A MACHINE THAT LEARNS TO THINK

September 8, 2000

SCIENCE & ENGINEERING NEWS

San Diego, CA — “Sure, he’s cute,” reports Lisa Krieger for Mercury News. “But is he conscious?”

The 2-foot-tall NOMAD – or Neurally Organized Mobile Adaptive Device – loves bright objects, electrical impulses and high-pitched beeping sounds.

Equipped with “ears,” “eyes,” “hands,” a “brain” and wheels, NOMAD is a robot that lives in a padded playpen filled with boxes decorated with shiny spots or stripes. His antics in this playpen make him unique. He is getting smarter, much as humans do over time. Based on the different characteristics of the blocks, he learns by experience whether to embrace them or leave them alone. Scurrying from block to block, he bears a striking resemblance to a metallic kid in a candy shop.

NOMAD was built to explore one of the last great scientific frontiers of our time: how the biology of our brains gives rise to that abstract concept of consciousness.

“Consciousness . . . is not a thing, but a process or stream that is changing on a time scale of fractions of seconds,” according to 70-year-old Nobel laureate Gerald Edelman, who conceived the project, in his new book. He is director of the Neurosciences Institute, an independent research center in La Jolla.

NOMAD began life with no instructions, just a network of connections between his computer-generated neurons, or nerve cells, that are set at “birth” with particular biases.

He learns out of his experiences with blocks, spontaneous events not written anywhere or preprogrammed by computer. The process of discovering what he likes, reinforced through repeated exposure to these blocks, builds up his neuronal circuits – just as a human baby builds up its neuronal circuits.

NOMAD’s behavior is, in the simplest sense, learning by conditioning. Not a robot in the conventional sense, NOMAD is an autonomous “being” created as a tool to study how the brain controls behavior.

He is an innovation in the new field of “machine psychology.” This combination of robotics and computer-driven neural modeling strives to better understand behaviors and underlying mechanisms that allow animals and their synthetic equivalents called “animats” to adapt and survive in complicated and uncertain environments.

While progress has been made by studying neurons, brain chemistry and brain imaging, less is known about what happens in the brain to produce the thoughts and actions of which our personalities are made.

By building synthetic neural modeling devices such as NOMAD, scientists believe they will ultimately understand how brain mechanisms produce the range of behaviors associated with higher brain functions, from perception and movement to memory and creative thought.

“Our main objective is to use NOMAD as a platform to test theories of the brain,” said Jeffrey Krichmar, a neuroscientist who with engineer Jim Snook built NOMAD.

“By analyzing NOMAD’s model of the brain, we hope to better understand how the human brain works,” said Krichmar. “We also have the ability with this simulated brain to model a neurological disease.”

The birth of NOMAD was based, in large part, on work Edelman did in the 1970s which stated that neurons are the equipment used to generate consciousness and awareness. Awarded the Nobel Prize in 1972 for his work in physiology, Edelman has devoted the past 20 years to research on the brain.

Neurons talk in groups, he asserts, continually cross-referencing the information pouring into the brain through the eyes, ears and other sensory organs, and composing a recognizable picture.

Self-awareness arises from the cross talk between groups of neurons, Edelman believes. For example, if one part of the brain that recognizes red “talks” to another part of the brain that recognizes curves and shapes, the cross talk can lead to the recognition of an apple.

It is well known that new stimuli create new connections between neurons; over time, these connections strengthen. So conscious thoughts are not located in any particular area of the brain but are evoked by this complex set of neuronal interactions.

Edelman disputes the notion of famed DNA co-discover and Nobel laureate Francis Crick, who believes specific neurons are associated with consciousness. Edelman uses the automobile analogy: “If I take the distributor out of your car, it won’t run. But that doesn’t mean that the distributor is the secret of the car.”

According to Edelman, stimuli can also trigger related connections that have already been made. Thus, each state of consciousness is a neuronal re-creation of all related experiences that have come before.

Edelman believes memories are not stored in the brain in any particular location, as they might be if we had computers in our heads. Instead, memories are re-creations of the neuronal activity originally produced by the experience. He says his lab’s experiments show that each time one of the 100 billion or so neurons in the brain fires, its connections with surrounding neurons are strengthened.

The fact that the song “The Wedding March” evokes sentimental images of flowers, churches and women dressed in white means only that such a neuronal connection was made and reinforced long ago, not that your brain has filed away the two perceptions in adjacent cells.

Edelman believes the brain is not a computer, processing encoded information – because who would have encoded the information in the first place? Unlike a computer, a brain can deal with random and novel information.

Edelman has described his challenge to the computer metaphor for the brain as this: “How do you explain the first time in all of history when, in an American-style diner, one waitress says to the other, `The ham sandwich left without paying’?”

His theory differs in two ways from others that attempt to explain our mental processes.

First, it rejects the characterization of the brain as a computer, working through a series of instructions. Second, it places consciousness squarely in the physical pathways of the brain itself without recourse to physical location – or metaphysical explanation.

In his new book, “Consciousness: How Matter Becomes Imagination,” Edelman writes that the brain circuits involved in this cross talk evolved over thousands of years and continue to evolve, on a smaller scale, within each of us during our lifetimes.

This is where NOMAD comes in. The robot operates on purely Darwinian principles. He hasn’t been told what to do. He picks up boxes because he likes them or ignores them because he’s learned he doesn’t like them. He’s not controlled by a computer program.

Some blocks, if picked up, cause a current to flow in NOMAD’s “brain,” which is equipped with 200,000 “neurons” in an adjacent computer. He likes the electrical current, so is attracted to it. He learns, over time, that the blocks with the stripes always carry current.

The blocks with spots do not create current, so his neurons firing at that moment get weakened. These neurons don’t learn.

Because blocks with stripes emit low-pitched beeps, NOMAD learns to link that sound to electric currents and stripes – in essence, creating a happy mental association. In contrast, blocks with spots emit a high-pitched beep, telling NOMAD that they can be ignored.

NOMAD starts life by enthusiastically picking up every block. But over the next hour, he becomes more discriminating, picking up the striped boxes that he knows he likes.

Over time, his behavior changes.

From an adjacent room, scientists can actually peer into his brain via computer screens, which serve as the equivalent of visual and auditory cortexes.

Neurosciences Institute scientists believe that it is impossible to effectively study the brain or any part of it completely in isolation. It is crucial for the brain to be attached to a body and interact with its environment. Using NOMAD, they can perform tests with all the noise and sophistication present in the real world and still record the activity from the entire brain – something that is impossible to do with live animals.

Until recent innovations in supercomputing, NOMAD would have demanded a prohibitive amount of computer time to solve problems. Complex tasks are no longer too difficult or too time-consuming, said Snook.

“It has the basic maturity of a 1 1/2-year-old,” said Snook. “But perhaps a better comparison is a reptilian brain. It’s very instinctive, very reactive.”

While other scientists study either artificial intelligence or robotics, no one else has built a creature that incorporates both to create a “real world” environment for brain simulation, said Snook.

NOMAD’s base unit was supplied from Nomadic Technologies Inc., a robot company in Mountain View. Its “pan-tilt unit,” allowing it to scan its environment, was built by Directed Perception Inc. of Burlingame. Microchip Inc. of Chandler, Ariz., supplied the microcontrollers that act as its spinal cord, transmitting information. Intel Corp. of Santa Clara provided the PC processor.

The NOMAD work supports Edelman’s big idea, known as Neural Darwinism – caustically dubbed “Neural Edelmanism” by critic Crick. He argues that the brain works along the same principles as natural selection in evolution, a concept first explored in his book “Bright Air, Brilliant Fire,” published 10 years ago.

Darwinian selection occurs not only in the evolution of the human brain, argues Edelman, but also within each brain over its lifetime – as evidenced by NOMAD. Common neuronal routines become habitually grooved circuits, so are favored in the future.

The real brain, of course, is interconnected in a way that human-made devices cannot possibly equal. This is the challenge: The brain of an adult human weighs about three pounds and contains about 30 billion neurons. The most recently evolved part of the brain, the cerebral cortex, contains about 10 billion neurons and 1 million billion connections, or synapses.

If we counted one synapse per second, we would finish the count 32 million years from now. And if we count the number of ways in which circuits or loops of connections could be excited, the numbers become hyperastronomical: 10 followed by at least a million zeros.

Nor are a human brain’s connections exact. Connections are not identical in any two brains of the same size, even those of identical twins. Furthermore, connections in the same brain are constantly changing. Some cells are dying; others are being born. While general patterns of connections are always being created, their routes may vary over time.

Because Edelman’s blueprint of a machine-brain involves the use of silicon, it is an imperfect model of messy biology – out of which emerged our human consciousness, he admits. Nor are the memories of NOMAD or other synthetic models influenced by strong emotions of fear, love or hate, for instance.

And because they lack the verbal abilities to explain what they’re experiencing, it is impossible to know what it feels like to be NOMAD.

But NOMAD can be built to be increasingly complex, offering a model that has not yet been achieved in science. In the near future, the institute team plans to give NOMAD areas of the brain that are important for navigation. They also hope to give NOMAD a long-term memory that will enable him to remember objects, places and events and even put them in context.

While he’ll never be an Einstein, NOMAD may gain the level of consciousness shared by life’s simple creatures, say scientists.

“It has a personality, in the sense that it develops behaviors based on experiences,” said Snook. “It has a value system, and rewards itself when it finds something it likes.”

But replicating a human brain – something with freewill and emotion – may be forever out of reach, Edelman and his team concede.

============================================================

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industry updates delivered to you every week!

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion XL — were added to the benchmark suite as MLPerf continues Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing power it brings to artificial intelligence.  Nvidia's DGX Read more…

Call for Participation in Workshop on Potential NSF CISE Quantum Initiative

March 26, 2024

Editor’s Note: Next month there will be a workshop to discuss what a quantum initiative led by NSF’s Computer, Information Science and Engineering (CISE) directorate could entail. The details are posted below in a Ca Read more…

Waseda U. Researchers Reports New Quantum Algorithm for Speeding Optimization

March 25, 2024

Optimization problems cover a wide range of applications and are often cited as good candidates for quantum computing. However, the execution time for constrained combinatorial optimization applications on quantum device Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at the network layer threatens to make bigger and brawnier pro Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HBM3E memory as well as the the ability to train 1 trillion pa Read more…

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing po Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HB Read more…

Nvidia Looks to Accelerate GenAI Adoption with NIM

March 19, 2024

Today at the GPU Technology Conference, Nvidia launched a new offering aimed at helping customers quickly deploy their generative AI applications in a secure, s Read more…

The Generative AI Future Is Now, Nvidia’s Huang Says

March 19, 2024

We are in the early days of a transformative shift in how business gets done thanks to the advent of generative AI, according to Nvidia CEO and cofounder Jensen Read more…

Nvidia’s New Blackwell GPU Can Train AI Models with Trillions of Parameters

March 18, 2024

Nvidia's latest and fastest GPU, codenamed Blackwell, is here and will underpin the company's AI plans this year. The chip offers performance improvements from Read more…

Nvidia Showcases Quantum Cloud, Expanding Quantum Portfolio at GTC24

March 18, 2024

Nvidia’s barrage of quantum news at GTC24 this week includes new products, signature collaborations, and a new Nvidia Quantum Cloud for quantum developers. Wh Read more…

Alibaba Shuts Down its Quantum Computing Effort

November 30, 2023

In case you missed it, China’s e-commerce giant Alibaba has shut down its quantum computing research effort. It’s not entirely clear what drove the change. Read more…

Nvidia H100: Are 550,000 GPUs Enough for This Year?

August 17, 2023

The GPU Squeeze continues to place a premium on Nvidia H100 GPUs. In a recent Financial Times article, Nvidia reports that it expects to ship 550,000 of its lat Read more…

Shutterstock 1285747942

AMD’s Horsepower-packed MI300X GPU Beats Nvidia’s Upcoming H200

December 7, 2023

AMD and Nvidia are locked in an AI performance battle – much like the gaming GPU performance clash the companies have waged for decades. AMD has claimed it Read more…

DoD Takes a Long View of Quantum Computing

December 19, 2023

Given the large sums tied to expensive weapon systems – think $100-million-plus per F-35 fighter – it’s easy to forget the U.S. Department of Defense is a Read more…

Synopsys Eats Ansys: Does HPC Get Indigestion?

February 8, 2024

Recently, it was announced that Synopsys is buying HPC tool developer Ansys. Started in Pittsburgh, Pa., in 1970 as Swanson Analysis Systems, Inc. (SASI) by John Swanson (and eventually renamed), Ansys serves the CAE (Computer Aided Engineering)/multiphysics engineering simulation market. Read more…

Choosing the Right GPU for LLM Inference and Training

December 11, 2023

Accelerating the training and inference processes of deep learning models is crucial for unleashing their true potential and NVIDIA GPUs have emerged as a game- Read more…

Intel’s Server and PC Chip Development Will Blur After 2025

January 15, 2024

Intel's dealing with much more than chip rivals breathing down its neck; it is simultaneously integrating a bevy of new technologies such as chiplets, artificia Read more…

Baidu Exits Quantum, Closely Following Alibaba’s Earlier Move

January 5, 2024

Reuters reported this week that Baidu, China’s giant e-commerce and services provider, is exiting the quantum computing development arena. Reuters reported � Read more…

Leading Solution Providers

Contributors

Comparing NVIDIA A100 and NVIDIA L40S: Which GPU is Ideal for AI and Graphics-Intensive Workloads?

October 30, 2023

With long lead times for the NVIDIA H100 and A100 GPUs, many organizations are looking at the new NVIDIA L40S GPU, which it’s a new GPU optimized for AI and g Read more…

Shutterstock 1179408610

Google Addresses the Mysteries of Its Hypercomputer 

December 28, 2023

When Google launched its Hypercomputer earlier this month (December 2023), the first reaction was, "Say what?" It turns out that the Hypercomputer is Google's t Read more…

AMD MI3000A

How AMD May Get Across the CUDA Moat

October 5, 2023

When discussing GenAI, the term "GPU" almost always enters the conversation and the topic often moves toward performance and access. Interestingly, the word "GPU" is assumed to mean "Nvidia" products. (As an aside, the popular Nvidia hardware used in GenAI are not technically... Read more…

Shutterstock 1606064203

Meta’s Zuckerberg Puts Its AI Future in the Hands of 600,000 GPUs

January 25, 2024

In under two minutes, Meta's CEO, Mark Zuckerberg, laid out the company's AI plans, which included a plan to build an artificial intelligence system with the eq Read more…

Google Introduces ‘Hypercomputer’ to Its AI Infrastructure

December 11, 2023

Google ran out of monikers to describe its new AI system released on December 7. Supercomputer perhaps wasn't an apt description, so it settled on Hypercomputer Read more…

China Is All In on a RISC-V Future

January 8, 2024

The state of RISC-V in China was discussed in a recent report released by the Jamestown Foundation, a Washington, D.C.-based think tank. The report, entitled "E Read more…

Intel Won’t Have a Xeon Max Chip with New Emerald Rapids CPU

December 14, 2023

As expected, Intel officially announced its 5th generation Xeon server chips codenamed Emerald Rapids at an event in New York City, where the focus was really o Read more…

IBM Quantum Summit: Two New QPUs, Upgraded Qiskit, 10-year Roadmap and More

December 4, 2023

IBM kicks off its annual Quantum Summit today and will announce a broad range of advances including its much-anticipated 1121-qubit Condor QPU, a smaller 133-qu Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire