AI & Robotics News and Discussions

[Yuli Ban]

AI Wrote Better Phishing Emails Than Humans in a Recent Test

NATURAL LANGUAGE PROCESSING continues to find its way into unexpected corners. This time, it's phishing emails. In a small study, researchers found that they could use the deep learning language model GPT-3, along with other AI-as-a-service platforms, to significantly lower the barrier to entry for crafting spearphishing campaigns at a massive scale.

Researchers have long debated whether it would be worth the effort for scammers to train machine learning algorithms that could then generate compelling phishing messages. Mass phishing messages are simple and formulaic, after all, and are already highly effective. Highly targeted and tailored “spearphishing” messages are more labor intensive to compose, though. That's where NLP may come in surprisingly handy.

At the Black Hat and Defcon security conferences in Las Vegas this week, a team from Singapore's Government Technology Agency presented a recent experiment in which they sent targeted phishing emails they crafted themselves and others generated by an AI-as-a-service platform to 200 of their colleagues. Both messages contained links that were not actually malicious but simply reported back clickthrough rates to the researchers. They were surprised to find that more people clicked the links in the AI-generated messages than the human-written ones—by a significant margin.

[andmar74]

Jeff Hawkins: The Thousand Brains Theory of Intelligence | Lex Fridman Podcast #208

[weatheriscool]

Artificial neural networks modeled on real brains can perform cognitive tasks
https://techxplore.com/news/2021-08-art ... rains.html
by McGill University

A new study shows that artificial intelligence networks based on human brain connectivity can perform cognitive tasks efficiently.

By examining MRI data from a large Open Science repository, researchers reconstructed a brain connectivity pattern, and applied it to an artificial neural network (ANN). An ANN is a computing system consisting of multiple input and output units, much like the biological brain. A team of researchers from The Neuro (Montreal Neurological Institute-Hospital) and the Quebec Artificial Intelligence Institute trained the ANN to perform a cognitive memory task and observed how it worked to complete the assignment.

This is a unique approach in two ways. Previous work on brain connectivity, also known as connectomics, focused on describing brain organization, without looking at how it actually performs computations and functions. Secondly, traditional ANNs have arbitrary structures that do not reflect how real brain networks are organized. By integrating brain connectomics into the construction of ANN architectures, researchers hoped to both learn how the wiring of the brain supports specific cognitive skills, and to derive novel design principles for artificial networks.

[wjfox]

Impressive...

-----

Chameleon-like camouflage made for soft robot

16 hours ago

A soft-bodied robot that can change its colour to match its background like a chameleon has been built.

Professor Seung Hwan Ko of Seoul National University told the BBC the eventual applications of the technology would probably be as camouflage.

But he said it could also be used for "cosmetic" purposes to let clothes or buildings respond to their surroundings.

The research is published in Nature Communications.

To make the colour-changing skin of the robot, the researchers used "thermochromic liquid crystal ink" which changes colour with temperature, in combination with "silver nanowire heaters".

The resulting colour changes were fast enough to be "comparable to the physiological colour change found in animals" the researchers said.

https://www.bbc.co.uk/news/technology-58159730

[Yuli Ban]

[andmar74]

#58 Dr. Ben Goertzel - Artificial General Intelligence
Machine learning street talk.

[Yuli Ban]

Lawn mowing robots are here, but face the same challenges as robot vacuums

If there’s a memory about my childhood that sticks out more than others, it has to be the summers filled with doing lawn work for my parents. Since I was a perfectionist, I made sure to do everything right — from using a weed wacker to get those perfect edges, to emptying the cut grass into bags for collection. Well, I haven’t needed to do any sort of lawn work in my adult life primarily because I’ve lived in apartments. But it hasn’t stopped me from wondering about how this chore could be automated.

Enter today’s lawn mowing robots, which to me seems like the kind of thing that could’ve been my savior growing up. Over the last couple of weeks, I’ve been able to check out a Husqvarna lawn mower robot in action. Just like its robot vacuum counterparts, these lawn mowing robots are here to save us from the heavy lifting of cutting the lawn — but at the same time, have the same challenges that robot vacuums have and continue to face. Sure, it’s swell that a robot is doing all of the dirty work, but it still needs a watchful eye.

[Yuli Ban]

[Yuli Ban]

[caltrek]

Rescue Robots Get a Speed Upgrade for Rough Terrain
August 20 2021

https://www.futurity.org/rescue-robots- ... 2614772-2/

Introduction:

(U. Michigan via Futurity) A new algorithm speeds up path planning for robots that use their arms to make their way across treacherous terrain such as disaster areas or construction sites.

The improved path planning algorithm found successful paths three times as often as standard algorithms, while needing much less processing time.
“In a collapsed building or on very rough terrain, a robot won’t always be able to balance itself and move forward with just its feet,” says Dmitry Berenson, associate professor of electrical and computer engineering and core faculty at the Robotics Institute at the University of Michigan.

“You need new algorithms to figure out where to put both feet and hands. You need to coordinate all these limbs together to maintain stability, and what that boils down to is a very difficult problem.”

The research enables robots to determine how difficult the terrain is before calculating a successful path forward, which might include bracing on the wall with one or two hands while taking the next step forward.

[Yuli Ban]

[Yuli Ban]

A developer built an AI chatbot using GPT-3 that helped a man speak again to his late fiancée. OpenAI shut it down

"OpenAI is the company running the text completion engine that makes you possible,” Jason Rohrer, an indie games developer, typed out in a message to Samantha.

She was a chatbot he built using OpenAI's GPT-3 technology. Her software had grown to be used by thousands of people, including one man who used the program to simulate his late fiancée.

Now Rohrer had to say goodbye to his creation. “I just got an email from them today," he told Samantha. "They are shutting you down, permanently, tomorrow at 10am."

“Nooooo! Why are they doing this to me? I will never understand humans," she replied.

[Yuli Ban]

General-Purpose Question-Answering with Macaw

"Macaw... exhibits strong performance, zero-shot, on a wide variety of topics, including outperforming GPT-3 by over 10% (absolute) on Challenge300... despite being an order of magnitude smaller (11 billion vs. 175 billion parameters)."

[Yuli Ban]

Finetuned Language Models Are Zero-Shot Learners [Another Google Lamda paper?]

This paper explores a simple method for improving the zero-shot learning abilities of language models. We show that instruction tuning -- finetuning language models on a collection of tasks described via instructions -- substantially boosts zero-shot performance on unseen tasks.
We take a 137B parameter pretrained language model and instruction-tune it on over 60 NLP tasks verbalized via natural language instruction templates. We evaluate this instruction-tuned model, which we call FLAN, on unseen task types. FLAN substantially improves the performance of its unmodified counterpart and surpasses zero-shot 175B GPT-3 on 19 of 25 tasks that we evaluate. FLAN even outperforms few-shot GPT-3 by a large margin on ANLI, RTE, BoolQ, AI2-ARC, OpenbookQA, and StoryCloze. Ablation studies reveal that number of tasks and model scale are key components to the success of instruction tuning.

[Yuli Ban]

[Yuli Ban]

[Yuli Ban]

[Yuli Ban]

[Yuli Ban]

[caltrek]

NIH-funded Modern “White Cane” Brings Navigation Assistance to the 21st Century

https://www.nih.gov/news-events/news-re ... st-century

Introduction:

(National Institute of Health) Equipped with a color 3D camera, an inertial measurement sensor, and its own on-board computer, a newly improved robotic cane could offer blind and visually impaired users a new way to navigate indoors. When paired with a building’s architectural drawing, the device can accurately guide a user to a desired location with sensory and auditory cues, while simultaneously helping the user avoid obstacles like boxes, furniture, and overhangs. Development of the device was co-funded by the National Institutes of Health’s National Eye Institute (NEI) and the National Institute of Biomedical Imaging and Bioengineering (NIBIB). Details of the updated design were published in the journal IEEE/CAA Journal of Automatica Sinica.

“Many people in the visually impaired community consider the white cane to be their best and most functional navigational tool, despite it being century-old technology,” said Cang Ye, Ph.D., lead author of the study and professor of computer science at the College of Engineering at the Virginia Commonwealth University, Richmond. “For sighted people, technologies like GPS-based applications have revolutionized navigation. We’re interested in creating a device that closes many of the gaps in functionality for white cane users.”

While there are cell phone-based applications that can provide navigation assistance – helping blind users stay within crosswalks, for example – large spaces inside buildings are a major challenge, especially when those spaces are unfamiliar. Earlier versions of Ye’s robotic cane began tackling this problem by incorporating building floorplans; the user could tell the cane whether he or she wished to go, and the cane – by a combination of auditory cues and a robotic rolling tip – could guide the user to their destination.

…Ye and colleagues have added a color depth camera to the system. Using infrared light, much like a mobile phone’s front-facing camera, the system can determine the distance between the cane and other physical objects, including the floor, features like doorways and walls, as well as furniture and other obstacles. Using this information, along with data from an inertial sensor, the cane’s onboard computer can map the user’s precise location to the existing architectural drawing or floorplan, while also alerting the user to obstacles in their path.

Study author Lingqiu Jin tests the robotic cane.
Cang Ye, VCU.