Africans should endorse quantum computing to tackle scientific problems specific to the continent.
There are many unsolved problems in science, business, and society. Since the beginning of modern computing, classical computing systems have helped address many of these thorny challenges.
However, certain complex problems, such as accurately simulating chemistry, have stubbornly remained out-of-reach even as classical computing power has grown in leaps and bounds. Ultimately, the ability of classical computing is limited by fundamental information processing principles. If we want to go beyond these limitations, we have to use systems that are built on revolutionarily different principles.
Nature relies on quantum mechanics
Classical computing theory is built on a Newtonian worldview. With the discovery of quantum mechanics, a strange counter-intuitive physical theory, the tantalizing question of whether it could form a new foundation for a powerful computing paradigm was raised. The question was answered in the affirmative in spectacular fashion with the discovery of quantum algorithms that are exponentially faster than their classical counterparts.
One of the best examples of this is the very same first inkling that quantum computers might be more powerful than classical systems: simulating nature. Classical computers with their clear-cut ‘0’ or ‘1’ building blocks are not well-suited to simulating nature because nature, being quantum mechanical, has some strange properties such as an electron being in multiple places at the same time.
An exponential number of possibilities
Fully keeping track of even one hundred interacting electrons is provably impossible for classical systems, because of the exponential number of possibilities. Quantum computers are up to the task, precisely because they leverage these strange phenomena.
This actually forms a useful heuristic for characterising the impressive, though limited, class of problems that quantum computers are slated to crack open in the near to medium term: those that are easy to specify (small data), but have an exponential amount of possible arrangements to run through (big compute) before arriving at an answer.
Generic applications of quantum computing
Chemistry simulations clearly fall into this category, by simply asking how a handful of electrons arrange themselves. But so do many other computational tasks such as optimization, where the answer sought is the optimal arrangement of some small set of variables, out of exponentially many possible combinations.
The diagram on top of this article gives a sense of the landscape of generic applications of quantum computing, particularly in relation to classical computing.
Easy problems, and hard problems
Problems that are “easy” to solve classically are represented by the white region. They range from simple arithmetic operations and sending emails to the most advanced GPU and supercomputing applications.
The light-blue area represents problems that cannot be addressed effectively using classical computers, requiring unachievable amounts of memory or time. The dark blue area represents problems that quantum computing can easily address, which encompasses the white “easy” classical ellipse.
Classically super-hard problems
This does not mean that quantum computers would completely replace classical computers, since many “easy” classical problems do not benefit from any quantum speed-up. The dark-blue region only partially overlaps with the hard-for-classical light-blue region.
Crucially, this implies that quantum computers cannot solve all problems that are hard for classical computers. Finally, the dark-blue region extends beyond the light-blue region representing the fact that there are known quantum-achievable problems that can even be classified as classically super-hard.
Three areas for applications
For another view of quantum applications and to drill down into domain-specific examples, it is helpful to group applications into three broad areas: Simulating Nature, Machine Learning (including AI), and Optimization.
- Simulating Nature
Quantum Mechanics is a framework that describes the physical world. For many systems, the quantum properties manifest themselves and need to be accurately accounted for in order to model the system.
Quantum computers offer a window into directly simulating such systems, from the tiny (quarks) to the gargantuan (black holes). Along the way, many practical domains are touched, including chemistry (e.g., the design of new catalysts, for carbon capture), biochemistry (e.g., the understanding of photosynthesis and drug discovery), and material science (e.g., the design of new batteries). Even for systems where the quantum nature is hidden and classical mechanics suffices, there is the promise for quantum computers, as linear algebra super-machines, to offer speed-ups, e.g., in fluid dynamics.
- Machine Learning
Many of the primitives behind today’s machine learning algorithms involve linear algebra, which may benefit from quantum computers. Beyond these there are distinctly new twists on familiar algorithms, such as Quantum SVM, Quantum Boltzmann machines, and Quantum Homology that take advantage of the huge compute space of quantum computers.
- Optimization
Discrete algorithms such as Linear and Quadratic programming and certain combinatorial algorithms are able to take advantage of quantum search. Problems that can be framed as Bayesian or Monte-Carlo problems also benefit from quantum techniques.
In general, optimization algorithms have applications far and wide, examples from the finance sector include portfolio optimization, risk analysis, and loan and credit scoring.
Race to build working quantum computers
The application discussion above has glossed over one important question, namely where are we now in the race to actually build these devices. Since quantum computers take advantage of delicate properties of nature, they are prone to destructive noise in ways that classical computers are not.
Fortunately, there is a promising, though challenging, long-term technique that solves the noise problem: quantum error-correction. In the short term, however, we have to contend with noise by directly designing algorithms that are robust to the noise.
Mitigate the effects of noise
At the very least, all such approaches have to complete the quantum part of the calculation within a narrow window, before noise dominates. In addition, certain algorithms creatively mitigate the effects of noise.
Chemistry and optimization are some of the applications that have found noise robust algorithms, opening the possibility of important real-world advantages within the near term.
Quantum applications for Africa
Most, if not all of the potential applications of quantum computers, are relevant to the African context. However, it is worthwhile contemplating which applications stand-out with respect to two development considerations.
Firstly, Africa is rich in natural resources and human talent. Historically, Africa has been unable or even prevented from beneficiating these natural resources on home soil.
African genetic diversity
A strange-to-state but priceless “natural resource” is African genetic diversity. There are quantum algorithms that offer significant speed-up in the field of genomics and Africa should try and position itself as one of the first movers in the field of applying quantum computers to genomics.
The second consideration is that African problems that do not overlap with immediate global concerns are often underserved in terms of research funding and effort. An example of this would again include genomics but also other problems such as HIV research on the strains found in Africa. Chemistry simulations on quantum computers promise to open doors for HIV drug discovery that may forever remain unreachable to classical computers.
Genomics
“Africa is the cradle of mankind, the African populations are the origin of others, and harbor the greatest genetic diversity on earth.”
Genomics is the qualitative study of the expression and interaction of genes from the entire genome. It has been surprisingly difficult to map genes to phenomes (the final biological expression).
Reasons for this current opaqueness include the difficulty in simulating the folding and docking of the proteins encoded for by genes (for which quantum chemistry can help) and the complicated independencies and interaction between genes.
Connections between genes and effects
In order to tease out the connections between genes and effects, genetic diversity is a veritable asset. Genetic diversity allows for genome-wide association studies to average out spurious connections and reveal true connections.
Furthermore, greater diversity increases the chances of discovering natural immunity to diseases. Such nature-bestowed resistance may be learnt from, in the quest for novel health solutions.
Genetically-tailored medication
“There is a host of challenges facing genomics research in Africa, from a shortage of skilled human resources, to a lagging cyberinfrastructure for high-performance analysis.”
One should, hopefully in the near future, include quantum computers as cutting-edge “cyberinfrastructure.” Quantum algorithms may be able to speed-up a number of calculations (for example, Sequence Alignment), related to the search for genotype-phenotype correlations including Quantum Hidden Markov Methods, Quantum Boltzmann Machines, Quantum SVM, Quantum Bayesian methods, and Quantum Homology.
Finally, genomics is also an example of the dangers of research neglect to Africa. Therapeutics designed for one population (“genetically-tailored personalized medication”) are sometimes not effective on other populations. If Africa wants to employ the latest technology for the improvement of the health of its populations, it seems it must master these tools for itself.
Drug Discovery
“Nowhere is the fight against HIV more critical than in Africa, which is home to 70% of the world’s [current infections] as well as 66% of all new infections occurring globally.”
As the world has seen with HIV and COVID, locally mutated diseases pose a serious threat to global health. Thus, ironically, the lack of research attention for finding effective treatments to African strains eventually has global impact and makes the underfunding of African disease research, short-sighted.
Part of what fuels this disparity is “market forces” combined with the fact that different strains sometimes predominate in different geographical regions. For example, the HIV-1-D and HIV-2 strains are found mostly only in Africa.
Quantum computing against HIV
With quantum computing promising spectacular improvements in drug-design, it would then seem prudent to employ it in the service of tackling African diseases. For example, with HIV’s ability to mutate rapidly, there is a constant search for new antiretrovirals as the virus develops resistance to the old.
In the process of designing new antiretrovirals, a crucial step is performing chemical simulations of the binding between the virus protein and the antiretroviral. Such simulations are notoriously hard classically, and sometimes completely ineffective. Quantum computers offer more accurate simulations allowing for a better drug-design workflow.
African scientists already undertake extensive research into HIV, it is time to consider adding quantum computers to their own arsenal, in the fight against the disease and neglect.
Ismail Yunus Akhalwaya and Waheeda Saib, IBM Research Africa
This article has first been published by the African Physics Newsletter. © American Physical Society.