imagery
https://techxplore.com/news/2022-05-sci ... agery.html
by Bob Yirka , Tech Xplore

A small team of researchers at Xiamen University has expressed alarm at the ease with which bad actors can now generate fake AI imagery for use in research projects. They have published an opinion piece outlining their concerns in the journal Patterns.

When researchers publish their work in established journals, they often include photographs to show the results of their work. But now the integrity of such photographs is under assault by certain entities who wish to circumvent standard research protocols. Instead of generating photographs of their actual work, they can instead generate them using artificial-intelligence applications. Generating fake photographs in this way, the researchers suggest, could allow miscreants to publish research papers without doing any real research.

To demonstrate the ease with which fake research imagery could be generated, the researchers generated some of their own using a generative adversarial network (GAN), in which two systems, one a generator, the other a discriminator, attempt to outcompete one another in creating a desired image. Prior research has shown that the approach can be used to create images of strikingly realistic human faces. In their work, the researchers generated two types of images. The first kind were of a western blot—an imaging approach used for detecting proteins in a blood sample. The second was of esophageal cancer images. The researchers then presented the images they had created to biomedical specialists—two out of three were unable to distinguish them from the real thing.

[caltrek]

This New AI Can Detect the Calls of Animals Swimming in an Ocean of Noise
by Carly Cassella
My 24, 2022

Introduction:

(Science Alert) The ocean is swimming in sound, and a new artificial intelligence tool could help scientists sift through all that noise to track and study marine mammals.

The tool is called DeepSqueak, not because it measures dolphin calls in the ocean underworld, but because it is based on a deep learning algorithm that was first used to categorize the different ultrasonic squeals of mice.

Now, researchers are applying the technology to vast datasets of marine bioacoustics.

Given that much of the ocean is out of our physical reach, underwater sound could help us understand where marine mammals swim, their density and abundance, and how they interact with one another.

Already, recordings of whale songs have helped identify an unknown population of blue whales in the Indian Ocean and a never-before-heard species of beaked whale

Read more here: https://www.sciencealert.com/new-ai-too ... deepsqueak

[caltrek]

The World's Smallest Remote-Controlled Walking Robot
by Eleanor Higgs
May 26, 2022

Introduction:

(IFL Science) This remote-controlled walking robot is the smallest ever created, and it could one day be scuttling around inside you.

Say hello to the peekytoe crab robot which is able to move without complex hardware, hydraulics, or electricity. Instead, the locomotion is created by the robot’s shape-memory alloy, which deforms into its remembered shape when heated.

A laser is used to rapidly heat the robot at different targeted locations, and a thin coating of glass elastically returns that part of the robot to its deformed shape when cooling. As the robot changes shape from deformed to remembered and back, this action creates the locomotion. The laser can control the direction of travel, and the robot can crawl, twist, walk, turn, and even jump. The research is published in the journal Science Robotics.

Researchers at Northwestern University were inspired by the mechanism used in children’s “pop-up” books to create the tiny robot crab. The weeny bot is 0.5 millimeters (0.02 inches) wide, making it even smaller than a flea, and it is hoped that similar robots could one day be used to perform tasks in tight spaces, including within the human body.

View more here: https://www.iflscience.com/technology/t ... ng-robot/

[raklian]

[Yuli Ban]

[Ozzie guy]

These deep mind updates are happening fast.

"Abstract
A longstanding goal of the field of AI is a strategy for compiling diverse experience into a highly capable, generalist agent. In the subfields of vision and language, this was largely achieved by scaling up transformer-based models and training them on large, diverse datasets. Motivated by this progress, we investigate whether the same strategy can be used to produce generalist reinforcement learning agents. Specifically, we show that a single transformer-based model -- with a single set of weights -- trained purely offline can play a suite of up to 46 Atari games simultaneously at close-to-human performance. When trained and evaluated appropriately, we find that the same trends observed in language and vision hold, including scaling of performance with model size and rapid adaptation to new games via fine-tuning. We compare several approaches in this multi-game setting, such as online and offline RL methods and behavioral cloning, and find that our Multi-Game Decision Transformer models offer the best scalability and performance. We release the pre-trained models and code to encourage further research in this direction."


https://sites.google.com/view/multi-game-transformers

[Nanotechandmorefuture]

Man definitely gotta look into cloud robotics or anything cool and future tech related to our very near future tech workforce. Its gonna be so drastically different it will be as awesome as it will be daunting at first.

[Ozzie guy]

This Chinese company or organization is working on AGI and seemingly made a breakthrough.





Should AGI related news be posted here or in the singularity thread?

[raklian]

[Yuli Ban]

Interesting! Would be nice to have another humanoid around after ASIMO was retired.

[weatheriscool]

Artificial skin gives robots sense of touch and beyond
https://techxplore.com/news/2022-06-art ... obots.html
by Emily Velasco, California Institute of Technology

We tend to take our sense of touch for granted in everyday settings, but it is vital for our ability to interact with our surroundings. Imagine reaching into the fridge to grab an egg for breakfast. As your fingers touch its shell, you can tell the egg is cold, that its shell is smooth, and how firmly you need to grip it to avoid crushing it. These are abilities that robots, even those directly controlled by humans, can struggle with.

A new artificial skin developed at Caltech can now give robots the ability to sense temperature, pressure, and even toxic chemicals through a simple touch.

This new skin technology is part of a robotic platform that integrates the artificial skin with a robotic arm and sensors that attach to human skin. A machine-learning system that interfaces the two allows the human user to control the robot with their own movements while receiving feedback through their own skin. The multimodal robotic-sensing platform, dubbed M-Bot, was developed in the lab of Wei Gao, Caltech's assistant professor of medical engineering, investigator with Heritage Medical Research Institute, and Ronald and JoAnne Willens Scholar. It aims to give humans more precise control over robots while also protecting the humans from potential hazards.

"Modern robots are playing a more and more important role in security, farming, and manufacturing," Gao says. "Can we give these robots a sense of touch and a sense of temperature? Can we also make them sense chemicals like explosives and nerve agents or biohazards like infectious bacteria and viruses? We're working on this."

[caltrek]

Note that this (see below) describes research at the University of Glasgow, as distinct from research at Caltech as described in the above post.

Percptron: Robots That Feel Pain and AI That Predicts Soccer Players Movements
by Kyle Wiggers & Devin Coldewey
May 31, 2022

Introduction:

(TechCrunch) Research in the field of machine learning and AI, now a key technology in practically every industry and company, is far too voluminous for anyone to read it all. This column, Perceptron (previously Deep Science), aims to collect some of the most relevant recent discoveries and papers — particularly in, but not limited to, artificial intelligence — and explain why they matter.

This week in AI, a team of engineers at the University of Glasgow developed “artificial skin” that can learn to experience and react to simulated pain. Elsewhere, researchers at DeepMind developed a machine learning system that predicts where soccer players will run on a field, while groups from The Chinese University of Hong Kong (CUHK) and Tsinghua University created algorithms that can generate realistic photos — and even videos — of human models.

According to a press release, the Glasgow team’s artificial skin leveraged a new type of processing system based on “synaptic transistors” designed to mimic the brain’s neural pathways. The transistors, made from zinc-oxide nanowires printed onto the surface of a flexible plastic, connected to a skin sensor that registered changes in electrical resistance.

While artificial skin has been attempted before, the team claims that their design differed in that it used a circuit built into the system to act as an “artificial synapse” — reducing input to a spike in voltage. This sped up processing and allowed the team to “teach” the skin how to respond to simulated pain by setting a threshold of input voltage whose frequency varied according to the level of pressure applied to the skin.

The team sees the skin being used in robotics, where it could, for example, prevent a robotic arm from coming into contact with dangerously high temperatures

Read more here: https://techcrunch.com/2022/06/04/perc ... ovements/

[Yuli Ban]

[Yuli Ban]