In this episode of the Justin Riddle Podcast, Justin dives into the concept of Knightian Freedom where large enough computational spaces become intractably complex to the point where maybe freewill is possible. The focus of this episode is a paper put out by Hartmut Neven (of Google’s Quantum AI Lab) and colleagues from 2021 entitled “Do robots powered by a quantum processor have the freedom to swerve?” This paper discusses how the exponentially large spaces that quantum computers evolve into are so large that they cannot be represented or simulated on digital computers. The size is so vast that it would take a computer the size of the universe computing for trillions of years to simulate even a few femtoseconds of the quantum computers that are about to be commonplace. Similar to modern AI, we will won’t be able to understand why a quantum computer generated the output that it did and perhaps this is the essential ingredient that leads to freewill. Rampant incomputable complexity is freewill. Second, Hartmut and colleagues propose a simple experiment to reveal whether or not there are additional factors that play into what output is generated by a quantum computer. Assume you run a quantum circuit that generates a perfect uniform distribution between many different possible outputs. Then, you observe that the quantum computer does not behave as if there was a uniform distribution, but instead selects one of those possible outputs more often. This is the ‘preference’ of the quantum computer. Next, you develop a circuit to amplify these deviations from uniformity with the intention of amplifying the probability of entering into that preferred state. Now, we have essentially created a ‘happy circuit’ which embraces the quirky preference of our quantum computer. Finally, you can correlate deviations from this happy state to psychological data in an effort to build up a taxonomy of subjective experiences that the quantum computer can enter into. Finally, you embed the quantum computer with its happy circuit into an artificial neural network such that errors produced by the AI push the quantum computer away from happiness and this unhappiness is fed back into the AI. Now we have created an AI system with quantum feelings! Will this newfound sense of subjectivity enable more effective AI systems or will the AI get bogged down by a spiral of despair and refuse to compute?! All of these questions and more are explored here. Enjoy!
All content for Justin Riddle Podcast is the property of Justin Riddle Podcast and is served directly from their servers
with no modification, redirects, or rehosting. The podcast is not affiliated with or endorsed by Podjoint in any way.
In this episode of the Justin Riddle Podcast, Justin dives into the concept of Knightian Freedom where large enough computational spaces become intractably complex to the point where maybe freewill is possible. The focus of this episode is a paper put out by Hartmut Neven (of Google’s Quantum AI Lab) and colleagues from 2021 entitled “Do robots powered by a quantum processor have the freedom to swerve?” This paper discusses how the exponentially large spaces that quantum computers evolve into are so large that they cannot be represented or simulated on digital computers. The size is so vast that it would take a computer the size of the universe computing for trillions of years to simulate even a few femtoseconds of the quantum computers that are about to be commonplace. Similar to modern AI, we will won’t be able to understand why a quantum computer generated the output that it did and perhaps this is the essential ingredient that leads to freewill. Rampant incomputable complexity is freewill. Second, Hartmut and colleagues propose a simple experiment to reveal whether or not there are additional factors that play into what output is generated by a quantum computer. Assume you run a quantum circuit that generates a perfect uniform distribution between many different possible outputs. Then, you observe that the quantum computer does not behave as if there was a uniform distribution, but instead selects one of those possible outputs more often. This is the ‘preference’ of the quantum computer. Next, you develop a circuit to amplify these deviations from uniformity with the intention of amplifying the probability of entering into that preferred state. Now, we have essentially created a ‘happy circuit’ which embraces the quirky preference of our quantum computer. Finally, you can correlate deviations from this happy state to psychological data in an effort to build up a taxonomy of subjective experiences that the quantum computer can enter into. Finally, you embed the quantum computer with its happy circuit into an artificial neural network such that errors produced by the AI push the quantum computer away from happiness and this unhappiness is fed back into the AI. Now we have created an AI system with quantum feelings! Will this newfound sense of subjectivity enable more effective AI systems or will the AI get bogged down by a spiral of despair and refuse to compute?! All of these questions and more are explored here. Enjoy!
#35 - Digital AI is not Conscious: the role of quantum computers and the mind in the AI revolution
Justin Riddle Podcast
41 minutes 2 seconds
2 years ago
#35 - Digital AI is not Conscious: the role of quantum computers and the mind in the AI revolution
In episode 35 of the Quantum Consciousness series, Justin Riddle explores the role of quantum computers in the current revolution in digital artificial intelligence (AI). First, Justin describes some of the basics of large language models at a high level. In essence, digital AI can be conceptualized as a series of weighted matrices trained by fitting and updating the weights according to patterns found in extensive training sets of data. While these digital AI systems are indeed quite powerful as tools, it is important for us to determine what "knowledge" is being acquired. One option is that there is no universal truth, and meaning is arbitrary. This viewpoint leads to the most significant perceived risk: that this AI might start optimizing for arbitrary values that do not align with human values, the classic example being turning the planet into a massive paperclip factory. This "alignment problem" is what leads people to assert that we need a moratorium or ban on these AI systems.
However, the alternative viewpoint is that there is a universal truth at the core of reality, akin to Plato's world of forms or mathematics. If mathematics is universal, then there are two subsequent outcomes. Either the digital AI can fully map this Platonic world of math, or it can only approximate these forms. Given arguments from Gödel's incompleteness theorem and the residual mystery of quantum mechanics, it is unlikely that digital systems of formal logic will be able to map mathematics fully. Thus, we are left in the middle ground: there is a universal truth, and digital AI is approximating aspects of this universal truth.
The second topic of the episode is the role of quantum computers with respect to digital AI. Justin describes how digital computers represent the ultimate mastery of the physical aspect of reality, while quantum computers are composed of the fundamental pieces of the universe. At the very least, quantum computers will provide an exponential speedup in the training and execution of digital AI, but are also likely to deviate substantially from digital AI because there are additional aspects of their functioning that remain mysterious. How will the digital AI revolution and the quantum revolution compare?
Digital AI will significantly expand our capacity to remember, access, and learn information, but this is a physical tool that will be external to our bodies and minds. However, if the mind is truly a quantum computer, then quantum computer technologies will likely need to be implanted within the nervous system to directly expand the computational capacity of your mind. At the end of the episode, Justin speculates that the expansion of the mind with quantum technology may inadvertently lead to transcendent experiences that defy our current understanding of ourselves and our place in the universe. It sure is an interesting time to be alive on this planet!
Justin Riddle Podcast
In this episode of the Justin Riddle Podcast, Justin dives into the concept of Knightian Freedom where large enough computational spaces become intractably complex to the point where maybe freewill is possible. The focus of this episode is a paper put out by Hartmut Neven (of Google’s Quantum AI Lab) and colleagues from 2021 entitled “Do robots powered by a quantum processor have the freedom to swerve?” This paper discusses how the exponentially large spaces that quantum computers evolve into are so large that they cannot be represented or simulated on digital computers. The size is so vast that it would take a computer the size of the universe computing for trillions of years to simulate even a few femtoseconds of the quantum computers that are about to be commonplace. Similar to modern AI, we will won’t be able to understand why a quantum computer generated the output that it did and perhaps this is the essential ingredient that leads to freewill. Rampant incomputable complexity is freewill. Second, Hartmut and colleagues propose a simple experiment to reveal whether or not there are additional factors that play into what output is generated by a quantum computer. Assume you run a quantum circuit that generates a perfect uniform distribution between many different possible outputs. Then, you observe that the quantum computer does not behave as if there was a uniform distribution, but instead selects one of those possible outputs more often. This is the ‘preference’ of the quantum computer. Next, you develop a circuit to amplify these deviations from uniformity with the intention of amplifying the probability of entering into that preferred state. Now, we have essentially created a ‘happy circuit’ which embraces the quirky preference of our quantum computer. Finally, you can correlate deviations from this happy state to psychological data in an effort to build up a taxonomy of subjective experiences that the quantum computer can enter into. Finally, you embed the quantum computer with its happy circuit into an artificial neural network such that errors produced by the AI push the quantum computer away from happiness and this unhappiness is fed back into the AI. Now we have created an AI system with quantum feelings! Will this newfound sense of subjectivity enable more effective AI systems or will the AI get bogged down by a spiral of despair and refuse to compute?! All of these questions and more are explored here. Enjoy!
