Image Credits: Yurchanka Siarhei/shutterstock.com
The landscape of quantum computing and artificial intelligence is and should remain a collaboration. Companies, including giants like Google, have been investing heavily in the realm of quantum computing and AI for years.
At the end of 2018, the spend on quantum computing stood at $2.2 billion, split regionally between the USA, UK, EU, China and Japan. The global spend on AI during the same period stood at $17.8 billion – a huge difference that would appear to indicate that developers and buyers have more confidence in AI then quantum computing. The $15.6 billion difference, however, does not recognize that the majority of AI spend lies within existing commercial systems (primarily commerce and marketing).
These types of AI are primarily reactive and have limited memory. The two types of AI that quantum computing can potentially advance are ‘theory of mind’ and the much-coveted ‘self-awareness’. Research from Google, IBM and others suggests that it is within these two types of AI that quantum computing can help.
Why is there a Perception of Hostility Between Quantum Computing and AI?
The answer lies within the organized chaos of quantum theory. Whereas a binary computer deals with absolutes (bits are always in a state of 0 or 1), quantum computing deals with certain-uncertainty (qubits can be in a state of 0, 1 or both – a superposition). Erwin Schrodinger is famous for saying he hated his theory which goes a long way in explaining why some scientists shy away from quantum research and computing. Yet Google, IBM and the Intel-acquired Mobileye are regularly publishing their latest theories and developments in quantum computing.
Back in 2017, Amnon Shashua along with a team from the Hebrew University in Israel argued that AI could help physicists better understand the quantum behavior of nature later producing mathematical evidence as proof. However, IBM, MIT and Oxford scientists also published a paper in ‘Nature’ titled ‘Supervised learning with quantum enhanced feature spaces’.
They claimed that by pushing ahead with more powerful quantum computers, the technology can perform feature mapping on incredibly complex data structures that binary computers are incapable of doing. Any scientist with a passing interest in artificial intelligence can tell you that feature mapping is a core component of machine learning. Now we begin to see the interlinked nature of quantum computing and AI.
The same authors went further, claiming quantum computing could create new classifiers to generate increasingly complex and sophisticated data maps. This would allow researchers to create more sophisticated AI able to identify data patterns that will always remain invisible to binary computers.
AI and Quantum Computing are Partners in Development
Elsewhere, scientists are already combining the two seemingly disparate technologies. Professor Michael Hartmann of Heriot-Watt University is a leading researcher in artificial neural networks. He hopes to use quantum computing to develop the first dedicated neural network computer (the theory that loosely models the makeup of processing power on the human brain).
Professor Hartmann believes quantum computing could lead to the birth of an AI capable of operating at speeds far beyond current technology. Because quantum computing takes advantage of sub-atomic particles that can exist in multiple states at the same time, theoretically the machines can think faster and wider than traditional computers.
To put it simply, quantum computers can see the spaces between the 0s and 1s that binary computers cannot. If successful, Hartmann’s AI could be one of the first artificial intelligence brains able to make highly complex decisions in nanoseconds or less.
The Uncertainty Pushback
The disparity in funding between AI and quantum computing makes it easy to insert an ‘us’ vs. ‘them’ mentality; a combative scenario that hinders rather than enhances development. AI researchers often point to existing quantum computers struggling to solve problems that binary computers solve in less than a second, to justify their dismissal of the technology. Scientists who adhere to this view risk missing out on the benefits of both.
However, it would be remiss to dismiss concerns about the potential of quantum computing. The uncertainty of the technology invites numerous errors. Quantum states are highly sensitive to interference from their environment leading to quantum errors difficult to predict. However, research conducted by a team at the Max Planck Institute for the Science of Light posits ways of using AI neural networks to correct quantum errors.
When the AI program ‘AlphaGo’ won 4 out of 5 games of ‘Go’ against the world’s best human player in 2016, quantum theorists and AI researchers rejoiced. Why? ‘Go’ has more combinations of moves than there are estimated atoms in the universe which required ‘AlphaGo’ to have more than just computer processing power on its side. It required neural networks to recognize, learn and predict potential moves. AlphaGo won because it was able to practice more potential combinations of moves before playing then a human could. AlphaGo displayed two factors that are essential in correcting quantum errors: speed and anticipation.
Researchers then took the learning from this and applied it to error correction in quantum computing. They discovered that AlphaGo inspired AI is capable of learning how to perform a task without being shown. As quantum error correction relies on a strategy that cannot always predict what will happen, the benefits of this type of AI in operating a quantum computer become exciting.
However, we cannot forget that the AI is not always able to predict and therefore correct quantum errors. How do we solve this? The solution arrived in the introduction of a second AI neural network. With the second network learning from the first, quantum errors become a little easier to anticipate and fix. The more AIs learn from each other, the stronger the potential of quantum computing.
The certain-uncertainty of quantum theory makes many R&D teams uncomfortable; even those that are actively developing quantum computers. If we allow ourselves to accept and perhaps become comfortable with this certain-uncertainty, we may yet create an AI capable of thinking, learning and even feeling as humans do.
Sources and Further Reading