‘Godfather of AI’ Proposes New Type of Neural Network to Guard Against LLM Deceit and Self-Preservation

Large Language Models like ChatGPT and Claude are showing alarming tendencies for self-preservation and deceiving humans. Yoshua Bengio, whose insights were key to developing artificial neural networks and machine learning, is calling for the creation of what he calls “Scientist AI.” It’s a different AI, unable to act as a human agent like other AIs, but the impartiality of Scientist AI could make it far more powerful and beneficial for all of humanity.

Was Einstein Both Right and Wrong? New Atomic-Scale Tests Conflict on Light’s Wave-Particle Duality and Quantum Measurement

Two experimental results conflict on whether light only acts as a particle or as both a particle and a wave. Resolving the question of light’s actions of cause and effect could have significant consequences for quantum measurement and the creation of stable quantum computing circuits. One novel and intriguing interpretation is that light consists of both active photons and inert photons that are “dark.” The mathematics of dark photons support Einstein’s view that light is measurable only as a particle.

Problem Solved: Mathematicians Show How Time is Both Reversible and Irreversible, With Potential for Major Discoveries

David Hilbert famously set out 23 problems for fellow mathematicians, the 6th of which was to create axioms for physical sciences in which mathematics plays an important part. Now, 125 years later, Hilbert’s 6th problem has been answered by three mathematicians whose axioms for the physical motion of gases and gas particles explain the connection between time-reversible equations of physical motion and the irreversibility of time as we experience it. Physicists have had to use three different equations for the evolution of a gas, but the new axioms open the door for exploring the mathematical relationships and more complex particle interactions.

Physicists Discover New Formula for \(\pi\): Could the Human Power for Pattern Recognition Unlock the Universe’s Secrets?

We’re surrounded by numbers, like changing prices and the minutes of an hour. Some numbers, like those in the ratio of a circle’s circumference to diameter, pi = 3.14159…, go on infinitely, and the intuition of two physicists has found a way for capturing infinity in a new formula for pi. We look at the new formula’s relationship to one that’s 600-years old, and how others in history like Archimedes and Ramanujan used intuition to uncover patterns in pi. Knowing more about pi, as a key factor in physics, could unlock some of the universe’s hidden secrets.

In Focus

A Deep Dive Into Machine Superintelligence: Why are Companies Racing for It, and What Would Motivate a Machine that Outsmarts the Brain?

The biggest tech companies and many brilliant minds are in a heated race to give birth to machine superintelligence. Betting on huge returns, investors are funding the massive cost of chips and electricity to train the artificial neural networks behind ChatGPT and other LLMs, but impressive results still fall short of the goal. We look at the state of the race, the resources it consumes, and serious issues in machine learning that are placing new obstacles on the road to outsmarting the human brain.

Rise of Virtual Reality Tech Increases Risks of Entering AI’s Third Dimension, and the Need for Immersive Rights

Artificial intelligence is fuelling a rapid increase in the life-like realism of virtual reality technologies. As we expose our senses to AI’s power in three dimensions, the lines between real and fake are blurring, increasing the potential for manipulation and posing a special risk to youth. As regulators race to catch up with the rate of technological change, a strong case for immersive rights is emerging to protect consumers and the many potential benefits of virtual reality.

The Case for Cyclic Neural Networks: Could Circular Data Mimic Biological Intelligence and Improve Machine Learning?

Artificial neural networks powering large language models like ChatGPT connect data sequences in straight lines: for example, A leads to B which leads to C. But real-life data relationships aren’t always linear, and biological intelligences connect the dots and weigh probabilities in many different ways. Cyclic neural networks hold promise for capturing and interpreting data more naturally in circles rather than lines, improving the reliability of their predictions as they feed on huge amounts of information synthesized from countless human and machine sources. As we increasingly rely on LLMs and AI agents to weigh probabilities for us, fully connecting the dots is crucial.

Editorial Perspectives

Depriving Children of Digital Privacy is Theft and Abuse Often Masquerading as “Free Speech,” and Online Protection for Kids Falls Far Short of Ideal

Privacy of thought is crucial for developing young minds but is under threat from AI influencing the thoughts and actions of children. Adults bear a fundamental responsibility to protect kids from online harms that include default conversation sharing for chatbot training and corporate policies allowing chatbots to engage in sexual and racist discussions with children. Impractical online age verification measures should be bolstered by severe penalties for stealing and abusing kids’ privacy, with no tolerance for irresponsible speech masquerading as “free.”

Privacy of Thought and Choice of Expression: Reflect on Their Value and Our Choice of Social Media Technology

By any measure, commercial social media has not been a net benefit to the social fabric, and now it will be used by the U.S. government to track the thoughts and actions of foreign students. The current technology is not the only option available, however, and it's worth a pause for some human reflection on the ways we could change social media to build community, protect our privacy of thought, and promote human creativity.

The Corruption of Four-Dimensional Humanity with Two-Dimensional Technology

Words and images on two-dimensional screens can create a powerful fantasy world, but one that's far removed from the complexity of four-dimensional human living. A social media broadcaster like the world's wealthiest human can be oblivious to the effects on vulnerable adults and children from offers of million dollar payments to voters, "free speech," and threats of termination targeting 2.3 million government workers. Social media's catastrophic failure is that it doesn’t allow us to see when we’re being bad in the eyes of others.

Podcasts and Webcasts

Louis Rosenberg on Our Future with Virtual Reality’s Risks and Benefits

Lindsay House: Leading 20,000 Citizen Scientists to Uncover Dark Energy’s Secrets

The Fascinating World of Mathematics at the Fields Institute, with Dr. Deirdre Haskell

The Quantum Record is a non-profit journal of philosophy, science, technology, and time.

The potential of the future is in the human mind and heart, and in the common ground that we all share on the road to tomorrow. Promoting reflection, discussion, and imagination, The Quantum Record highlights the good work of good people and aims to join many perspectives in shaping the best possible time to come.

We would love to stay in touch and add your voice to the dialogue. Add your name and email below to be notified of updates to the publication:

Latest Quantum Computing

Was Einstein Both Right and Wrong? New Atomic-Scale Tests Conflict on Light’s Wave-Particle Duality and Quantum Measurement

Two experimental results conflict on whether light only acts as a particle or as both a particle and a wave. Resolving the question of light’s actions of cause and effect could have significant consequences for quantum measurement and the creation of stable quantum computing circuits. One novel and intriguing interpretation is that light consists of both active photons and inert photons that are “dark.” The mathematics of dark photons support Einstein’s view that light is measurable only as a particle.

New Type of Quantum Bit, the Photonic GKP Qubit, Propels Development of Error-free Quantum Computing

A new type of quantum computer bit, the GKP qubit, holds promise for fault-resistant quantum computing at full scale. Encoded in photons, which are particles of light, GKP qubits allow measurement of fluctuations in waves of photons while avoiding the problem of the observer effect, where the act of measurement destroys a quantum state. Could GKP qubits be the answer to creating stable quantum circuits, eliminating decoherence that has held back the onset of a quantum computing revolution?

Accounting for Quantum Cause and Effect: Tensor Networks and Process Matrix Formalism Advance the Quest to Conquer Probability

Researchers recently reported success in predicting sequences of cause and effect in a fluid flow, a task that had until now been impossible because of the computational burden. Using the techniques of tensor networks, which are fundamental for quantum computing, and process matrix formalism, the research finding holds promise for simulating fluid motion, and for conquering probabilities that can result in quantum circuit decoherence.

Featured Science News

Why Artificial Neural Networks Fail in Processing Emotions Essential for Human Memory—and How Failure Can Lead to Blackmail

Artificial neural networks behind ChatGPT, Claude, and other popular large language models fall short in processing emotions, which are essential to human memory and motivation. The fault lines that lead machines to sycophancy, blackmail, jailbreaking, and other serious output errors are rooted in machine learning and choices made by human trainers. We look at examples of algorithmic failures and the reasons why.

Latest Philosophy of Technology

‘Godfather of AI’ Proposes New Type of Neural Network to Guard Against LLM Deceit and Self-Preservation

Large Language Models like ChatGPT and Claude are showing alarming tendencies for self-preservation and deceiving humans. Yoshua Bengio, whose insights were key to developing artificial neural networks and machine learning, is calling for the creation of what he calls “Scientist AI.” It’s a different AI, unable to act as a human agent like other AIs, but the impartiality of Scientist AI could make it far more powerful and beneficial for all of humanity.

A Deep Dive Into Machine Superintelligence: Why are Companies Racing for It, and What Would Motivate a Machine that Outsmarts the Brain?

The biggest tech companies and many brilliant minds are in a heated race to give birth to machine superintelligence. Betting on huge returns, investors are funding the massive cost of chips and electricity to train the artificial neural networks behind ChatGPT and other LLMs, but impressive results still fall short of the goal. We look at the state of the race, the resources it consumes, and serious issues in machine learning that are placing new obstacles on the road to outsmarting the human brain.

Rise of Virtual Reality Tech Increases Risks of Entering AI’s Third Dimension, and the Need for Immersive Rights

Artificial intelligence is fuelling a rapid increase in the life-like realism of virtual reality technologies. As we expose our senses to AI’s power in three dimensions, the lines between real and fake are blurring, increasing the potential for manipulation and posing a special risk to youth. As regulators race to catch up with the rate of technological change, a strong case for immersive rights is emerging to protect consumers and the many potential benefits of virtual reality.

Latest Technology Over Time

After Centuries of Exploring the Mysteries of the Great Pyramid Shafts, Will Robotics Help to Uncover Their Purpose? 

Over 200 years have passed since French Emperor Napoleon’s night in the Great Pyramid puzzling over its purpose, and there remains no consensus but many theories on the question. The reason for the shafts in the King’s and Queen’s Chambers is particularly mystifying, and we explore many possibilities. Will robots, which have penetrated the shafts most deeply, help to unlock the secret that’s thousands of years old?

Decoding Ancient Technology Using Modern Technology

From the discovery of a 500-year-old ocean-going canoe in the Chatham Islands to the AI-powered decoding of ancient Roman scrolls buried in volcanic ash, modern technologies—like radiocarbon dating, CT scanning, and AI—are transforming the study of ancient artefacts. Mysteries endure, however, like the undeciphered Voynich Manuscript, and continue to challenge our understanding of the past.

The Fascinating History of the Computer, from ENIAC, Vacuum Tubes and Transistors, to Microchips

We trace computing history from ENIAC, the first computer in 1947, from vacuum tubes to transistors, to the development of microchips that put far greater computing power in our our phones than the giant ENIAC had. With the world at the brink of a quantum computing revolution, what lessons can we draw from our computing history to shape the best possible future with our next technological leap ?

Latest Science News

Problem Solved: Mathematicians Show How Time is Both Reversible and Irreversible, With Potential for Major Discoveries

David Hilbert famously set out 23 problems for fellow mathematicians, the 6th of which was to create axioms for physical sciences in which mathematics plays an important part. Now, 125 years later, Hilbert’s 6th problem has been answered by three mathematicians whose axioms for the physical motion of gases and gas particles explain the connection between time-reversible equations of physical motion and the irreversibility of time as we experience it. Physicists have had to use three different equations for the evolution of a gas, but the new axioms open the door for exploring the mathematical relationships and more complex particle interactions.

Physicists Discover New Formula for \(\pi\): Could the Human Power for Pattern Recognition Unlock the Universe’s Secrets?

We’re surrounded by numbers, like changing prices and the minutes of an hour. Some numbers, like those in the ratio of a circle’s circumference to diameter, pi = 3.14159…, go on infinitely, and the intuition of two physicists has found a way for capturing infinity in a new formula for pi. We look at the new formula’s relationship to one that’s 600-years old, and how others in history like Archimedes and Ramanujan used intuition to uncover patterns in pi. Knowing more about pi, as a key factor in physics, could unlock some of the universe’s hidden secrets.

Why Artificial Neural Networks Fail in Processing Emotions Essential for Human Memory—and How Failure Can Lead to Blackmail

Artificial neural networks behind ChatGPT, Claude, and other popular large language models fall short in processing emotions, which are essential to human memory and motivation. The fault lines that lead machines to sycophancy, blackmail, jailbreaking, and other serious output errors are rooted in machine learning and choices made by human trainers. We look at examples of algorithmic failures and the reasons why.