AI & Robotics News and Discussions

[Yuli Ban]

[Yuli Ban]

A certain type of artificial intelligence agent can learn the cause-and-effect basis of a navigation task during training.

Neural networks can learn to solve all sorts of problems, from identifying cats in photographs to steering a self-driving car. But whether these powerful, pattern-recognizing algorithms actually understand the tasks they are performing remains an open question.

For example, a neural network tasked with keeping a self-driving car in its lane might learn to do so by watching the bushes at the side of the road, rather than learning to detect the lanes and focus on the road’s horizon.

Researchers at MIT have now shown that a certain type of neural network is able to learn the true cause-and-effect structure of the navigation task it is being trained to perform. Because these networks can understand the task directly from visual data, they should be more effective than other neural networks when navigating in a complex environment, like a location with dense trees or rapidly changing weather conditions.

In the future, this work could improve the reliability and trustworthiness of machine learning agents that are performing high-stakes tasks, like driving an autonomous vehicle on a busy highway.

[Yuli Ban]

Introducing Pathways: A next-generation AI architecture

Pathways is a new way of thinking about AI that addresses many of the weaknesses of existing systems and synthesizes their strengths. To show you what I mean, let’s walk through some of AI’s current shortcomings and how Pathways can improve upon them.

Today's AI models are typically trained to do only one thing. Pathways will enable us to train a single model to do thousands or millions of things.

Today’s AI systems are often trained from scratch for each new problem – the mathematical model’s parameters are initiated literally with random numbers. Imagine if, every time you learned a new skill (jumping rope, for example), you forgot everything you’d learned – how to balance, how to leap, how to coordinate the movement of your hands – and started learning each new skill from nothing.

That’s more or less how we train most machine learning models today. Rather than extending existing models to learn new tasks, we train each new model from nothing to do one thing and one thing only (or we sometimes specialize a general model to a specific task). The result is that we end up developing thousands of models for thousands of individual tasks. Not only does learning each new task take longer this way, but it also requires much more data to learn each new task, since we’re trying to learn everything about the world and the specifics of that task from nothing (completely unlike how people approach new tasks).

Instead, we’d like to train one model that can not only handle many separate tasks, but also draw upon and combine its existing skills to learn new tasks faster and more effectively. That way what a model learns by training on one task – say, learning how aerial images can predict the elevation of a landscape – could help it learn another task -- say, predicting how flood waters will flow through that terrain.

We want a model to have different capabilities that can be called upon as needed, and stitched together to perform new, more complex tasks – a bit closer to the way the mammalian brain generalizes across tasks.

[Ozzie guy]

Introducing Pathways: A next-generation AI architecture

Pathways is a new way of thinking about AI that addresses many of the weaknesses of existing systems and synthesizes their strengths. To show you what I mean, let’s walk through some of AI’s current shortcomings and how Pathways can improve upon them.

Today's AI models are typically trained to do only one thing. Pathways will enable us to train a single model to do thousands or millions of things.

Today’s AI systems are often trained from scratch for each new problem – the mathematical model’s parameters are initiated literally with random numbers. Imagine if, every time you learned a new skill (jumping rope, for example), you forgot everything you’d learned – how to balance, how to leap, how to coordinate the movement of your hands – and started learning each new skill from nothing.

That’s more or less how we train most machine learning models today. Rather than extending existing models to learn new tasks, we train each new model from nothing to do one thing and one thing only (or we sometimes specialize a general model to a specific task). The result is that we end up developing thousands of models for thousands of individual tasks. Not only does learning each new task take longer this way, but it also requires much more data to learn each new task, since we’re trying to learn everything about the world and the specifics of that task from nothing (completely unlike how people approach new tasks).

Instead, we’d like to train one model that can not only handle many separate tasks, but also draw upon and combine its existing skills to learn new tasks faster and more effectively. That way what a model learns by training on one task – say, learning how aerial images can predict the elevation of a landscape – could help it learn another task -- say, predicting how flood waters will flow through that terrain.

We want a model to have different capabilities that can be called upon as needed, and stitched together to perform new, more complex tasks – a bit closer to the way the mammalian brain generalizes across tasks.

This is what I have been hyping up the last few months as potentially real AGI (just not human level as it has to learn so many things to be human level including the ability to self learn).

I would love to know your personal thoughts Yuli now that google as made a more public release.

[Yuli Ban]

It's literally just an announcement. No details of what exactly they're doing specifically that's different from, say, the hype over OpenAI. Hopefully this is indeed a multimodal, general-purpose world-modeling transformer as they're promising, but until they give us a more substantial update, it's hard to say anything that isn't just wishful hopes.

[wjfox]

‘Yeah, we’re spooked’: AI starting to have big real-world impact, says expert

Fri 29 Oct 2021 16.00 BST

A scientist who wrote a leading textbook on artificial intelligence has said experts are “spooked” by their own success in the field, comparing the advance of AI to the development of the atom bomb.

Prof Stuart Russell, the founder of the Center for Human-Compatible Artificial Intelligence at the University of California, Berkeley, said most experts believed that machines more intelligent than humans would be developed this century, and he called for international treaties to regulate the development of the technology.

“The AI community has not yet adjusted to the fact that we are now starting to have a really big impact in the real world,” he told the Guardian. “That simply wasn’t the case for most of the history of the field – we were just in the lab, developing things, trying to get stuff to work, mostly failing to get stuff to work. So the question of real-world impact was just not germane at all. And we have to grow up very quickly to catch up.”

[...]

Russell said there was still a big gap between the AI of today and that depicted in films such as Ex Machina, but a future with machines that are more intelligent than humans was on the cards.

“I think numbers range from 10 years for the most optimistic to a few hundred years,” said Russell. “But almost all AI researchers would say it’s going to happen in this century.”

https://www.theguardian.com/technology/ ... ays-expert

[Yuli Ban]

‘Yeah, we’re spooked’: AI starting to have big real-world impact, says expert

Fri 29 Oct 2021 16.00 BST

A scientist who wrote a leading textbook on artificial intelligence has said experts are “spooked” by their own success in the field, comparing the advance of AI to the development of the atom bomb.

Prof Stuart Russell, the founder of the Center for Human-Compatible Artificial Intelligence at the University of California, Berkeley, said most experts believed that machines more intelligent than humans would be developed this century, and he called for international treaties to regulate the development of the technology.

“The AI community has not yet adjusted to the fact that we are now starting to have a really big impact in the real world,” he told the Guardian. “That simply wasn’t the case for most of the history of the field – we were just in the lab, developing things, trying to get stuff to work, mostly failing to get stuff to work. So the question of real-world impact was just not germane at all. And we have to grow up very quickly to catch up.”

[...]

Russell said there was still a big gap between the AI of today and that depicted in films such as Ex Machina, but a future with machines that are more intelligent than humans was on the cards.

“I think numbers range from 10 years for the most optimistic to a few hundred years,” said Russell. “But almost all AI researchers would say it’s going to happen in this century.”

https://www.theguardian.com/technology/ ... ays-expert

What's tragic is that we indeed HAVE been discussing this for years. Heck, I remember my shock back in the mid 2010s when I discovered that the World Economic Forum was shifting focus to discuss things like AI and transhumanism because it was such a niche topic just five years prior. Now even Sunday school church groups are talking about AI. Ten years ago, you were a schizophrenic loon if you claimed we'd be where we are right now. I know because I debated some people about robots and their effect on the economy back in 2010, and even when I was under the impression it was 100 years away, it was STILL a "Star Trek fantasy."

We had all the time in the world to do something, but we didn't because the status quo was too comfortable and we rationalized often that it wasn't actually as close as it seemed. Sure, AI is getting better, but we still don't understand the brain well enough to make it human level.

Except, we don't need human level AI for it to be transformative. And the pandemic upended that comfortable status quo with which we had grown complacent. Right on time for transformative AI to arrive.

We had every chance to prepare, and we didn't.

[Yuli Ban]

It's time for the machines to take over

In the sixty-five years since John McCarthy first coined the term "artificial intelligence," one of the most surprising discoveries in the field has been that things that people do that are easy, and that should be easy for a computer to do, actually turn out to be very, very difficult for machines.

The average adult can, with relative ease, perceive a doorknob, reach out and grasp it, turn it, and open the door. The same task is still incredibly hard to engineer for even the most sophisticated robotic armatures using cutting-edge deep learning AI approaches.

By contrast, and equally surprising, machines have surpassed scientists' expectations at tasks that are hard for a human. After about five-hundred years of development by humans of the game of chess, a machine built by DeepMind called AlphaZero was able, in a matter of a couple of years, to develop to a point where it could defeat all human grandmasters, and also beat them at an even older game, Go.

That divide between the easy and the hard typically defines the debate over what's known as "human-level" AI, also known as "artificial general intelligence," or AGI, the quest to make a machine equal to a human. Many people think that the divide means that AGI won't be achieved for decades, if it is ever achieved.

AI scientist Melanie Mitchell has written, "AI is harder than we think because we are largely unconscious of the complexity of our own thought processes."

But what if the very definition of what is "human-level" intelligence is changing? What if AI is no longer measured against the quality of human thought and action in the real world, but rather the all-too-predictable behavior of spending all day staring into a smartphone?

[Yuli Ban]

The first time I interviewed Eric Schmidt, a dozen years ago when he was the C.E.O. of Google, I had a simple question about the technology that has grown capable of spying on and monetizing all our movements, opinions, relationships and tastes.

“Friend or foe?” I asked.

“We claim we’re friends,” Schmidt replied coolly.

Now that the former Google executive has a book out Tuesday on “The Age of AI,” written with Henry Kissinger and Daniel Huttenlocher, I wanted to ask him the same question about A.I.: “Friend or foe?”

“A.I. is imprecise, which means that it can be unreliable as a partner,” he said when we met at his Chelsea office. “It’s dynamic in the sense that it’s changing all the time. It’s emergent and does things that you don’t expect. And, most importantly, it’s capable of learning.

“It will be everywhere. What does an A.I.-enabled best friend look like, especially to a child? What does A.I.-enabled war look like? Does A.I. perceive aspects of reality that we don’t? Is it possible that A.I. will see things that humans cannot comprehend?”

I agree with Elon Musk that when we build A.I. without a kill switch, we are “summoning the demon” and that humans could end up, as Steve Wozniak said, as the family pets. (If we’re lucky.)

[raklian]

The first time I interviewed Eric Schmidt, a dozen years ago when he was the C.E.O. of Google, I had a simple question about the technology that has grown capable of spying on and monetizing all our movements, opinions, relationships and tastes.

“Friend or foe?” I asked.

“We claim we’re friends,” Schmidt replied coolly.

Now that the former Google executive has a book out Tuesday on “The Age of AI,” written with Henry Kissinger and Daniel Huttenlocher, I wanted to ask him the same question about A.I.: “Friend or foe?”

“A.I. is imprecise, which means that it can be unreliable as a partner,” he said when we met at his Chelsea office. “It’s dynamic in the sense that it’s changing all the time. It’s emergent and does things that you don’t expect. And, most importantly, it’s capable of learning.

“It will be everywhere. What does an A.I.-enabled best friend look like, especially to a child? What does A.I.-enabled war look like? Does A.I. perceive aspects of reality that we don’t? Is it possible that A.I. will see things that humans cannot comprehend?”

I agree with Elon Musk that when we build A.I. without a kill switch, we are “summoning the demon” and that humans could end up, as Steve Wozniak said, as the family pets. (If we’re lucky.)

Before too long, we won't be able to equally coexist with them for obvious reasons. Instead, we'll have to make the choice whether to join them or not.

[weatheriscool]

A new machine-learning system helps robots understand and perform certain social interactions
https://techxplore.com/news/2021-11-mac ... tions.html
by Adam Zewe, Massachusetts Institute of Technology

Robots can deliver food on a college campus and hit a hole in one on the golf course, but even the most sophisticated robot can't perform basic social interactions that are critical to everyday human life.

MIT researchers have now incorporated certain social interactions into a framework for robotics, enabling machines to understand what it means to help or hinder one another, and to learn to perform these social behaviors on their own. In a simulated environment, a robot watches its companion, guesses what task it wants to accomplish, and then helps or hinders this other robot based on its own goals.

The researchers also showed that their model creates realistic and predictable social interactions. When they showed videos of these simulated robots interacting with one another to humans, the human viewers mostly agreed with the model about what type of social behavior was occurring.

Enabling robots to exhibit social skills could lead to smoother and more positive human-robot interactions. For instance, a robot in an assisted living facility could use these capabilities to help create a more caring environment for elderly individuals. The new model may also enable scientists to measure social interactions quantitatively, which could help psychologists study autism or analyze the effects of antidepressants.

"Robots will live in our world soon enough and they really need to learn how to communicate with us on human terms. They need to understand when it is time for them to help and when it is time for them to see what they can do to prevent something from happening. This is very early work and we are barely scratching the surface, but I feel like this is the first very serious attempt for understanding what it means for humans and machines to interact socially," says Boris Katz, principal research scientist and head of the InfoLab Group in the Computer Science and Artificial Intelligence Laboratory (CSAIL) and a member of the Center for Brains, Minds, and Machines (CBMM).

Joining Katz on the paper are co-lead author Ravi Tejwani, a research assistant at CSAIL; co-lead author Yen-Ling Kuo, a CSAIL Ph.D. student; Tianmin Shu, a postdoc in the Department of Brain and Cognitive Sciences; and senior author Andrei Barbu, a research scientist at CSAIL and CBMM. The research will be presented at the Conference on Robot Learning in November.

[Yuli Ban]

[Ozzie guy]

This week in AI becoming a professional news source.

This week in AI is now charging low fees turning it into a professional paper and podcast.
In theory this should improve content quality.
Note I don't think the podcast is weekly like the article but it might be.

Personally I am tempted to subscribe maybe I could make it my weekly AI news and then stop deliberately googling around for news (unless there is a giant breakthrough).

https://lastweekin.ai/p/last-week-in-ai ... al-message

[Yuli Ban]

At its fall 2021 GPU Technology Conference (GTC) today, Nvidia announced that it’s making Megatron 530B, one of the world’s largest language models, available to enterprises for training to serve new domains and languages. First detailed in early October, Megatron 530B — also known as Megatron-Turing Natural Language Generation (MT-NLP) — contains 530 billion parameters and achieves high accuracy in a broad set of natural language tasks, including reading comprehension, commonsense reasoning, and natural language inference.

“Today, we provide recipes for customers to build, train, and customize large language models, including Megatron 530B. This includes scripts, code, and 530B untrained model. Customers can start from smaller models and scale up to larger models as they see fit,” Nvidia VP of AI software product management Kari Briski told VentureBeat via email. “Our researchers [worked] together with Microsoft [to train] the Megatron 530B model in six weeks.”

[Yuli Ban]

NVIDIA has launched a follow-up to the Jetson AGX Xavier, its $1,100 AI brain for robots that it released back in 2018. The new module, called the Jetson AGX Orin, has six times the processing power of Xavier even though it has the same form factor and can still fit in the palm of one's hand. NVIDIA designed Orin to be an "energy-efficient AI supercomputer" meant for use in robotics, autonomous and medical devices, as well as edge AI applications that may seem impossible at the moment.

The chipmaker says Orin is capable of 200 trillion operations per second. It's built on the NVIDIA Ampere architecture GPU, features Arm Cortex-A78AE CPUs and comes with next-gen deep learning and vision accelerators, giving it the ability to run multiple AI applications. Orin will give users access to the company's software and tools, including the NVIDIA Isaac Sim scalable robotics simulation application, which enables photorealistic, physically-accurate virtual environments where developers can test and manage their AI-powered robots. For users in the healthcare industry, there's NVIDIA Clara for AI-powered imaging and genomics. And for autonomous vehicle developers, there's NVIDIA Drive.

The company has yet to reveal what the Orin will cost, but it intends to make the Jetson AGX Orin module and developer kit available in the first quarter of 2022. Those interested can register to be notified about its availability on NVIDIA's website. The company will also talk about Orin at NVIDIA GTC, which will take place from November 8th through 11th.

[Yuli Ban]

NVIDIA NeMo is a framework for building, training, and fine-tuning GPU-accelerated speech and natural language understanding (NLU) models with a simple Python interface. Using NeMo, developers can create new model architectures and train them using mixed- precision compute on Tensor Cores in NVIDIA GPUs through easy-to-use application programming interfaces (APIs).

NeMo Megatron is a part of the framework that provides parallelization technologies such as pipeline and tensor parallelism from the Megatron-LM research project for training large-scale language models.

With NeMo, you can build models for real-time automated speech recognition (ASR), natural language processing (NLP), and text-to-speech (TTS) applications such as video call transcriptions, intelligent video assistants, and automated call center support across healthcare, finance, retail, and telecommunications.

[weatheriscool]

Braingate Translates in Realtime Thoughts to Text With 94% Accuracy
November 10, 2021 by Brian Wang

https://www.nextbigfuture.com/2021/11/b ... uracy.html

A paper in Nature reports that a Braingate Brain Computer Interface (BCI) enabled a man paralyzed from the neck down to have his thoughts translated to text with 94% accuracy.

An intracortical BCI tdecodes attempted handwriting movements from neural activity in the motor cortex and translates it to text in real-time, using a recurrent neural network decoding approach. The patient paralysed from spinal cord injury, achieved typing speeds of 90 characters per minute with 94.1% raw accuracy online, and greater than 99% accuracy offline with a general-purpose autocorrect. These typing speeds exceed those reported for any other BCI, and are comparable to typical smartphone typing speeds of individuals in the age group of our participant (115 characters per minute). Finally, theoretical considerations explain why temporally complex movements, such as handwriting, may be fundamentally easier to decode than point-to-point movements. This is a new approach for BCIs and demonstrates the feasibility of accurately decoding rapid, dexterous movements years after paralysis.

Neural activity recorded during the attempted handwriting of 1,000 sentences (43,501 characters) over 10.7 hours.

BrainGate is an interdisciplinary research team involving Brown University, Massachusetts General Hospital, Stanford University, Case Western Reserve University and Providence VA Medical Center. With a focus on practical applications and reliability, the team aims to develop assistive BCI technology to restore independence and communication in individuals with impaired movement abilities.

[Nanotechandmorefuture]

Braingate Translates in Realtime Thoughts to Text With 94% Accuracy
November 10, 2021 by Brian Wang

https://www.nextbigfuture.com/2021/11/b ... uracy.html

A paper in Nature reports that a Braingate Brain Computer Interface (BCI) enabled a man paralyzed from the neck down to have his thoughts translated to text with 94% accuracy.

An intracortical BCI tdecodes attempted handwriting movements from neural activity in the motor cortex and translates it to text in real-time, using a recurrent neural network decoding approach. The patient paralysed from spinal cord injury, achieved typing speeds of 90 characters per minute with 94.1% raw accuracy online, and greater than 99% accuracy offline with a general-purpose autocorrect. These typing speeds exceed those reported for any other BCI, and are comparable to typical smartphone typing speeds of individuals in the age group of our participant (115 characters per minute). Finally, theoretical considerations explain why temporally complex movements, such as handwriting, may be fundamentally easier to decode than point-to-point movements. This is a new approach for BCIs and demonstrates the feasibility of accurately decoding rapid, dexterous movements years after paralysis.

Neural activity recorded during the attempted handwriting of 1,000 sentences (43,501 characters) over 10.7 hours.

BrainGate is an interdisciplinary research team involving Brown University, Massachusetts General Hospital, Stanford University, Case Western Reserve University and Providence VA Medical Center. With a focus on practical applications and reliability, the team aims to develop assistive BCI technology to restore independence and communication in individuals with impaired movement abilities.

Now we're talking!

[Yuli Ban]

One afternoon in late September, a yellow four-legged robot called Spot pranced and pirouetted on a replica of a dingy subway platform that had been constructed inside a vast limestone cavern burrowed beneath the Louisville Zoo.

Spot snooped around the platform, inhaling data through cameras and sensors arrayed on its vacuum-cleaner-size torso. The robot's little feet kept darting perilously close to the edge of the platform, then back to safety. Finally, apparently satisfied by what it had learned, Spot nimbly descended a staircase to make further investigations on the track bed. Back on the now-deserted platform, a poster on the wall declared: "The Future Is Now."

And what a future.

[Yuli Ban]

Even as economies struggle with the chaos of the pandemic, the AI startup space continues to grow stronger with increased investments and M&A deals.

According to the latest State of AI report from CB Insights, the global funding in the segment has seen a significant surge, growing from $16.6 billion across 588 deals in Q2 2021 (figures show $20B due to the inclusion of two public subsidiary fundings) to $17.9 billion across 841 deals in the third quarter. Throughout the year (which is yet to end), AI startups around the world raised $50 billion across 2000+ deals with 138 mega-rounds of 100+ million. As much as $8.5 billion of the total investment went into healthcare AI, $3.1 billion went into fintech AI, while $2.6 billion went into retail AI.

The findings show how AI has become a driving force across nearly every industry and is drawing significant attention from VCs, CVCs, and other investors.