The Race for Quantum Computing — The End of Humanity?

Edward Teller, known as the father of the Hydrogen bomb, was concerned during its creation that the explosion could create a chain reaction in which the initial explosion would lead to hydrogen atoms in the water and air fusing together into helium. Ultimately, he feared this could lead to a runaway reaction which would “engulf the globe, extinguishing all life”. [1]

Today, humanity is faced with a similar but unique dilemma, Quantum Computing (“QC”). The implications of QC, given how integrated technology and cyber security have become in our day-to-day life, are far reaching and the technology has the potential to completely undo all current security measures in place. Recently, NASA hit the emergency brakes on their QC project “sending shockwaves through the scientific community” as the implication of them doing so likely means that there are concerns about the merging of AI and quantum computers.[2] Specifically, concerns that doing so would create technology that dominants human capabilities and pose a threat to cybersecurity. So, what is QC?

QC at its core deals with how very tiny particles such as atoms interact. Unlike traditional computing, QC can handle a huge number of possibilities simultaneously. QC can do this as, unlike traditional computing which operates off of bits (0 or 1, based on voltage or charge), QC operates off of quantum bits (exist in a superposition of 0 and 1 simultaneously that enables parallel computation). Particularly, this would allow computers to solve matrixes quicker and allow for a greater range of computations that would greatly impact the Machine Learning industry given its reliance on these functions for computation. Given the complexities associated with creating and maintaining computers in a quantum state, the industry has remained in its infancy. However, given recent developments in Large Language Model (“LLMs”) capabilities and progression in the QC industry, there is concern about the combination of these two capabilities would create programs that would effectively have all the intuition and decision-making capabilities of humans.

As mentioned, if QC is able to achieve its theoretical goals in terms of speed, capabilities and storage, it would lead to a substantial leap in the capabilities of LLM programs; as it would dramatically reduce the time required “for generative AI models to learn, evolve, and create”.[3] Under current constraints, training LLM’s may take days or weeks. With the capabilities of QC, it is estimated that this could be achieved in hours or minutes — allowing for a substantial increase to the number of parameters that these programs can be trained on. The increased computing capabilities could allow LLMs to handle more complex optimization problems currently unimaginable, thus running the risk that these programs could become more intelligent and powerful than expected or planned for from an ethics and risk management perspective. Given the currently lax regulation on the management, controls, and ethics associated with LLM technology, rapid expansion of the technological capabilities would likely inhibit regulation from ever keeping pace.

While the advancement in AI associated with QC remains well within the control of individuals and regulators, QC could be used for far more nefarious purposes, namely hacking. As these computers will have the ability to manage simultaneous processes, they will be incredibly effective at solving the mathematical problems that support traditional cyber-security infrastructure at a rate which prevents traditional measures from stopping a breach. Additionally, quantum computers may be able to impact the core of data integrity and digital signatures, hash functions which are used to map data, and break them to provide access to a system. The functionality outlined above poses a substantial risk to the traditional cybersecurity that we rely on to ensure the privacy of not only personal sensitive information, but information critical to national security. Ultimately, if this technology was to be developed and utilized by bad actors, we would be woefully underprepared. In a sense, the development of this technology creates an arms race similar to that of Edward Teller.

There are a plethora of security measures, regulatory measures and planning steps available to us to decrease the potentially negative implications of implementing QC. Given the huge risk posed to traditional cyber security, the industry needs to create encryption that is resistance to quantum attacks, particularly around information that is critical to national security or sensitive. Further, we can ensure the privacy of our data by utilizing post-quantum cryptography, unique algorithms that rely on multiple layers of algorithms / mathematical code, which is thought to be secure against attacks from quantum computers. However, cryptography is only useful at today’s processing power and may become obsolete soon. Additionally, Quantum Key Distribution can be used to ensure secure communication between the cryptographic protocols, it “enables two parties to produce and share a key that is used to encrypt and decrypt messages”[4]. Like the ethical quandaries facing machine learning, we must ensure that there is a similar focus in the development of this technology to establish ethical guide rails for its use and development, either at an industry, regulator or intergovernmental level. Finally, today we need to begin developing a transition plan from traditional computing to QC for when the technology moves from testing to deployment, to industry/consumers and when / how existing systems and programs reach “quantum-safe” standards.

While we may not know exactly what the future holds it is likely that given what is at stake with QC, nations and companies will seek to continue to develop these capabilities as it will likely create the next generation of technology leaders. We can only hope that the cybersecurity industry and regulators act quickly in managing its influence, so we are able to maintain a level of personal security and control over LLM/machine learning.

[1]https://www.realclearscience.com/blog/2019/09/12/the_fear_that_a_nuclear_bomb_could_ignite_the_atmosphere.html#:~:text=He%20reasoned%20that%20a%20nuclear,the%20Earth%20into%20a%20miniature

[2] https://medium.com/@pareto_investor/nasa-just-shut-down-quantum-computer-after-something-insane-happened-6ddd6ff1d105

[3] https://gafowler.medium.com/the-revolutionary-quantum-leap-what-impact-will-quantum-computers-have-on-generative-ai-8f332bbfb61b#:~:text=Quantum%20computers%2C%20with%20their%20incredible,learn%2C%20evolve%2C%20and%20create.&text=Training%20AI%20models%20might%20take%20days%20or%20even%20weeks%20with%20classical%20computers.

[4] https://www.techtarget.com/searchsecurity/definition/quantum-key-distribution-QKD