Page Nav


Classic Header




Quantum computing for artificial intelligence

 There are two major, future-defining fields in science and technology today: quantum computing and artificial intelligence. Both have the p...

 There are two major, future-defining fields in science and technology today: quantum computing and artificial intelligence. Both have the potential to revolutionize our lives, and their development can have a major impact on human society. The creation of quantum computers for everyday use is still a long way off, and even household use may not make sense, but the field is rapidly evolving to play an important role in computing. Artificial intelligence systems have already undergone major development and have established significant roles in our daily lives, sometimes in an unnoticeable way. Can these two fields work together?

Artificial intelligence could help find ways to make quantum computers, just as AI can help in any field of science and technology, but can quantum computers make better AI?

Would AI be more powerful using quantum computing technology? Can quantum computers help achieve the desired goal and help create the most feared of all, Artificial General Intelligence? If so, it would be the ultimate, most capable machine created, it could be the most capable machine ever created.

We already have a general intelligence machine, the biological brain. Although less complex than human brains, animal brains are still capable of high-level artificial intelligence tasks such as image and sound processing and recognition, autonomous navigation, and learning capabilities. Animal brains are even capable of generating consciousness. The human brain is the most capable of all natural brains. It is capable of creating abstract thought and all that goes with it, conscious recognition, complex language. It builds and uses society's knowledge and creates technical civilization.

Artificial intelligence technology attempts to mimic the workings of the biological brain. Its development has made significant achievements using learning functions modeled on biological neural networks. The artificial deep learning systems have already surpassed human capabilities in the function of recognition. They are able to identify correlations in large data sets better than the human brain. As a result, artificial intelligence has conquered humans in areas once considered human territory, including chess, Go, Jeopardy, and complex strategy games.

These are significant achievements, but artificial intelligence systems lag behind human and even animal brain capabilities in areas where correlations are not so clear, where decisions are necessary and based on insufficient information, not to mention consciousness and all related capabilities, such as attention and volition. We can see the struggle of artificial intelligence in these areas. These functions require more conceptual and independent functional mechanisms, in areas where today's artificial intelligence systems are less developed.

To move forward, AI systems will probably need more computing power, or more likely, new system architectures and methods of working. The core and paths of today's AI are correct, otherwise we could not have achieved all that we have, yet biological brains are capable of accomplishing more complex-looking, qualitatively different tasks with seemingly less computational power.

Quantum computers with enhanced computational capabilities with necessarily different computing architecture could come to the rescue. Quantum computing technology is a quantum leap in computing power for certain computational tasks. Quantum computers are built on different theoretical foundations than digital computers. Digital computers use only two basic states, quantum computers use these two states and infinite more between them as basic states, hence their computational power can be exponentially higher than regular digital computers.

Quantum computers may help artificial intelligence systems to work more efficiently, but the advanced biological brain functions are not based fundamentally on quantum effects. Sometimes it is fashionable to think about the brain as if it were a quantum computer, especially with the ability and working method to create consciousness. Both fields, the quantum world and consciousness, look mysterious, and so we might think that they are somehow connected, but biological brains are not fundamentally quantum computers. Even though everything in our world is built on quantum foundations, a light switch is not considered a quantum machine, even though it utilizes quantum tunneling effects.

Biological brains are similarly built on quantum effects because biological processes are chemical reactions between molecules, and these interactions fundamentally take place in the quantum realm. However, neurons, the building blocks of all biological brains, work strictly in the classical sense, and we are able to simulate neuron functions efficiently with classical computers. The power of the biological brain does not lie in its supposed quantum nature, but rather in its structure and its complex neuronal interactions.

But can we use quantum computing and techniques that mimic biological brain functions to achieve higher-level AI capabilities? The scientific instinct says yes, but how to do it is the real question

The quantum world is based on probabilities. Quantum computers are evolving quantum systems of quantum superpositions. Superpositions are complex mixtures of different quantum states. A superposition can combine many quantum states. After this combined superposition, the whole system behaves as one, evolves together, and represents a multi-complex quantum state. This multi-complex quantum state is capable of representing a complex information structure at once. How this complex state evolves corresponds to the calculations of the quantum computer.

A quantum system - like a quantum computer - remains in this complex state until a measurement is made, until the complex state collapses and takes on a definite value according to its specific probability distribution. In the case of the quantum computer, this collapse is itself the result of the computation. The result will be a state value according to the specific probability distribution of the given quantum system. If the quantum computer is properly set up, then the most likely result will be the value we are looking for.

Theoretically, quantum computers cannot do what regular digital computers cannot do, but quantum computers can perform certain calculations exponentially more efficiently than regular digital computers. Quantum computers are particularly well suited for "needle in a haystack" tasks such as reverse search.

Today's quantum computers attempt to use quantum logic gates, similar to the logic gates of classical computers. Quantum logic gates, like their classical counterparts, take inputs and produce outputs. The difference is the calculation of the output from the input. Classical logic gates use simple binomial logic, quantum logic gates use quantum logic to calculate the output from the inputs. The calculations are different, but the structure that connects logic gates is similar. The quantum logic gate computing structure can be extremely useful for some specific computing tasks, such as classifying sets based on parameters. 

As computer theories suggest, quantum computational methods are not the magic tool that we inevitably need to perform high-level cognitive functions, assuming that biological brains are fundamentally special computers, and high-level cognitive functions are fundamentally specific computational methods. However, if we don't stick to classical logic gate architectures, quantum computing can provide remarkably useful working methods for effective artificial intelligence computation.

The most promising field of today's artificial intelligence systems is the deep learning architecture. The deep learning architecture is modeled similar to the biological brain structure. 

Artificial intelligence systems are systems optimized for performing large amounts of similar calculations, computers that use specific mathematical procedures based on probability calculations. Today's artificial intelligence is essentially pattern recognition. It is capable of recognizing undefined or undefinable correlations in large data sets.

Two fundamental mathematical computational functions in AI are global minimum search and parameter trial and error. An ideal tool for both functions is the quantum computer. The ability of artificial systems to detect hidden or unknown correlations is based on methods of mathematical probability calculations. Their limit is the speed and efficiency of these calculations. Quantum computers can perform these specific calculations exponentially faster than digital computers. Because the build-up and collapse of the superposition is instantaneous, they can provide fast solutions to AI-specific mathematical problems that would take extremely long to solve using regular computational methods.

There is a striking overlap between the purpose, function, and essence of operation of quantum computers and artificial intelligence systems. Today's primary applications planned for quantum computing are communication security and secrecy. However, the most suitable application of quantum computers seems to be the realization of intelligent functions, the creation of non-biological artificial intelligence, and ultimately the creation of Artificial General Intelligence.

No comments