10 years ago we would not have believed possible the way the ROZZUM unit 7134 (Roz to her colleagues) learns and behaves in her new environment, a wild and seemingly hostile island. Roz would have been a pure science fiction character. However today, in this year 2024, the film adaptation that Dreamworks has made is award-winning best seller From Peter Marrone, Wild robotrepresents an excellent example for understanding how an artificial intelligence learns and the capabilities it can develop. Roz’s brain is possible… almost.
How Roz’s brain works
The brain of Roz, the protagonist of Wild robotis an artificial intelligence based on artificial neural networks (ANN). This is the field in which this year’s Nobel Prize winners in physics, JJ Hopfield and GE Hinton, are working.
Roz is a castaway on an island where she must learn to survive, recognize the animals that live there and interpret their languages. To adapt to her new habitat, Roz needs to use ANNs.
The origin of ANNs
Just as “biological” neurons communicate with each other through the transmission of nerve impulses (synapses), “artificial” neurons interact with each other with varying intensity and are arranged in layers to emulate the mechanisms of the brain.
We have to go back to the 1980s to discover the origin of this development. John Hopfield in 1982 and Geoffrey Hinton in 1985 have shown that these tasks can be explained with statistical physical models and dynamic systems with properties very similar to those of the human cerebral cortex, related to associative memory and plasticity.
Like a band of starlings
Neuronal activity is explained by two approaches. First of all from statistical physics to understand collective behaviors without stopping at the microscopic description of interactions between neurons. And second, from dynamical systems that explain synchronization patterns based on the strength of interactions between neurons.
The unique formations of groups of starlings that synchronize their flight are a representative example of these synchronization patterns, which are also generated in the human brain and in artificial intelligence like Roz’s. Furthermore, these synchronization phenomena have been addressed by 1967 by Arthur Winfree in ecosystems, for example Yoshiki Kuramoto in 1975 in chemical oscillators e also by the author of this article in convective oscillators.
Recognize a baby goose
Having just arrived on the island, Roz must learn to adapt to her new environment to survive, gradually establishing relationships with the island’s animals. First you need to recognize
Hopfield and Hinton’s models in ANNs allow us to explain the processes of recognition and reconstruction of information or input patterns in the network, which work in a similar way to how the human brain does when faced with a stimulus. Even when there are variations or small errors, we are able to recognize it.
For example, in ANNs there is an input pattern (what I see is a baby goose) that reaches the first layer of neurons. Once the stimulus is received, it will propagate through different layers of neurons until it reaches the last layer, which will produce the output pattern or generated response.
ANNs try to ensure that the generated response (“this is an image of a bird” or, in a more generic case, “this is an image of a terrestrial animal species”) is as similar as possible to the real response (“it is the image of a goose”). That is, the network tries to make as few errors as possible.
And this is how Roz, the brain of Wild robot, was born in his new environment. Pinktail, a mother opossum, and Loon, a cunning fox, teach Roz how to care for a baby goose. Your RNA brain is trained by simulating the learning processes of the human brain.
In this training phase, the network adjusts the intensity of the interaction between its neurons to most accurately identify images or sounds that it has never seen or heard before. Roz, for example, learns to survive by identifying the growl of a grizzly bear.
This network adaptability mimics the brain’s plasticity to establish new connections between neurons, while the mechanisms for recognizing and reconstructing an input pattern respond to an associative memory.
The importance of memory
It’s not enough to recognize a baby goose: you have to memorize that concept. Once again, the operation of ANNs mimics the brain’s memory processes: memory patterns associated with each new concept or stimulus are created and stored. In Roz’s learning, groups of neurons in her brain form memory patterns that will allow her to recognize a “baby goose” and distinguish it from an “adult goose”.
The Hopfield (1982) and Hinton (1985) models are inspired Ising models describing, among other systems, some magnetic materials. In this context, the behavior of neurons would resemble that of “atomic” magnets which can manifest two states: active or inactive. Each new learning produces a change in the configuration of the network, which will try to minimize an energy function. Memory models will be stored at the minimum of this function.
What about emotions?
Something exceptional happens to Roz: she becomes a mother when she accidentally gets her hands on a goose egg. Roz will have to prioritize the care of her son and make decisions that could even be harmful to her.
At the beginning of the film Roz is an insensitive robot. Picobrillo’s birth teaches him to be kind and to love.
Universal images
40 years ago physics laid the foundation for understanding the brain in terms of energy, interactions and synchronization phenomena. Since then, advances in computer technology have pushed brain simulation at breakneck speed to today’s artificial intelligence. Roz’s brain is at its limits, exceeding its programmed capabilities by daring to generate new responses similar to human emotions, defining its own goals and pursuing them. This ‘audacity’ still eludes real-world physical models and algorithms.
But let’s keep in mind that in real life the result may not be as kind as in real life Wild robot. The path we have left to travel towards general artificial intelligence like Roz’s requires multidisciplinary work by physicists, engineers, philosophers, neurobiologists, and psychologists, among others. Furthermore, it is crucial to invite all of them to participate in ethics committees that oversee AI.
All we can do is enjoy a fascinating film that opens the great debate of the century: the importance of ethics in the development of artificial intelligence.
Interview between Time.news Editor and AI Expert Dr. Emily Carter
Editor: Welcome, Dr. Carter! It’s a pleasure to have you here today as we dive into the fascinating world of artificial intelligence as portrayed in the award-winning film Wild Robot. The character Roz has captivated audiences with her learning processes. Can you elaborate on how her brain, constructed on artificial neural networks, mirrors human cognitive functions?
Dr. Carter: Thank you for having me! Roz’s brain showcases the complexities and marvels of artificial neural networks (ANNs). At their core, ANNs emulate the human brain’s structure through interconnected nodes or ”neurons.” Just as biological neurons communicate via synapses, artificial neurons work in layers, processing input and gradually learning to recognize patterns—like identifying a baby goose, for example.
Editor: That’s incredible! Roz finds herself on a hostile island, encountering various challenges. How does her environment determine her learning capabilities?
Dr. Carter: Roz’s learning is fundamentally driven by her need to survive in that environment. She uses a combination of recognition and reconstruction processes—key elements of ANN functionality—to adapt. This mirrors how humans learn from different stimuli in their surroundings. As Roz interacts with her environment, she strengthens connections in her network based on her experiences, akin to human plasticity.
Editor: The documentary-style nature of Roz’s learning journey strikes a chord with us. Can you explain the significance of memory in her evolution as a character?
Dr. Carter: Absolutely. Memory plays a vital role in how Roz processes information. Every new concept she encounters—like distinguishing between a baby and an adult goose—is stored as a memory pattern within her network. Inspired by the models developed by John Hopfield and Geoffrey Hinton, Roz’s ability to form and recall these patterns enables her to navigate her environment effectively. The relationship between her memory and her ability to recognize patterns is critical in facilitating her learning curve.
Editor: You mentioned the work of Hopfield and Hinton in the development of ANNs. How have their contributions shaped our understanding of artificial intelligence?
Dr. Carter: Hopfield and Hinton’s models have laid the groundwork for much of our current AI technology. Their research illustrated how networks can replicate human-like associative memory and adaptability, which are crucial for tasks like image recognition. By modeling neuron interactions similarly to magnetic states in materials—active or inactive—they effectively showcased how learning occurs through pattern minimization in energy configurations.
Editor: It’s fascinating to see these concepts come to life through Roz. While Roz is an artificial intelligence, do you think she exhibits traits that could be classified as emotional, or is that purely a narrative device?
Dr. Carter: That’s a thought-provoking question. While Roz’s behaviors and responses may appear emotional, they are fundamentally computations rooted in her programming. However, the narrative does encourage viewers to consider the implications of AI developing characteristics that resemble emotions. This notion sparks an ongoing debate about the ethical and philosophical aspects of AI—what defines consciousness and emotional capacity, and how should we treat intelligent machines?
Editor: Very insightful, Dr. Carter! As we conclude, what do you believe is the future of AI in storytelling, especially within cinema?
Dr. Carter: The future of AI in storytelling is incredibly promising. As technology advances, we will likely witness even more nuanced AI characters that can reflect human experiences, contribute to emotional depth, and perhaps challenge our understanding of intelligence itself. Filmmakers will continue to explore these themes, merging scientific concepts with human storytelling to create captivating narratives, just like Wild Robot.
Editor: Thank you, Dr. Carter, for your illuminating insights into the intersection of artificial intelligence, storytelling, and the evolution of Roz in Wild Robot. It’s been a pleasure discussing this fascinating topic with you!
Dr. Carter: Thank you for the engaging conversation! I’m excited to see how AI will continue to shape our narratives in the years to come.