AI & Robotics News and Discussions

Yuli Ban · Post by **Yuli Ban** » Thu Jun 30, 2022 8:02 pm

Language models have demonstrated remarkable performance on a variety of natural language tasks — indeed, a general lesson from many works, including BERT, GPT-3, Gopher, and PaLM, has been that neural networks trained on diverse data at large scale in an unsupervised way can perform well on a variety of tasks.

Quantitative reasoning is one area in which language models still fall far short of human-level performance. Solving mathematical and scientific questions requires a combination of skills, including correctly parsing a question with natural language and mathematical notation, recalling relevant formulas and constants, and generating step-by-step solutions involving numerical calculations and symbolic manipulation. Due to these challenges, it is often believed that solving quantitative reasoning problems using machine learning will require significant advancements in model architecture and training techniques, granting models access to external tools such as Python interpreters, or possibly a more profound paradigm shift.

In “Solving Quantitative Reasoning Problems With Language Models” (to be released soon on the arXiv), we present Minerva, a language model capable of solving mathematical and scientific questions using step-by-step reasoning. We show that by focusing on collecting training data that is relevant for quantitative reasoning problems, training models at scale, and employing best-in-class inference techniques, we achieve significant performance gains on a variety of difficult quantitative reasoning tasks. Minerva solves such problems by generating solutions that include numerical calculations and symbolic manipulation without relying on external tools such as a calculator. The model parses and answers mathematical questions using a mix of natural language and mathematical notation. Minerva combines several techniques, including few-shot prompting, chain of thought or scratchpad prompting, and majority voting, to achieve state-of-the-art performance on STEM reasoning tasks. You can explore Minerva’s output with our interactive sample explorer!

Yuli Ban · Post by **Yuli Ban** » Fri Jul 01, 2022 4:30 am

Yuli Ban · Post by **Yuli Ban** » Fri Jul 01, 2022 11:52 am

Yuli Ban · Post by **Yuli Ban** » Fri Jul 01, 2022 12:53 pm

Yuli Ban · Post by **Yuli Ban** » Fri Jul 01, 2022 4:32 pm

Yuli Ban · Post by **Yuli Ban** » Sat Jul 02, 2022 4:55 am

weatheriscool · Post by **weatheriscool** » Sat Jul 02, 2022 7:48 pm

A model that allows robots to follow and guide humans in crowded environments
https://techxplore.com/news/2022-06-rob ... ments.html
by Ingrid Fadelli , Tech Xplore

Assistance robots are typically mobile robots designed to assist humans in malls, airports, health care facilities, home environments and various other settings. Among other things, these robots could help users to find their way around unknown environments, for instance guiding them to a specific location or sharing important information with them.

While the capabilities of assistance robots have improved significantly over the past decade, the systems that have so far been implemented in real-world environments are not yet capable of following or guiding humans efficiently within crowded spaces. In fact, training robots to track a specific user while navigating a dynamic environment characterized by many randomly moving "obstacles" is far from a simple task.

Researchers at the Berlin Institute of Technology have recently introduced a new model based on deep reinforcement learning that could allow mobile robots to guide a specific user to a desired location or follow him/her around while carrying their belongings, all within a crowded environment. This model, introduced in a paper pre-published on arXiv, could help to significantly enhance the capabilities of robots in malls, airports and other public places.

"The task of guiding or following a human in crowded environments, such as airports or train stations, to carry weight or goods is still an open problem," Linh Kästner , Bassel Fatloun , Zhengcheng Shen , Daniel Gawrisch and Jens Lambrecht wrote in their paper. "In these use cases, the robot is not only required to intelligently interact with humans, but also to navigate safely among crowds."

Post by **caltrek** » Wed Jul 06, 2022 10:34 pm

Carnegie Mellon Engineering Is Reimagining Nanosatellite Capabilities With Orbital Edge Computing
July 6, 2022

Introduction:

(EurekAlert) PITTSBURGH — Researchers at Carnegie Mellon University’s College of Engineering are setting out on a mission to reimagine the capabilities of nanosatellites in low-Earth orbit. Backed by a $7 million grant from the National Science Foundation’s (NSF) Cyber-Physical Systems (CPS) Frontiers Program, the CMU initiative will transform constellations of nanosatellites into sophisticated distributed computing platforms, building the foundation for a wide range of novel applications in public safety, defense and intelligence, carbon mapping, traffic management and precision agriculture, among others.

Today’s nanosatellites collect enormous amounts of raw data, so much that it’s impossible to downlink all of it to earth. The long loop required to beam just a portion of the data to the ground and then make sense of it also creates many latency issues.

With the team’s new approach, called orbital edge computing, researchers at CMU will work to develop computationally capable constellations of nanosatellites, equipped with machine learning techniques that extract valuable insights from data while still in orbit. This will not only reduce the amount of information being sent to earth but will build the foundation for a wide array of possible responsive applications that operate entirely from space.

The new technology will help detect the initial signs of problems before they occur, according to principal investigator Brandon Lucia, associate professor of electrical and computer engineering. For example, it could allow for monitoring suspicious activity at large-scale events like the upcoming 2028 Summer Olympics in Los Angeles or identify early signs of wildfires, enabling response teams to make mitigation efforts before forests are ablaze.

The project comprises world-leading experts in critical areas like federated learning, wireless communications, security and networking, and nanosatellite design, including Carnegie Mellon professors Gauri Joshi, Swarun Kumar, Zac Manchester and Vyas Sekar.

Read more here: https://www.eurekalert.org/news-releases/958067

Post by **caltrek** » Fri Jul 08, 2022 2:37 pm

So, a Robot Walks Into a Nursing Home...
by Jennifer A. Kingson
July 8, 2022

Introduction:

(Axios) A 4-foot-tall droid named Pepper — preprogrammed with hundreds of jokes — is one of two robots now working at a nursing home in Roseville, Minnesota, entertaining residents and helping monitor their health.

• A sample from its sometimes-salty repertoire: "I went on a date with a Roomba last week — it totally sucked."

The big picture: Household robots are growing in utility and ubiquity, and this latest use seems to blend two trends we've noted: The availability of humanoid robots as party guests and entertainers and their use as companions and health monitors for the elderly.

Driving the news: As part of a study by the University of Minnesota Duluth, two robots (designed by SoftBank Robotics) have been deployed at a nursing home owned by Monarch Healthcare Management.

• When not delivering a standup routine, Pepper will mingle with residents, to "remind them to eat and exercise, and react to their facial expressions or tone of voice," per the Minnesota Star Tribune.

Read more here: https://www.axios.com/2022/07/08/robot ... obotics

Post by **caltrek** » Fri Jul 08, 2022 8:55 pm

Responsive Soft Robots Inspired by Sputtering Ketchup Bottle
July 8, 2022

Introduction:

(EurekAlert) A smartly designed pressure valve allows soft robots to respond to their environment without the need for computer control, reveal AMOLF researchers in their article in the journal Matter. That brings robots with natural movements and tactile responses similar to those of living organisms one step closer to reality. Such developments render soft robots more suitable for exploring rough and unknown terrain or for medical applications.

Read more here: https://www.eurekalert.org/news-releases/957996

weatheriscool · Post by **weatheriscool** » Sat Jul 09, 2022 6:07 am

AI as a 'wise counsel' for synthetic biology
https://techxplore.com/news/2022-07-ai- ... ology.html
by Max Planck Society

Machine learning is transforming all areas of biological science and industry, but is typically limited to a few users and scenarios. A team of researchers at the Max Planck Institute for Terrestrial Microbiology led by Tobias Erb has developed METIS, a modular software system for optimizing biological systems. The research team demonstrates its usability and versatility with a variety of biological examples.

Though engineering of biological systems is truly indispensable in biotechnology and synthetic biology, today machine learning has become useful in all fields of biology. However, it is obvious that application and improvement of algorithms, computational procedures made of lists of instructions, is not easily accessible. Not only are they limited by programming skills but often also insufficient experimentally-labeled data. At the intersection of computational and experimental works, there is a need for efficient approaches to bridge the gap between machine learning algorithms and their applications for biological systems.

Now a team at the Max Planck Institute for Terrestrial Microbiology led by Tobias Erb has succeeded in democratizing machine learning. In their recent publication in Nature Communications, the team presented together with collaboration partners from the INRAe Institute in Paris, their tool METIS. The application is built in such a versatile and modular architecture that it does not require computational skills and can be applied on different biological systems and with different lab equipment. METIS is short from Machine-learning guided Experimental Trials for Improvement of Systems and also named after the ancient goddess of wisdom and crafts Μῆτις, or "wise counsel."

Post by **caltrek** » Sat Jul 09, 2022 8:17 pm

Swarms of Tiny Robots Could One Day Floss Your Teeth
by Katherine Unger
July 7, 2022

Introduction:

(Futurity) A shapeshifting swarm of tiny robots could one day brush, rinse, and floss your teeth.

The system could be particularly valuable for those who lack the manual dexterity to clean their teeth effectively themselves.

The building blocks of these microrobots are iron oxide nanoparticles that have both catalytic and magnetic activity. Using a magnetic field, researchers could direct their motion and configuration to form either bristle-like structures that sweep away dental plaque from the broad surfaces of teeth, or elongated strings that can slip between teeth like a length of floss. In both instances, a catalytic reaction drives the nanoparticles to produce antimicrobials that kill harmful oral bacteria on site.

Experiments using this system on mock and real human teeth showed that the robotic assemblies can conform to a variety of shapes to nearly eliminate the sticky biofilms that lead to cavities and gum disease. The research team shares their findings establishing a proof-of-concept for the robotic system in the journal ACS Nano.

“Routine oral care is cumbersome and can pose challenges for many people, especially those who have hard time cleaning their teeth,” says Hyun (Michel) Koo, a professor in the orthodontics department and divisions of Community Oral Health and Pediatric Dentistry in the University of Pennsylvania’s School of Dental Medicine and co-corresponding author of the study.

Read more here: https://www.futurity.org/robots-swarm-brush-floss

weatheriscool · Post by **weatheriscool** » Wed Jul 13, 2022 5:45 am

Artificial intelligence model finds potential drug molecules a thousand times faster
https://techxplore.com/news/2022-07-art ... cules.html
by Alex Ouyang, Massachusetts Institute of Technology

The entirety of the known universe is teeming with an infinite number of molecules. But what fraction of these molecules have potential drug-like traits that can be used to develop life-saving drug treatments? Millions? Billions? Trillions? The answer: novemdecillion, or 1060. This gargantuan number prolongs the drug development process for fast-spreading diseases like COVID-19 because it is far beyond what existing drug design models can compute. To put it into perspective, the Milky Way has about 100 thousand million, or 108, stars.

In a paper that will be presented at the International Conference on Machine Learning (ICML), MIT researchers developed a geometric deep-learning model called EquiBind that is 1,200 times faster than one of the fastest existing computational molecular docking models, QuickVina2-W, in successfully binding drug-like molecules to proteins. EquiBind is based on its predecessor, EquiDock, which specializes in binding two proteins using a technique developed by the late Octavian-Eugen Ganea, a recent MIT Computer Science and Artificial Intelligence Laboratory and Abdul Latif Jameel Clinic for Machine Learning in Health (Jameel Clinic) postdoc, who also co-authored the EquiBind paper.

weatheriscool · Post by **weatheriscool** » Wed Jul 13, 2022 5:54 am

Researchers trained an AI model to 'think' like a baby, and it suddenly excelled

by Susan Hespos, The Conversation
https://techxplore.com/news/2022-07-ai- ... elled.html

In a world rife with opposing views, let's draw attention to something we can all agree on: If I show you my pen, and then hide it behind my back, my pen still exists—even though you can't see it anymore. We can all agree it still exists, and probably has the same shape and color it did before it went behind my back. This is just common sense.

These common-sense laws of the physical world are universally understood by humans. Even two-month-old infants share this understanding. But scientists are still puzzled by some aspects of how we achieve this fundamental understanding. And we've yet to build a computer that can rival the common-sense abilities of a typically developing infant.

New research by Luis Piloto and colleagues at Princeton University—which I'm reviewing for an article in Nature Human Behaviour—takes a step towards filling this gap. The researchers created a deep-learning artificial intelligence (AI) system that acquired an understanding of some common-sense laws of the physical world.

The findings will help build better computer models that simulate the human mind, by approaching a task with the same assumptions as an infant.

Childish behavior

Typically, AI models start with a blank slate and are trained on data with many different examples, from which the model constructs knowledge. But research on infants suggests this is not what babies do. Instead of building knowledge from scratch, infants start with some principled expectations about objects.

For instance, they expect if they attend to an object that is then hidden behind another object, the first object will continue to exist. This is a core assumption that starts them off in the right direction. Their knowledge then becomes more refined with time and experience.

The exciting finding by Piloto and colleagues is that a deep-learning AI system modeled on what babies do, outperforms a system that begins with a blank slate and tries to learn based on experience alone.

weatheriscool · Post by **weatheriscool** » Fri Jul 15, 2022 6:36 am

Providing embedded artificial intelligence with a capacity for palimpsest memory storage
https://techxplore.com/news/2022-07-emb ... psest.html
by Thamarasee Jeewandara , Tech Xplore

Biological synapses are known to store multiple memories on top of each other at different time scales, much like representations of the early techniques of manuscript writing known as "palimpsest," where annotations can be superimposed alongside traces of earlier writing.

Biological palimpsest consolidation occurs via hidden biochemical processes that govern synaptic efficacy at varying lifetimes. The arrangement can facilitate idle memories to be overwritten without forgetting them, while using previously unseen memories short-term. Embedded artificial intelligence can significantly benefit from such functionality; however, the hardware has yet to be demonstrated in practice.

In a new report, now published in Science Advances, Christos Giotis and a team of scientists in Electronics and Computer Science at the University of Southampton and the University of Edinburgh, U.K., showed how the intrinsic properties of metal-oxide volatile memristors mimicked the process of biological palimpsest consolidation.

Memristors are devices that can regulate the flow of electric current in a circuit to remember the charge flowing through it. Without implementing special instructions, the experimental memristor synapses displayed expanded doubled capacity while protecting a consolidated memory, as up to hundreds of uncorrelated short-term memories temporarily overwrote it. The outcomes highlighted how the emerging memory technologies can effectively expand the capacity of artificial intelligence hardware to conceptualize learning memory.

weatheriscool · Post by **weatheriscool** » Fri Jul 15, 2022 6:45 am

Using AI to diagnose birth defect in fetal ultrasound images
https://medicalxpress.com/news/2022-07- ... sound.html
by David McFadden, University of Ottawa

In a new proof-of-concept study led by Dr. Mark Walker at the uOttawa Faculty of Medicine, researchers are pioneering the use of a unique AI-based deep learning model as an assistive tool for the rapid and accurate reading of ultrasound images.

It's trailblazing work because although deep learning models have become increasingly popular in interpreting medical images and detecting disorders, figuring out how its application can work in obstetric ultrasonography is in its nascent stages. Few AI-enabled studies have been published in this field.

The goal of the team's study was to demonstrate the potential for deep-learning architecture to support early and reliable identification of cystic hygroma from first trimester ultrasound scans. Cystic hygroma is an embryonic condition that causes the lymphatic vascular system to develop abnormally. It's a rare and potentially life-threatening disorder that leads to fluid swelling around the head and neck.

The birth defect can typically be easily diagnosed prenatally during an ultrasound appointment, but Dr. Walker and his research group wanted to test how well AI-driven pattern recognition could do the job. The findings are promising.

Yuli Ban · Post by **Yuli Ban** » Mon Jul 18, 2022 12:38 am

A robot can create a model of itself to plan how to move and reach a goal – something its developers say makes it self-aware, though others disagree.

Every robot is trained in some way to do a task, often in a simulation. By seeing what to do, robots can then mimic the task. But they do so unthinkingly, perhaps relying on sensors to try to reduce collision risks, rather than having any understanding of why they are performing the task or a true awareness of where they are within physical space. It means they will often make mistakes – bashing their arm into an obstacle, for instance – that humans wouldn’t because they would compensate for changes.

“It’s a very essential capability of humans that we normally take for granted,” says Boyuan Chen at Duke University, North Carolina.

I don't know about this one. LaMDA wasn't sentient either but it offered a tantalizing hint of something more, a proof of concept that chatbots can fool humans into thinking they're self-aware through pure emergence from the sheer scale of LLMs.

This, on the other hand, feels like one of those 2000s/2010s-era tech stories about a "robot that has self-awareness!" because it was programmed to recognize itself.

Yuli Ban · Post by **Yuli Ban** » Mon Jul 18, 2022 6:21 am

Post by **caltrek** » Wed Jul 20, 2022 5:51 pm

Look What’s Happening Tomorrow at TechCrunch Sessions: Robotics
by Lauren Simonds
July 20, 2022

Introduction:

(TechCrunch) Who’s ready for robots? Tomorrow, July 21, is the big day when the global community of people who live, breathe and build robots gather to talk about the latest developments in the field and get a look at demos of the newest robotics products and artificial intelligence applications.
If you haven’t registered for TC Sessions: Robotics, a free online event, just click here to reserve your seat: https://techcrunch.com/events/tc-sessi ... splay=true.

Bonus: Not only do you get a deep download on robotics and AI news, products and trends, you can also connect and chat with other enthusiasts from around the world using our event app: https://techcrunch.com/events/tc-sessi ... splay=true.

The app takes the pain out of finding the right people who share your business goals. Expand your network, explore collaboration opportunities and build your community.

Let’s take a look at some of the highlights you can expect tomorrow. Check out the complete agenda: https://techcrunch.com/events/tc-sessi ... w&display= to plan your day and prioritize your networking strategy.

Read more here: https://techcrunch.com/2022/07/20/loo ... obotics/

Post by **caltrek** » Fri Jul 22, 2022 5:26 pm

Here Is Everything You Missed at the TechCrunch Session: Robotics 2022
by Matt Burns
July 22, 2022

Introduction:

(TechCrunch) In case you missed it, robots took over TechCrunch on Thursday, July 21. We played host to the robotic industry’s leading startups, researchers and academics at TC Sessions: Robotics. The event was a blockbuster success, and we hope you enjoyed the show. All the features, panels, interviews and podcasts are embedded below.

TechCrunch Editor Brian Heater organized and hosted the event. Subscribe to his robotics newsletter, Actuatorhttps://techcrunch.com/newsletters[url] ... cs-2022/