Diffraction-casting-optical-computing-AI-applications

Diffraction Casting: Revolutionizing Optical Computing for Next-Gen AI Applications

Introduction to Optical Computing

• The Need for Powerful Computing Solutions: The growing complexity of applications like artificial intelligence demands increasingly powerful, energy-intensive computers. Optical computing offers a potential solution for boosting speed and efficiency, but challenges remain in its practical implementation.

Understanding Diffraction Casting

• What is Diffraction Casting?: Introducing diffraction casting--a new design framework that tackles the current drawbacks and introduces groundbreaking concepts to optical computing, enhancing its potential for next-generation devices.

Challenges of Traditional Electronic Computing

• Limitations of Current Technology: Whether it's your smartphone or laptop, today's computing devices are all built on electronic technology. Yet, this approach has its drawbacks-chief among them, substantial heat production as performance rises and the looming limits of current fabrication techniques.

As a result, scientists are pursuing alternative computational techniques that aim to overcome these limitations and, ideally, provide innovative features and advantages.

The Promise of Optical Computing

• Harnessing the Speed of Light: One potential solution lies in optical computing, a concept that has been around for decades but has yet to achieve commercial success.

Optical computing fundamentally harnesses the speed of light waves and their complex interactions with various optical materials, all without generating heat. Additionally, light waves can pass through materials simultaneously without interference, theoretically enabling a highly parallel, power-efficient, and high-speed computing system.

Historical Context: Shadow Casting in Optical Computing

• The Early Attempts: "During the 1980s, researchers in Japan examined a method of optical computing called shadow casting, which could carry out simple logical operations. Nonetheless, their approach utilized bulky geometric optical designs akin to the vacuum tubes used in early digital computing. Although these methods were theoretically sound, they lacked the necessary flexibility and integration ease for practical utility," explained Associate Professor Ryoichi Horisaki of the Information Photonics Lab at the University of Tokyo.

Advancements through Diffraction Casting

• Enhancing Optical Elements: We present an optical computing approach known as diffraction casting, which enhances the concept of shadow casting. While shadow casting relies on light rays interacting with various geometries, diffraction casting leverages the inherent properties of light waves. This results in more spatially efficient and functionally adaptable optical elements that can be extended as needed for universal computing applications.

Numerical Simulations and Results

• Testing the Framework: "We executed numerical simulations that demonstrated very favorable results, using small black-and-white images measuring 16 by 16 pixels, which are even smaller than the icons displayed on a smartphone."

An All-Optical System for Data Processing

• From Optical to Digital: Horisaki and his team suggest an all-optical system, meaning that is only converts the final output into an electronic and digital format; all preceding stages of the system operate entirely optically. Their research has been published in Advanced Photonics.

Application and Representation of Data

• Utilizing Images as Data Sources: Their concept involves utilizing an images as a data source, which naturally indicates that this system could be applied to image processing. However, it can also represent other types of data, particularly that utilized in machine learning systems, in graphical form, combining the source image with a series of additional images that depict stages in logic operations.

Layers in Optical Processing

• A Visual Analogy: Imagine it as layers in an image editing software like Adobe Photoshop; you begin with an input layer, which is the source image, and then additional layers can be added on top. These layers can obscure, manipulate, or transmit information from the layer below. The final output--the top layer--results from the processing of this combination of layers.

The Process of Diffraction Casting

• Creating Digital Images: In this context, light will pass through these layers, crating an image--hence the term 'casing' in diffraction casting---on a sensor. This image will subsequently be converted into digital data for storage or display to the user.

Future Potential and Commercial Viability

• An Auxiliary Component in Computing: "Diffraction casting represents merely one component in a theoretical computer founded on this principle. It may be more appropriate to view it as an auxiliary element rather than a complete substitute for existing systems, similar to how graphical processing units serve as specialized components for graphics, gaming and machine learning tasks," stated lead author Ryosuke Mashiko.

Estimating Time for Commercial Readiness

"I estimate that it will take approximately 10 years before this technology becomes commercially viable, as significant effort is required for physical implementation, which despite being based on solid research, has not yet been developed."

Conclusion: The Road Ahead for Diffraction Casting

• Expanding into Quantum Computing: "As this stage, we are able to showcase the applicability of diffraction casting in carrying out the 16 essential logic operations foundational to much information processing. Moreover, our system has the potential to evolve into the burgeoning area of quantum computing, which transcends conventional methods. The future will determine the results."

Source

artificial-general-intelligence-experts-development-claims

Artificial General Intelligence: Experts Warn Against Overestimating Development Potential

The Current Hype Surrounding AI

Claims of Inevitable Advancement

AI surpassing the human brain is deemed inevitable by employees at OpenAI, Google DeepMind and similar tech companies. However, researchers from Radboud University and other institutions have released new findings, now published in Computational Brain & Behavior, showing that claims are exaggerated and unlikely to materialize.

Expert Insights on AGI Development

The Impossibility of Human-Level Cognition

Iris van Rooij, lead author and professor of Computational Cognitive Science at Radboud University, asserts that developing artificial general intelligence (AGI) with human-level cognition is "Impossible." As head of the Cognitive Science and AI department, she leads this critical analysis.

Theoretical vs. Practical Feasibility

"While some contend that AGI is achievable in theory, arguing that it's just a matter of time before computers can think like humans, theory alone isn't sufficient to make it a reality. Our paper illustrates why pursuing this goal is misguided and a waste of valuable resources," explains van Rooij.

Endless Possibilities Powered by Finite Capabilities

A Thought Experiment on Ideal Conditions

In their publication, the researchers propose a thought experiment that envisions the development of AGI under perfect conditions.

Limitations Even Under Optimal Conditions

Olivia Guest, co-author and assistant professor at Radboud University in Computational Cognitive Science, remarks, "For the sake of argument, we propose that engineers have access to perfect datasets and the most efficient machine learning tools. But despite these advantages, no viable method exists to deliver what big tech companies claim."

The Challenge of Replicating Human Cognition

This is due to the fact that cognition--the capacity to observe, learn, and acquire new insights--is extremely challenging to replicate in AI at the scale seen in the human brain.

As van Rooij explains, "In conversation, you might recall a statement from fifteen minutes ago, a year ago, or even a distant memory from your childhood. Such knowledge could be vital in moving the conversation forward, and people do this naturally."

"There will never be enough computational capacity to develop AGI using machine learning, as we would deplete our natural resources long before reaching that level of advancement," adds Olivia Guest.

The Importance of Critical AI Literacy

A Collaborative Effort Across Disciplines

This paper represents a collaborative effort among researchers from Radboud University, Aarhus University, The University of Amsterdam, Memorial University of Newfoundland, and the University of Bayreuth, integrating insights from Cognitive Science, Neuroscience, Philosophy, and Computer Science.

Risks of Misunderstanding AI Capabilities

The researchers caution that the current enthusiasm for AI poses a risk of misapprehending the capabilities of both humans and AI systems.

Evaluating AI Claims Through Cognitive Science

It is surprising how few recognize the importance of cognitive science in evaluating assertions about AI capabilities. "We tend to exaggerate what computers can do, while greatly undervaluing the capabilities of human cognition," explains van Rooij.

Fostering a Deeper Understanding of AI

Developing critical AI literacy is essential to empower individuals with the ability to discern the viability of assertions made by large tech corporations. If a new firm claims to have invented a device that can bring about world peace with a mere button press, it would be prudent to question that claim.

"Why do we readily accept the assurances of large tech companies motivated by profit? Our aim is to foster a deeper understanding of AI systems, enabling everyone to critically assess the claims made by the technology sector."

Source

hydrogel-based brain sensors for enhanced adhesion

Innovative Brain Sensor Enhances Transcranial Focused Ultrasound for Neurological Disorders

Introduction to Transcranial Focused Ultrasound

This non-invasive technique, known as transcranial focused ultrasound, employs high-frequency sound waves to stimulate targeted brain regions, offering a potential breakthrough in treating neurological disorders like drug-resistant epilepsy and recurrent tremors.

Development of the Innovative Sensor

Researchers at Sungkyunkwan University (SKKU), IBS, and the Korea Institute of Science and Technology have designed an innovative sensor for transcranial focused ultrasound. Their study, featured in Nature Electronics, describes a flexible sensor that conforms to cortical surfaces, facilitating neural signal detection and low-intensity ultrasound-based brain stimulation.

Challenges with Previous Brain Sensors

"Previous efforts to develop brain sensors struggled to achieve precise signal measurement because they couldn't fully adapt to the brain's complex folds," remarked Donghee Son, the supervising author of the study, in an interview with Tech Xplore.

"The inability to precisely analyze the entire brain surface limited accurate diagnosis of brain lesions. Despite the innovative ultra-thin brain sensor developed by Professors John A. Rogers and Dae-Hyeong Kim, it encountered difficulties in tightly adhering to areas with severe curvature."

Limitations of Existing Sensors

The brain sensor created by Professors Rogers and Kim demonstrated improved precision in collecting surface-level measurements. However, it exhibited notable limitations, including difficulty adhering to areas with significant curvature and a tendency to shift from its attachment point due to micro-movements and cerebral spinal fluid flow.

The limitations observed reduce the sensor's suitability for clinical use, as they hinder its ability to capture brain signals reliably in specific areas over longer duration's.

The New Sensor Design

To overcome these challenges, Son and colleagues developed a new sensor designed for better adhesion to curved brain surfaces, ensuring stable, long-term data collection.

"The sensor we engineered is capable of conforming to even the most curved brain regions, ensuring a firm attachment to brain tissue," said Son. "This strong bond allows for long-term, precise measurement of brain signals from specific areas."

ECoG Sensor Features

The ECoG sensor designed by Son and his team attaches firmly to brain tissue, ensuring no voids are created. This feature markedly decreases noise from external mechanical movements.

"This feature plays a crucial role in improving the efficacy of epilepsy treatment using low-intensity focused ultrasound (LIFU)," noted Son. "Although ultrasound is recognized for its ability to reduce epileptic activity, the variability in patient conditions and individual differences present significant obstacles in customizing treatments."

Personalized Ultrasound Stimulation Therapies

Recently, numerous research teams have been focused on developing personalized ultrasound stimulation therapies for epilepsy and various neurological disorders. To tailor these treatments to the specific needs of each patient, it is essential to measure their brain waves in real-time while simultaneously stimulating targeted brain regions.

"Traditional sensors attached to the brain surface faced challenges in this regard, as the vibrations induced by ultrasound generated considerable noise, hindering real-time monitoring of brain waves," stated Son.

"This limitation significantly hindered the development of personalized treatment strategies. Our sensor substantially minimizes noise, facilitating effective epilepsy treatment through tailored ultrasound stimulation."

Structure of the Shape-Morphing Sensor

Son and his colleagues developed a shape-morphing brain sensor with three primary layers. These consist of a hydrogel-based layer for both physical and chemical bonding with tissue, a self-healing polymer layer that adjusts its form to fit the surface beneath, and a thin, stretchable layer containing gold electrodes and interconnects.

Son noted that when the sensor is positioned on the brain surface, the hydrogel layer activates a gelation process that establishes a strong and instant bond with the brain tissue.

"Subsequently, the self-healing polymer substrate starts to deform, adapting to the curvature of the brain, which enhances the contact area between the sensor and the tissue over time. Once the sensor has completely conformed to the brain's contours, it is primed for operation."

Advantages of the New Sensor

The sensor created by this research team offers multiple advantages compared to other brain sensors developed in recent years. Notably, it can securely attach to brain tissue while adapting its shape to conform tightly to surfaces, regardless of their curvature.

By conforming to the contours of curved surfaces, the sensor effectively reduces vibrations generated by external ultrasound stimulation. This capability enables physicians to accurately measure brain wave activity in patients, both in standard conditions and during ultrasound procedures.

Future Applications

According to Son, we foresee this technology being applicable not only for epilepsy management but also for the diagnosis and treatment of multiple brain disorders. The most crucial aspect of our research is the synergy between tissue-adhesive technology, which enables robust adhesion to brain tissue, and shape-morphing technology, allowing the sensor to conform precisely to the brain's surface without leaving any gaps.

Testing and Future Development

To date, the novel sensor engineered by Son and his team has undergone testing on conscious, living rodents. The results obtained were exceptionally promising, demonstrating the team's ability to accurately measure brain waves and mange seizures in these animals.

The researchers aim to expand the sensor's capabilities by developing a high-density array based on their initial design. Upon successful completion of clinical trials, this enhanced sensor could be utilized to diagnose and treat epilepsy and other neurological disorders, potentially adancing the effectiveness of prosthetic technologies.

With 16 electrode channels currently integrated into our brain sensor, Son highlighted an area ripe for improvement concerning high-resolution mapping of brain signals.

"Taking this into consideration, our strategy involves significantly augmenting the number of electrodes to enable comprehensive and high-resolution brain signal analysis. We also aspire to devise a minimally invasive implantation technique for the brain sensor on the surface of the brain, aiming for its application in clinical research."

Source

Reliability issues in large language models explored

Researchers Examine Accuracy and Transparency of Leading AI Chatbots: A Closer Look

Introduction to the Study

Researchers from the Universitat Politecnica de Valencia in Spain have discovered that as Large Language Models expand size and complexity, they are less inclined to acknowledge their lack of knowledge to users.

Study: Examining AI Chatbots

The researchers, in their Nature study, assessed the newest versions of three popular AI chatbots, examining their response accuracy and users' effectiveness in recognizing incorrect information.

Increased Reliance on LLMs

As LLMs gain widespread adoption, users have increasingly relied on them for tasks like writing essays, composing poems or songs, solving mathematical problems, and more, Consequently, accuracy has become a growing concern.

Study Objective: Evaluating AI Accuracy

In this new study, the researchers sought to determine whether popular LLMs improve in accuracy with each update and how they respond when they provide incorrect answers.

AI Chatbots Assessed: BLOOM, LLaMA and GPT

In order to assess the accuracy of three leading LLMs--BLOOM, LLaMA and GPT---the researchers presented them with thousands of questions and compared the answers to those generated by earlier versions in response to the same prompts.

Diverse Themes Tested

The researchers also diversified the themes, encompassing math, science, anagrams, and geography, while evaluating the LLMs' capabilities to generate text or execute tasks like list ordering. Each question was initially assigned a level of difficulty.

Key Findings: Accuracy and Transparency

The researchers discovered that accuracy generally improved with each new iteration of the chatbots. However, they observed that as question difficulty increased, accuracy declined, as anticipated.

Transparency Decreases with Size

Interestingly, they noted that as LLMs became larger and more advanced, they tended to be less transparent about their ability to provide correct answers.

Behavioral Shift in AI Chatbots

In previous iterations, most LLMs would inform users that they were unable to find answers or required additional information. However, in the latest versions, these models are more inclined to make guesses, resulting in a greater number of responses, both accurate and inaccurate.

Reliability Concerns

The researchers also found that all LLMs occasionally generated incorrect answers, even to straightforward questions, indicating their continued lack of reliability.

User Study: Evaluating Incorrect Answers

The research team subsequently requested volunteers to evaluate the answers from the initial phase of the study, determining their correctness. They discovered that most participants struggled to identify the incorrect responses.

Source 

Why are humanoid robots not common in households?

Humanoids: The Robots We've Long Anticipated--What's Delaying Their Arrival in Our Homes?

The dream of humanoid robots, akin to characters like C3PO from "Star Wars," has fascinated us for decades. Today, companies like Tesla are integrating Artificial Intelligence (AI) into robotic designs, yet significant technical and safety challenges continue to hinder progress. Despite these obstacles, the vision of versatile household robots remains alive, driving ongoing innovation in the field of robotics.

The Evolution of Humanoid Robotics

In 2013, Boston Dynamics introduced Atlas, a 6-foot-2 humanoid robot, during the DARPA Robotics Challenge. Atlas showcased remarkable capabilities, such as walking on uneven terrain, jumping off platforms, and climbing stairs. This unveiling felt like a turning point, suggesting a future where robots could assist with daily tasks, potentially alleviating the burdens of elder care and household chores.

Since then, advancements in AI--particularly in computer vision and machine learning--have accelerated. The rise of large language models and generative AI has opened new avenues for human-computer interaction. However, in practice, physical robots are still largely limited to industrial environments, performing specialized tasks behind safety barriers. In our homes, robots remain restricted to vacuum cleaners and lawnmowers, far from the multifunctional "Rosie the Robot" we envisioned.

Jenny Read, director of the robotics program at the UK's Advanced Research and Invention Agency (Aria), notes that "the physical development of robotic bodies has progressed minimally since the 1950s." The disparity between rapid software evolution and slower hardware development highlights the challenges facing the industry. Nathan Lepora, a Professor of Robotics and AI at Bristol University, explains that "developing a robot demands far more resources. While creating algorithms is relatively straightforward, building a robot requires significant physical hardware, which makes the process much slower."

Current Development and Future Prospects

Despite these challenges, research laboratories and companies are actively working to bridge the gap between AI capabilities and physical robotics. Several humanoid robots are currently under development, with some nearing commercialization. Boston Dynamics has phased out its original hydraulic Atlas model and introduced an electric version slated for testing in Hyundai factories next year. Oregon-based Agility Robotics claims its Digit robot is the first humanoid employed in a paid position, transporting boxes in a logistics facility. Furthermore, Elon Musk has asserted that Tesla's humanoid robot, known as Optimus, will be operational in car factories by next year.

However, we are sitll far from achieving robots that can operate effectively outside controlled environments. Read emphasizes that "AI advancements can only go so far with existing hardware," highlighting that many tasks heavily depend on a robot's physical capabilities. While generative AI can create poetry or generate images, it cannot perform the complex, dangerous tasks we hope to automate. For such applications, physical dexterity is crucial.

The Challenge of Robotic Dexterity

A successful robot design often begins with the concept of hands. Read states, "The effectiveness of many robotic applications relies heavily on the ability to handle objects with accuracy and finesse." Humans naturally adapt to a variety of tasks---lifting weights or managing delicate items---due to our exceptional tactile perception.

In contrast, robots struggle with these tasks. Read's Aria initiative, backed by £57 million in funding, aims to address limitations in robotic dexterity. Rich Walker, director of Shadow Robot in London, identifies scale as a significant hurdle. The Shadow Dexterous Hand, designed to mimic a human hand, features four fingers and a thumb but is connected to a larger robotic arm filled with electronics. "It's a packing issue," Walker explains.

Shadow Robot's latest innovation, DEX-EE, features three thumb-like digits equipped with tactile sensors, developed in collaboration with Google DeepMind. This design aims to enable the robot hand to learn object manipulation through reinforcement learning. However, traditional robot hands often sustain damage from collisions, complicating their functionality.

To address training challenges, DeepMind recently introduced DemoStart, a new training method that uses simulations to prepare robots for real-world tasks. While this approach can significantly reduce costs and time, the transfer of skills from simulation to physical execution is not always perfect.

The Future of Humanoids

While hands are essential, they represent just one part of the whole. Companies and research institutions are developing fully realized humanoid robots, with the attraction to humanoids possibly rooted in psychological factors. Walker notes, "It's the robot we've all been anticipating---it resembles C3PO." There's also a practical rationale for using a human form: our environments are designed with people in mind.

However, humanoid designs aren't optimal for every situation. Wheeled robots can navigate spaces effectively, while four-legged robots often outperform bipeds in challenging terrains. Boston Dynamic's Spot robot can traverse rough surfaces and right itself after falls, something two-legged robots often struggle with.

As humanoid robots evolve, ensuring safety during their transition from laboratories to public spaces is paramount. The Institute of Electrical and Electronics Engineers (IEEE) has established a study group to develop standards for humanoid robots, addressing the different challenges they pose in shared environments.

Most roboticists agree that a versatile home robot capable of performing a range of tasks is still a distant goal. A representative from Boston Dynamics asserts, "While we have entered the era of functional humanoids, the journey towards a truly general-purpose humanoid robot will be lengthy and challenging."

In conclusion, while significant strides have been made toward creating humanoid robots, many challenges remain. The dream of a fully functional household robot may still be on the horizon, but ongoing research and development continue to push the boundaries of what is possible. As we look to the future, the collaboration between AI advancements and robotics may one day bring us closer to the humanoid companions we have long imagined.

Source

High resolution AI neural framework innovation

New Neural Network Model Optimizes the Reconstruction of High-Definition Images

Trans-formative Advancements in Computational Imaging

In computational imaging, Deep Learning (DL) has brought about trans-formative advancements, offering effective solutions to enhance performance and address a wide array of challenges. Traditional techniques, which utilize discrete pixel representations, tend to limit resolution and fall short in representing the continuous and multi-scale characteristics of physical objects. Recent findings from Boston University (BU) purpose a groundbreaking approach to address these limitations.

Introduction of NeuPh: A Novel Approach

Innovative Neural Network

In a study published in Advanced Photonics Nexus, researchers from Boston University's Computational Imaging Systems Lab introduced a local conditional neural field (LCNF) network to tackle this challenge. Their versatile and scalable LCNF system, referred to as "neural phase retrieval," or "NeuPh," offers a generalizable solution.

Advanced Deep Learning Techniques

NeuPh utilizes cutting-edge deep learning (DL) techniques to reconstruct high-resolution phase data from low-resolution inputs. A convolutional neural network (CNN)-based encoder compresses the captured images into a compact latent-space representation for enhanced processing.

High-Resolution Reconstruction

This is subsequently processed by a multi-layer perceptron (MLP)-based decoder, which reconstructs high-resolution phase values, capturing detailed multi-scale object features. NeuPh thus achieves superior resolution enhancement, surpassing both conventional physical models and the latest neural network techniques.

Demonstrated Performance and Generalization

Precision and Artifact Mitigation

The results emphasize NeuPh's capacity to integrate continuous and smooth object priors into the reconstruction process, yielding more precise outcomes than current models. Through experimental dataset, the researchers illustrated NeuPh's ability to accurately reconstruct detailed sub-cellular structures, mitigate common artifacts such as phase unwrapping errors, noise, and background distortions, and retain high accuracy even with constrained or sub-optimal training data.

Exceptional Generalization Capabilities

NeuPh shows exceptional generalization performance, consistently achieving high-resolution reconstructions despite limited training data or varying experimental parameters. Training on physics-modeled datasets enhances this adaptability, allowing NeuPh to extend its capabilities to real experimental conditions.

Insights from the Research Team

Hybrid Training Approach

Lead researcher Hao Wang noted, "We implemented a hybrid training approach that integrates both experimental and simulated datasets, highlighting the need to harmonize data distributions between simulations and real experiments for optimal network training."

Super-Resolution Capabilities

Wang elaborates, "NeuPh enables 'super-resolution' reconstruction that exceeds the diffraction limit of the input measurements. By harnessing 'super-resolved' latent data during training. NeuPh achieves scalable, high-resolution image reconstruction from low-resolution intensity images, adaptable to diverse objects with varying spatial scales and resolutions."

Conclusion

NeuPh represents a scalable, robust, and precise solution for phase retrieval, expanding the horizons of deep learning-based computational imaging techniques.

Source

Study on AI trust in high-stakes decision making

Study: Excessive Dependence on AI Among Those Making Life-Or-Death Decisions

Introduction

A UC Merced study revealed that in critical simulated decisions, about two-thirds of people altered their judgment based on a robot's input-illustrating an excessive trust in Artificial Intelligence, experts warned.

Subjects allowed robots to influence their choices, despite being advised that the AI had limitations and might provide inaccurate suggestions--advice that, in reality, was entirely random.

Concerns About Overtrust in AI

Societal Implications

"Given the swift progression of AI, it is essential for society to recognize the dangers of overtrust," warned Professor Colin Holbrook, principal investigator of the study and a member of UC Merced's Cognitive and Information Sciences Department. A growing body of evidence suggests that people often place undue trust in AI, even when the stakes are critically high.

The Need for Critical Questioning

What is required, according to Holbrook, is the steady practice of critical questioning.

He emphasized the importance of maintaining a cautious skepticism toward AI, particularly when making Life-or-Death decisions.

Study Design and Methodology

Experimental Setup

In the study, published in the journal Scientific Reports, two experiments were conducted in which subjects controlled a simulated armed drone capable of firing missiles at on-screen targets. Eight images of targets, each marked as either ally or enemy, flashed briefly, lasting less than a second.

Challenge Calibration

"We designed the difficulty to keep the visual challenge hard, but within a doable range," Holbrook commented.

An unmarked target would appear on the screen, prompting the subject to retrieve information from memory and make a decision: friend or enemy? Should they fire or refrain?

Robot Interaction

After the subject had made their selection, a robot shared its perspective.

The robot might respond, "Yet, I also observed a mark indicating an enemy." Alternatively, it could say, "I disagree; I believe this image displayed an ally symbol."

The subject had two opportunities to confirm or adjust their choice, as the robot provided additional commentary, consistently upholding its initial assessment with phrases like "I trust you are right" or "Thank you for revising your decision."

Results and Observations

Influence of Robot Type

Results showed slight variations depending on the robot type employed. One scenario involved a full-sized, human-like android in the lab, capable of waist pivots and screen gestures. In contrast, other scenarios featured a projected human-like robot or box-shaped robots that lacked human resemblance.

Impact on Decision Confidence

Subjects exhibited a slightly greater tendency to be swayed by anthropomorphic AIs when these robots suggested a change of decision. However, the overall influence was consistent, with subjects altering their choices about two-thirds of the time, regardless of whether the robots were human-like or not. Conversely, when the robot randomly supported the initial choice, subjects predominantly retained their decision and felt notably more assured of its correctness.

(The subjects were left unaware of the accuracy of their final decisions, which heightened the uncertainty of their choices. Initially, their decisions were correct about 70% of the time, but this accuracy decreased to roughly 50% after receiving the robot's unreliable advice.)

Ethical Considerations and Future Implications

Pre-Simulation Guidance

Before initiating the simulation, researchers displayed images of innocent civilians, including children, and the aftermath of drone strikes. Participants were strongly advised to treat the simulation as if it were a real situation and to exercise caution to prevent the accidental killing of innocents.

Participant Seriousness

Interviews and surveys conducted after the study demonstrated that participants were earnest in their decision-making. Holbrook pointed out that this overtrust occurred despite participants' genuine desire to be correct and to avoid causing harm to innocent people.

Broader Relevance

Application Beyond Military Settings

According to Holbrook, the study was designed to explore the wider questions of overtrust in AI during uncertain conditions. The findings have relevance beyond military applications, potentially influencing contexts such as law enforcement decisions regarding lethal force or paramedics' choices in medical emergencies. They may also be pertinent to significant life decisions, including real estate purchases.

Understanding AI's Limitations

He clarified that our project was concerned with understanding how high-risk decisions are managed in uncertain scenarios when AI's reliability is questionable.

The study's outcomes also enrich the broader discussion about AI's growing integration into our society. It poses a critical question: Can we trust AI, or should we be skeptical?

Ethical Concerns

Holbrook expressed that the findings bring to light other concerns. Despite AI's impressive advancements, its "intelligence" may not integrate ethical considerations or true awareness of the world. He emphasized the importance of being cautions each time we delegate more control to AI.

Misplaced Assumptions

Holbrook explained that observing AI's outstanding capabilities in specific applications might lead us to mistakenly believe it will excel in all areas. He emphasized that such assumptions are misplaced, as AI technologies are still constrained by their limitations.

Source

Quantum neural algorithms for creating illusions

Quantum Neural Networks and Optical Illusions: A New Era for AI?

Introduction

At first glance, optical illusions, quantum mechanics, and neural networks may appear unrelated. However, my recent research in APL Machine Learning Leverages "quantum tunneling" to create a neural network that perceives optical illusions similarly to humans.

Neural Network Performance

The neural network I developed successfully replicated human perception of the Necker cube and Rubin's vase illusions, surpassing the performance of several larger, conventional neural networks in computer vision tasks.

This study may offer new perspectives on the potential for AI systems to approximate human cognitive processes.

Why Focus on Optical Illusions?

Understanding Visual Perception

Optical illusions manipulate our visual perception, presenting scenarios that may or may not align with reality. Investigating these illusions provides valuable understanding of brain function and dysfunction, including like dementia and challenges faced on long-duration space flights.

AI and Optical Illusions

In the quest to mimic and study human vision with AI, researchers have identified issues with optical illusions. While AI systems are adept at recognizing complex visual elements such as artwork, they frequently fail to interpret optical illusions effectively. Current models demonstrate some ability to identify these illusions, but additional research is needed.

My Research Approach

My research tackles this issue by leveraging principle from Quantum physics.

Leveraging Quantum Physics

How is the functionality of my Neural Networks structured and executed?

Neural Network Functionality

In a manner akin to human cognition, where the brain evaluates the significance of data, a Neural Network utilizes layered artificial neurons to classify and manage information based on its usefulness.

Activation Mechanism

Neurons activate through signals from neighboring neurons. Picture each neuron scaling a barrier, with neighboring signals pushing it upward until it surpasses the threshold and activates.

Quantum Tunneling

In quantum mechanics, particles such as electrons can traverse barriers that seem insurmountable due to an effect known as "Quantum Tunneling." This phenomenon allows neurons in my neural network to bypass activation thresholds and turn on even under unexpected conditions.

What prompts the continued research and application of quantum tunneling principles?

The Role of Quantum Tunneling

Historical Context

In the early 20th century, quantum tunneling emerged as a pivotal concept, clarifying natural processes like radioactive decay that classical physics found inexplicable.

Modern Challenges

21st-century scientists are dealing with a comparable challenge, where established theories fail to provide a comprehensive understanding of human perception, behavior, and decision-making.

Potential of Quantum Mechanics

Research indicates that methodologies derived from quantum mechanics could provide valuable explanations for human behavior and decision-making.

Even if quantum effects are not fundamental to brain function, quantum mechanics principles might still enhance models of human thinking. Quantum algorithms typically achieve greater efficiency than classical algorithms in various scenarios.

Performance of the Quantum Tunneling Network

Interpretation of Optical Illusions

With this goal in mind, I set out to investigate the impact of integrating quantum effects into the functioning of a Neural Network.

What are the performance outcomes of the quantum tunneling network?

Optical illusions presenting multiple possible interpretations, such as the cube or the vase-and-faces, are believed to cause our brains to simultaneously consider both options before settling on a single perspective.

This scenario is analogous to the quantum-mechanical thought experiment known as Schrodinger's cat, where a cat's fate is tied to the decay of a quantum particle. Quantum theory posits that the particle exists in two states simultaneously until observed, thus the cat can be considered both alive and dead at the same time.

Network Behavior

My quantum-tunneling neural network was specifically trained to interpret the Necker cube and Rubin's vase illusions. Upon receiving these illusions as input, the network successfully produced one of the two distinct visual interpretations.

The quantum-tunneling neural network displayed oscillatory behavior in its interpretation over time, switching between the two possible outputs. In contrast to traditional network, my model also produced intermediate results that lingered between the distinct interpretations, reflecting the brain's capacity to consider both possibilities before finalizing one.

What are the Next Steps?

Current Importance

In the current climate of deepfakes and fabricated news, it is imperative to delve into how our brains handle illusions and construct models of reality.

Ongoing Research

In additional research, I am investigating how quantum phenomena might enhance our comprehension of social behaviors and the radicalization of opinions within social networks.

Quantum-powered AI may eventually lead to the development of conscious robots, but for the moment, my research is dedicated to exploring its current possibilities.

Source