astro + quantum

Artificial intelligence and quantum computing (AI & QC) are simultaneously rapidly developing fields with immense potential directed toward revolutionizing science and technology. Astronomy, an ancient science, has seen tremendous expansion in the century, and these two technologies of the last decade have significant implications for reconstructing our understanding of the universe. Within this vast field of research there are a variety of forms in which the industry progresses, including construction of aerospace activity, telescopic tools, etc. The intersection of machine learning and astronomical research is the fuel for the fire of the next line of disruptive technologies in the industry; it offers incredibly creative possibilities for problem solving and redesigning preexisting methodologies. Writing this piece has allowed me to delve deeper into the modern application of the study of space and how my chosen area of interest, computer science, extends and has begun to dip into this vast field. The scope of machine learning conducted research has only recently been observed at a surface level, and I hope to be able to contribute to its incorporation in the space industry’s next-generation projects.

To bring this idea to life, we can observe NASA’s recently launched Webb telescope. Natasha Batalha, an astrophysicist and computer scientist working at the Ames Research Center, spoke in an interview about how the computer self-classifies and observes data received from the telescope – the beginning of high-level artificial intelligent integration into such projects. Essentially, the computer determines the atmospheric properties and detects spectra in order to classify and interpret it through its own software. As computers are built with increasingly smarter algorithms, they have become exponentially more involved by consuming and interpreting incredible volumes of data with no explicit programming done prior – all of this is accredited to the systematic models that utilize algorithms to automatically parse data and produce highly accurate and specified predictions, calculations, or outcomes. Listening to her work had acted as an inspirational catalyst for me and further motivated me to develop my base of research and introductory work. In a project that I did during the summer program last year, we explored this by coding an exoplanet detection tool, where telescopic data was used as a set to derive cycles of recurring light flux’s from in order to determine a possible revolving pattern of a planetary body - I extensively practiced the usage of data collected from NASA’s Kepler telescope to train AI model to detect exoplanets. In the following months, I completed an apprenticeship with the AEOP where I worked individually with a professor to conduct research using computer science to design high entropy alloys which could be used in the materials design of aerospace machinery. These compositions were later used to train machine learning models on the simulated data for melting points. I tested multi-component atomic systems by code through simulation software and developed an abstract. This allowed me to explore two potential applications of AI in the space industry through an immersive experience, and I am yet to interact with the possibilities of quantum computing.

As a basepoint, I’ll begin by defining the terms quantum computing & artificial intelligence in a broader context to clearly apply them effectively within astronomy. AI is the development of computational software systems designed to perform tasks that are typically performed solely by the requirement of human intelligence. This would include highly cognitive skills including visual perception, speech recognition, critical thinking, and similar concepts. QC, however, stems from a quantum-mechanical phenomena that performs operations on data far faster than the classical computing systems, which enables extensive data processing in an exceptionally efficient time frame. Quantum computing may define the bounds and efficiency of the program, and machine learning is the core system of the program itself which determines the functionality. When compounding the two concepts together, quantum computing can feed artificial intelligence algorithms with the power it requires to run complex programs, proving it the essential in machine learning experiments. Binary code is the basis of all computer science, including AI, as it is performed on computers which encode information as independently either a 0 or 1 bit. Quantum computers, however, are of another nature themselves. Information can be encoded with an additional feature where 0s and 1s can be stored simultaneously in the form of qubits. Similar to how a quantum of light energy or a photon can exist as a particle or wave based on behavior, qubits exceed beyond the necessity to encode in either a 0 or 1 independently. Thinking of quantum behavior in physics, this mimics the capacity to superimpose the binary states and allows the machine to run exponentially faster. Companies such as Sandbox AQ, Quantinuum, Xanadu, Zapata, and others have already built quantum tools purposed for artificial intelligence application for a variety of fields such as cryptography, drug discovery, neurochips, cybersecurity, material sciences, and natural sciences – including application in space research. As I discuss the possibilities for AI in cosmological exploration, it will supplement the information as to why we would require the power of quantum computing systems. Astronomy is the study of celestial objects and the universe in its entirety; as this field remains incredibly relevant in the usage of new technologies and has historically incorporated them to advance our understanding, therefore the integration of AI & QC will inevitably be expected to have a significant impact on the field.

As previously established, there is an immense depth to the capability of AI integration in the space industry. In the form of a summative list, I’ll discuss some potential projects after classifying the applicable branches of AI. Artificial intelligence can be broken into two subcategories, one of which is narrow AI. This AI is typically deemed weaker and is designed to perform specified tasks but is incapable of initiating independent learning beyond its programming. An example could include the AI “Eye” an oceanographer at NASA developed to analyze image streams for data patterns and discover relationships between variables. This technology currently scans and interprets images from Earth’s complex aquatic and coastal regions in search for significant features through its optical sensor in infrared and other frequencies. It cannot, however, make critical decisions or initiate further learning. As we move up in capability or power on the spectrum of AI, we approach technologies capable of mimicking human behavior effectively (notably not yet to the extent of emotional capacity or sentience). This stage of artificial intelligence is theoretical at its current point in certain aspects of development but has the capacity of being developed with the purpose of performing accurate intellectual tasks. Researchers are largely working on developing algorithms required for such a system. This category of AI falls under the subfield of machine learning (ML) which involves the development of computers with the ability to make inferences and learn based on provided data. By training a system on various sets of strong data, the system will eventually adopt pattern recognition itself in order to assert precise predictions stemming from the information it was fed. The key advantage of AI is the capacity of analyzing and process heavy datasets with relative efficiency, making it furthermore important to emphasize the potential role of quantum computer systems within these functions. In application to astronomy, this ability can be utilized to analyze vast numbers of data collected from telescopic instruments and can be taught to identify patterns in order to reach accurate conclusions on the behavior of celestial objects in significantly shortened time periods. Interestingly, in the initial calibration and testing of such programming, it detected two exoplanets via sifting through the Kepler mission data which human professionals had missed. AI can be used to craft astronomic maps by processing the influx of data from telescopes and replace the human effort that typically goes into this process: regional images taken by the telescopes are accurately arranged and digested by the algorithms. It can also locate and detect certain formations such as stellar clusters. Shifting over to primarily quantum computing systems (rather than the intersection with machine learning) which uses phenomena such as superposition and entanglement to perform operations on data, it serves as a probable replacement for high-level artificial intelligence usages. The ability for the quantum computer’s unclassical qubits to exist simultaneously in states of 0 and 1 provide the ability to perform a multitude of operations simultaneously. An issue within the concept development, however, is quantum decoherence. Essentially, the term is defined by the required monitoring and isolation of qubits due to their nature – in the occurrence of qubits interacting with their environment, they will lose their quantum state and revert to the behavior of a classical qubit. Due to this, quantum computers must be painstakingly operated and sophisticated in design in order to minimize the effects of decoherence. Regardless of such challenges, the progress of quantum computing has been outstanding as a research field, and a demonstration from 2018 displayed that the milestone of “quantum supremacy” has been achieved: a quantum computer could solve a provided problem in 200 seconds that would take a powerful supercomputer 10,000 years.  

Finally, tying the two concepts back to astronomy, I’ll overview some essential aspects of the study of our universe where the two factors of quantum computing and artificial intelligence may play in. A key tool in all subsections of space sciences is the telescope, which collects invaluable data on light and other forms of radiation emitted from certain locations or objects in the broader universe. This information can typically be utilized to provide information on properties and behavior post analysis. Other instruments could include spectrographs and advanced cameras designed to additionally provide information on elements such as composition, temperature, and velocity of the observed objects. As these instruments advance, so do does the development of data and ability to comprehend it. Significant credit is due to the usage of data storage in computers and increasingly automated systems – analysis of these tremendous datasets is being performed by increasing integrated computational systems at a majority of labs. As a science, astronomy will typically generate vast amounts of data that is typically far too complex and specified for manual analysis by humans, and an especially lengthy process susceptible to error. Algorithmic parsing and identification of these patterns significantly increases accuracy, and as mentioned above, can search for bodies such as exoplanets (planets orbiting stars outside our solar system) which is a naturally time-consuming and complex process sped up by the automatic processing of computers which will scan potential exoplanets and perform secondary screening to confirm via property analyzation. Extending this process, the AI can also be used to help classify objects based on the discovered properties and therefore classify galaxies, for example, based on their shape or classify stars based on their spectral signature, which reveals information such as composition, luminosity, radius, etc. AI can also be used in the application of developing predictive models for occasions such as deciphering the likelihood and predicted the time of a star undergoing a supernova explosion. As this technology advances, it is projected to be tasked with far more complex prediction and simulation of events. This is where quantum computing is most effectively introduced: simulation. Simulating the incredibly complex behavior of certain celestial objects is a difficult task requiring significant computational resources, which is a process quantum computing can assist in and further allowing researchers to examine behaviors in far more detail than possible before. A strong example could be the simulation of the activities of black holes, considered one of the most obscure and complicated objects existing in the universe – in order to properly simulate these, complex equations must be solved which are difficult for classical computers to approach and therefore disenable researchers from thoroughly understanding the subject. Quantum computing may also be applied in the optimization of telescopic arrays, which are collections of telescopes that work in unison to provide a more detailed view of the observable universe. In order to optimize the placement and configuration of these tools, significant computer resources are required once again. Overall, the combination of AI and quantum computing as research methods can prove to be a remarkable reinvention of the work of the space industry. AI algorithms could be used to analyze data collected by telescopes and other instruments, while quantum computers could be used to simulate the behavior of celestial objects and optimize telescope arrays.

Their convergence is already being integrated into the space industry, as observed in both NASA’s Quantum Artificial Intelligence Laboratory (QuAIL) and Frontier Development Lab Programs. To develop cutting edge tools, a Frontier Development Lab has been created for an 8-week collaboration between technology and space innovators each summer, where they brainstorm and develop project code. The program has been running since around 2016 and is partnered with the SETI Institute and Ames Research Center both based in Silicon Valley and aiming to pair young science and computer engineering doctoral students with experts from the space agency, academia, and renowned technology corporations. Their contributions arise in various combinations of hardware, algorithms, super-compute resources, funding, and subject-matter experts and have already been assisting in identify asteroids, analyze exoplanet atmospheres (for promising chemistry associated with habitability to narrow exploration costs), and predict solar radiative events according to NASA’s official site. An example that especially intrigued me was the Bayesian neural network that had been tested prior to 2020. They deployed the neural network technique against the machine learning technique called “random forest” in order analyze the atmosphere of the exoplanet WASP-12b and discovered it was far more accurate at identifying an abundance of various molecules and was also capable of providing a measure of certainty on its on predictions. By 2017, students from France, South Africa, and the United States in the lab had also presented a technique adopted in the real world of a machine learning program that created increasingly efficient 3D models of asteroids with accurate estimation of their shapes, sizes, and spin rates which are critical to efforts for detecting and deflecting potentially threatening asteroids for our planet. This rendering could be completed within four days rather than over two months with classic systems and is being applied by astronomers at the Arecibo Observatory in Puerto Rico; the lab still continues continue to work on applications of sophisticated algorithms in similar volumes of data as the agency gathers around 2 gigabytes of data every 12 seconds from its fleet of spacecraft and the areas of potential increase. As a second example, NASA’s QuAIL is the space agency’s hub for developing quantum computing techniques and algorithms that may address difficult optimization and machine learning problems arising in aeronautics, space science, and exploration missions with expertise ranging in physics, computer science, mathematics, chemistry, and engineering. Their work centers on the utilization of quantum hardware – they research and evaluate different prototypes for experimentation and sampling. Currently, a key component to their work includes a D Wave machine (one of the initial quantum computational devices) at NASA Ames. An essential aspect of their work is devising methods to utilize emerging quantum hardware most effectively such as systematic methods. For example, using statistical sampling of the ground-state of electrons to avoid potentially significant bias due to limited flexibility of classical computation, this implementation can perform unbiased simulation on chemical systems with up to 120 orbitals. Essentially, with the ability to combine constrained calculations and higher accuracy for the machine learning process with QC, a result of more accurate ground-state mechanical wavefunction can be derived. This specific application is called the unbiased fermionic Monte Carlo method and can use experimental devices such as the noisy intermediate-scale quantum computers or NISQ. As NASA missions require increasingly complex solutions to a multitude of challenging computational problems, the ambitiousness of each project’s limits is defined by the ability to further such technologies in applications such as space vehicle design, anomaly detection, data analysis and fusion, and advanced logistics.

Yet, while the integration of AI and quantum computing with astronomy has significant potential, there are limited challenges that may need to be addressed in the near future. One key challenge may appear to be the development of algorithms and techniques that are specifically tailored to the unique properties of quantum computing. This will require a significant amount of research and development, as well as collaboration between researchers in different fields. Expanding on that, there is an increasingly greater need for further collaboration between researchers in these specific different fields: astronomy, AI, and quantum computing. This will require the development of new interdisciplinary research programs and the sharing of knowledge and resources between different fields. Regardless of the significant challenges that need to be addressed, including the development of algorithms and more powerful quantum computing systems, the integration of these technologies with astronomy could lead to new discoveries and insights into the properties and behavior of the universe. This could have significant implications for our understanding of the universe and the development of new technologies and applications.