IEEE websites place cookies on your device to give you the best user experience. By using our websites, you agree to the placement of these cookies. To learn more, read our Privacy Policy.
Can we study AI the same way we study lab rats? Researchers at DeepMind and Harvard University seem to think so. They built an AI-powered virtual rat that can carry out multiple complex tasks. Then, they used neuroscience techniques to understand how its artificial “brain” controls its movements.
Today’s most advanced AI is powered by artificial neural networks—machine learning algorithms made up of layers of interconnected components called “neurons” that are loosely inspired by the structure of the brain. While they operate in very different ways, a growing number of researchers believe drawing parallels between the two could both improve our understanding of neuroscience and make smarter AI.
Now the authors of a new paper due to be presented this week at the International Conference on Learning Representations have created a biologically accurate 3D model of a rat that can be controlled by a neural network in a simulated environment. They also showed that they could use neuroscience techniques for analyzing biological brain activity to understand how the neural net controlled the rat’s movements.
The platform could be the neuroscience equivalent of a wind tunnel, says Jesse Marshall, coauthor and postdoctoral researcher at Harvard, by letting researchers test different neural networks with varying degrees of biological realism to see how well they tackle complex challenges.
“Typical experiments in neuroscience probe the brains of animals performing single behaviors, like lever tapping, while most robots are tailor-made to solve specific tasks, like home vacuuming,” he says. “This paper is the start of our effort to understand how flexibility arises and is implemented in the brain, and use the insights we gain to design artificial agents with similar capabilities.”
The virtual rodent features muscles and joints based on measurements from real-life rats, as well as vision and a sense of proprioception, which refers to the feedback system that tells animals where their body parts are and how they’re moving. The researchers then trained a neural network to guide the rat through four tasks—jumping over a series of gaps, foraging in a maze, trying to escape a hilly environment, and performing precisely timed pairs of taps on a ball.
Once the rat could successfully complete the tasks, the research team then analyzed recordings of its neural activity using techniques borrowed from neuroscience to understand how the neural network was achieving the motor control required to complete the tasks.
Because the researchers had built the AI that powered the rat, much of what they found was expected. But one interesting insight they gained was that the neural activity seemed to occur over longer time scales than would be expected if it were directly controlling muscle forces and limb movements, says Diego Aldarondo, a coauthor and graduate student at Harvard.
“This implies that the network represents behaviors at an abstract scale of running, jumping, spinning, and other intuitive behavioral categories,” he says, a cognitive model that has previously been proposed to exist in animals.

The neural network appeared to reuse some such representations across tasks, and the neural activity encoding them often took the form of sequences, a phenomenon that has been observed in both rodents and songbirds.
The researchers have open sourced the virtual rat in the hopes that other researchers will build on their findings, says Josh Merel, a coauthor and a senior research scientist at DeepMind.
“This paper is the start of our effort to understand how flexibility arises and is implemented in the brain.”
While neural networks don’t have the physiological realism of some models, Blake Richards, a neuroscientist from McGill University in Canada who was not involved in the work, says they capture enough important features of neural processing to generate useful predictions about how neural activity impacts behavior. The big contribution of this paper, he says, is to come up with a way to train these networks in a realistic manner that makes it much easier to compare against biological data.
“[The authors] are providing a platform for training [neural networks] in a realistic body and set of tasks, which will make comparisons to real brains in rodents far more valuable,” he adds.
While one must be cautious about making overly broad comparisons between artificial and biological neural networks, this approach could be a fruitful way to probe the neural underpinnings of behavior, says Stephen Scott, a neuroscientist at Queen’s University in Canada who was not involved in the work.
The complexity of recording neural activity in animals and linking it to specific behaviors means most experiments are done on relatively simple tasks in rigid experimental settings, Scott says. In contrast, the virtual rat can carry out complex, multipart behaviors like foraging that can be linked to its sensory input and neural activity with high precision.
The only problem is that collecting neural data from animals on tasks this complicated is very difficult, says Scott. He would like to see the authors test the virtual rat on some of the simpler tasks used in laboratory settings so that the neural activity patterns could be compared against those found in animals to see where they diverge.
About the author
Reader responsesThe Conversation (0)
Add comment...
Sort by
READ ALSO

Andrew Ng: Unbiggen AI

The AI pioneer says it’s time for smart-sized, “data-centric” solutions to big issues

09 Feb 2022
10 min read
​Andrew Ng listens during the Power of Data: Sooner Than You Think global technology conference in Brooklyn, New York, on Wednesday, October 30, 2019.
Andrew Ng was involved in the rise of massive deep learning models trained on vast amounts of data, but now he’s preaching small-data solutions.
Cate Dingley/Bloomberg/Getty Images
Andrew Ng has serious street cred in artificial intelligence. He pioneered the use of graphics processing units (GPUs) to train deep learning models in the late 2000s with his students at Stanford University, cofounded Google Brain in 2011, and then served for three years as chief scientist for Baidu, where he helped build the Chinese tech giant’s AI group. So when he says he has identified the next big shift in artificial intelligence, people listen. And that’s what he told IEEE Spectrum in an exclusive Q&A.
Keep Reading ↓Show less

Can a Seattle Startup Launch a Fusion Reactor Into Space?

Avalanche Energy Design’s compact power and propulsion tech raises funds—and eyebrows

13 Jun 2022
4 min read
A patent drawing of a new fusion reactor developed by Avalanche a Seattle based start-up.
Avalanche’s inertial electrostatic confinement reactor hinges on recirculating ions millions of times per second to force them to undergo fusion.
Avalanche
share this story
Practical nuclear fusion is, famously, always 10 years in the future. Except that the Pentagon recently gave an award to a tiny startup to launch a fusion power system into space in just five.
There is no shortage of organizations, from VC-backed startups to nation states, trying to realize the dream of cheap, clean, and reliable power from nuclear fusion. But Avalanche Energy Designs, based near a Boeing facility in Seattle, is even more ambitious. It is working on modular “micro fusion packs,” small enough to hold in your hand yet capable of powering everything from electric cars to spaceships.
Keep Reading ↓ Show less

AI Tool for COVID Monitoring Offers Solution for Urban Congestion

Researchers at NYU have developed an AI solution that can leverage public video feeds to better inform decision makers

09 Jun 2022
7 min read
C2SMART Center/New York University
share this story
This is a sponsored article brought to you by NYU’s Tandon School of Engineering.
In the midst of the COVID-19 pandemic, in 2020, many research groups sought an effective method to determine mobility patterns and crowd densities on the streets of major cities like New York City to give insight into the effectiveness of stay-at-home and social distancing strategies. But sending teams of researchers out into the streets to observe and tabulate these numbers would have involved putting those researchers at risk of exposure to the very infection the strategies were meant to curb.
Researchers at New York University’s (NYU) Connected Cities for Smart Mobility towards Accessible and Resilient Transportation (C2SMART) Center, a Tier 1 USDOT-funded University Transportation Center, developed a solution that not only eliminated the risk of infection to researchers, and which could easily be plugged into already existing public traffic camera feeds infrastructure, but also provided the most comprehensive data on crowd and traffic densities that had ever been compiled previously and cannot be easily detected by conventional traffic sensors.
Keep Reading ↓ Show less
{"imageShortcodeIds":["29948834"]}

Quantum Computers Exponentially Faster at Untangling Insights

Classical computers cannot overcome the “quantum advantage” in simulating chemistry and physics experiments

13 Jun 2022
2 min read
Two sets of iterations of images of a cat. On the left, the images are grayscale and even the sharpest front iteration is pixellated and incomplete. On the right, the cat is in color, sharp and each version is relatively complete.
Google Quantum AI Hook
share this story
Using Google’s Sycamore quantum processor, a new study reveals that quantum computers need exponentially fewer experiments than classical machines to reveal insights about viruses, black holes, and more.
Quantum computers can theoretically achieve a quantum advantage where they can find the answers to problems no classical computer could solve even if given thousands of years or more. This advantage can grow exponentially when a quantum computer links together a greater number of qubits—the quantum-mechanically entangled bits that such a computer uses.
Keep Reading ↓ Show less

Fixing the Future

On IEEE Spectrum's Fixing the Future podcast, host Steven Cherry talks with the brightest minds in technology about sustainable solutions for the world's biggest challenges

The Femtojoule Promise of Analog AI

To cut power by orders of magnitude, do your processing with analog circuits

20 Nov 2021
12 min read
A list of characters and letters in blue, white black and yellow.
Machine learning and artificial intelligence (AI) have already penetrated so deeply into our life and work that you might have forgotten what interactions with machines used to be like. We used to ask only for precise quantitative answers to questions conveyed with numeric keypads, spreadsheets, or programming languages: "What is the square root of 10?" "At this rate of interest, what will be my gain over the next five years?"
But in the past 10 years, we've become accustomed to machines that can answer the kind of qualitative, fuzzy questions we'd only ever asked of other people: "Will I like this movie?" "How does traffic look today?" "Was that transaction fraudulent?"
Keep Reading ↓ Show less
{"imageShortcodeIds":[]}

Flea-Sized Robots Walk a Coin-Edge-Sized Path

These crab-inspired crawlers are just half a millimeter wide

By
Payal Dhar
Payal Dhar (she/they) is a freelance journalist on science, technology, and society. They write about AI, cybersecurity, surveillance, space, online communities, games, and any shiny new technology that catches their eye. You can find and DM Payal on Twitter (@payaldhar).
11 Jun 2022
3 min read
Close-up image of crab shaped robot
Laser heating (and ambient cooling) of joints made from shape memory alloy allows this flea-sized 'bot to flex its legs and ambulate across a surface.
Northwestern University
share this story
Microrobotics engineers often turn to nature to inspire their builds. A group of researchers at Northwestern University have picked the peekytoe crab to build a remote-controlled microbot that is tiny enough to walk comfortably on the edge of a coin.
According to John A. Rogers, the lead investigator of the study, their work complements that of other scientists who are working on millimeter-scale robots, for example, worm-like structures that can move through liquid media with flagella. But to the best of his knowledge, their crab microbots are the smallest terrestrial robots—just half a millimeter wide—to walk on solid surfaces in open air.
Keep Reading ↓ Show less

Edge Learning: Adapting to a Changing Environment at the Edge

Going above and beyond the call of duty at the edge

12 May 2022
1 min read
share this story
Edge Learning is the capability of an edge device to adapt and learn to new data points/objects that have not been part of its initial training dataset. AlphaICs, a leading AI company developed edge learning PoC with a grant from a US Government Research Organization.

Video Friday: Robot Soccer Finals

Your weekly selection of awesome robot videos

By
Evan Ackerman
is a senior editor at . Since 2007, he has written over 6,000 articles on robotics and technology. He has a degree in Martian geology and is excellent at playing bagpipes.
10 Jun 2022
3 min read
Five large orange robots with wheels line up on an indoor soccer field
share this story
Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE Spectrum robotics. We also post a weekly calendar of upcoming robotics events for the next few months. Please send us your events for inclusion.
RSS 2022: 21 June–1 July 2022, NEW YORK CITY
ERF 2022: 28–30 June 2022, ROTTERDAM, NETHERLANDS
RoboCup 2022: 11–17 July 2022, BANGKOK
IEEE CASE 2022: 20–24 August 2022, MEXICO CITY
CLAWAR 2022: 12–14 September 2022, AZORES, PORTUGAL
ANA Avatar XPRIZE Finals: 4–5 November 2022, LOS ANGELES
CoRL 2022: 14–18 December 2022, AUCKLAND, NEW ZEALAND
Enjoy today’s videos!
Keep Reading ↓ Show less
A robot arm being pushed down by a very big dollar icon
Eddie Guy
LightGreen
Deep learning is now being used to translate between languages, predict how proteins fold, analyze medical scans, and play games as complex as Go, to name just a few applications of a technique that is now becoming pervasive. Success in those and other realms has brought this machine-learning technique from obscurity in the early 2000s to dominance today.
Although deep learning's rise to fame is relatively recent, its origins are not. In 1958, back when mainframe computers filled rooms and ran on vacuum tubes, knowledge of the interconnections between neurons in the brain inspired Frank Rosenblatt at Cornell to design the first artificial neural network, which he presciently described as a "pattern-recognizing device." But Rosenblatt's ambitions outpaced the capabilities of his era—and he knew it. Even his inaugural paper was forced to acknowledge the voracious appetite of neural networks for computational power, bemoaning that "as the number of connections in the network increases...the burden on a conventional digital computer soon becomes excessive."
Keep Reading ↓ Show less

Regulators Rush to Become “Crypto-Friendly”

Blockchain-conscious policies are proliferating worldwide—balancing innovation against protection

10 Jun 2022
5 min read
A collection of cryptocurrency coins with a gavel
share this story
Regulation might seem anathema to cryptocurrencies, whose entire raison d’être is to circumvent the power of governments and banks to control how people use their money. But as the technology pushes into the mainstream, there’s a growing sense that new rules are both necessary and even beneficial, leading to a boom in national experiments in crypto regulation.
Some countries clearly see cryptocurrencies as a direct threat to the power of the state and have instituted outright bans, most famously in China, which had previously been a major hub for crypto activity. But others are attempting a careful balancing act, trying to manage the potential dangers without stifling innovation in a potentially lucrative new industry.
Keep Reading ↓ Show less
share this story
As global ideation patterns change and the way they are measured evolves, the composition of the Top 100 Global Innovators changes too. The pace of innovation is breaching new frontiers of speed and complexity. The landscape of technical ideas is evolving more rapidly, and in more directions, than at any time in history.
Join us for the webinar, Top 100 Global Innovators: Calculating excellence, to hear directly from companies that made the list—Fredrik Egrelius, Patent Unit Director, Ericsson and Benjamin Bai, Vice President and Chief IP Counsel, Ant Group. These key decision-makers will provide insights into the future of real innovation that will create new economic value. Register now.

“PACMAN” Hack Can Break Apple M1’s Last Line of Defense

How many dominos could fall if this centerpiece CPU’s weakness pans out?

By
Samuel K. Moore
is the senior editor at in charge of semiconductors coverage. An IEEE member, he has a bachelor's degree in biomedical engineering from Brown University and a master's degree in journalism from New York University.
10 Jun 2022
4 min read
product photograph of a microprocessor
Apple’s M1 processor is a powerful, high-efficiency chip, though perhaps not as impervious to attacks as its initial safety record might suggest.
Apple
share this story
Apple’s M1 processor made a big splash on its November 2020 release, noteworthy for its eye-popping performance and miserly power consumption. But the value of its security may not be as obvious at first blush. A lack of serious attacks since its launch nearly two years ago indicates that its security systems, among them a last line of defense called pointer authentication codes, are working well. But its honeymoon period could possibly be coming to an end.
At the International Symposium on Computer Architecture later this month, researchers led by MIT’s Mengjia Yan will present a mode of attack that so weakens the pointer authentication code (PAC) defense that the core of a computer’s operating system is made vulnerable. And because PACs may be incorporated in future processors built from the 64-bit Arm architecture, the vulnerability could become more widespread. It’s possible that other processors are already using PACs, but the M1 was the only one available to Yan’s lab.
Keep Reading ↓ Show less

7 Revealing Ways AIs Fail

Neural networks can be disastrously brittle, forgetful, and surprisingly bad at math

21 Sep 2021
8 min read
A robot falling in the sky and a plane flying away
Chris Philpot
LightBlue
Artificial intelligence could perform more quickly, accurately, reliably, and impartially than humans on a wide range of problems, from detecting cancer to deciding who receives an interview for a job. But AIs have also suffered numerous, sometimes deadly, failures. And the increasing ubiquity of AI means that failures can affect not just individuals but millions of people.
Increasingly, the AI community is cataloging these failures with an eye toward monitoring the risks they may pose. "There tends to be very little information for users to understand how these systems work and what it means to them," says Charlie Pownall, founder of the AI, Algorithmic and Automation Incident & Controversy Repository. "I think this directly impacts trust and confidence in these systems. There are lots of possible reasons why organizations are reluctant to get into the nitty-gritty of what exactly happened in an AI incident or controversy, not the least being potential legal exposure, but if looked at through the lens of trustworthiness, it's in their best interest to do so."
Keep Reading ↓ Show less

Data Is Vulnerable to Quantum Computers That Don’t Exist Yet

A new spin-off from Alphabet has a plan to transition to postquantum cryptography

09 Jun 2022
5 min read
Pixelated key and lock that are disintegrating
Donald Iain Smith
share this story
Future quantum computers may rapidly break modern cryptography. Now a new spin-off from Google’s parent company Alphabet warns that sensitive data is already vulnerable to quantum computers that don’t exist yet, courtesy of codebreaking attacks that steal that data now and could decrypt it in the future. Therefore, it has developed a road map to help businesses, governments and other organizations begin the shift to post-quantum cryptography now.
The new startup, Sandbox AQ (which stands for AI and quantum), has already attracted clients including Mount Sinai Health System, telecommunications firm Softbank Mobile, communications technology company Vodafone Business, and Web developer Wix. It has also reeled in investors including the CIA’s venture capital arm In-Q-Tel and cybersecurity-focused investment firm Paladin Capital Group. Former Google CEO Eric Schmidt is serving as the chairman of its board of directors.
Keep Reading ↓ Show less
Transitioning AI Models from Proof-of-Concept to Full Scale Deployment
share this story
When deploying AI models to the real world many data scientists have found it difficult to move beyond the proof-of-concept stage for industrial applications such as manufacturing, healthcare, food and beverage, and more. In this webinar, NVIDIA, Landing AI, and Advantech will share our insights on resolving these AI obstacles.
Speakers:
Keep Reading ↓ Show less

From Fixing Farm Equipment to Becoming a Director at 3M

At 100, Gus Gaynor reflects on his career and volunteerism with IEEE

09 Jun 2022
5 min read
A photo of an older smiling man in glasses.
Rosie Gaynor
share this story
Gerard “Gus” Gaynor says he knew he would become an engineer when he was 7 years old, inspired by his father’s monthly Popular Mechanics magazines. As his fascination with different engineering fields grew, he set out to explore them. He visited a Ford auto manufacturing plant in his home state of Michigan, watched the printing presses roll at the Detroit Free Press plant, and listened on his crystal radio receiver to the broadcast of Charles Lindbergh’s transatlantic flight in 1927.
His enthusiasm for understanding how things work continued when, during the Great Depression, his family moved from his hometown of Detroit to a small farm in Livonia, Mich., about 32 kilometers away. He and his older brother repaired farm equipment and cars there, and they built their own chicken-plucking machine to automate the process.
Keep Reading ↓ Show less
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%