Neuronet and quantum dots: how it will change our life
A lot of developments, which on the present day are actively implied in the everyday life, are based on the neuronet work. American physicists came up with an idea how the next generation neuronets can be built with the use of the quantum dots. What is it and why their development is so important we’re going to explain further.
What is quantum neuronet?
Researchers from the USA carried out experiments on the matter of сonductivity of contiguous semiconductor (quantum) nanoparticles in quasi-two-dimensional spaces. And they did an amazing discovery — quantum dots in the mesoscale can become the base for the construction of a new type of artificial intelligence (AI).
To simplify — now within the number of technologies, that are imitating the brain work, came up a new one. Recently some other researchers informed about the development of photonic chips, which are imitating the workflow of biological neurons. Now it seems like they have competitors with a different concept.
To understand the way it works and the reason why it is so important for the high-tech world, we need to figure out the work of quantum neuronets of artificial origin. The idea of quantum neuronet is search for a certain objective function of its minimum through the adiabatic evolution of the quantum system from a predetermined initial state.
“Quantum dots” are semiconductor parts in size of a couple of nanometers. Charge carriers are very limited in space unlike larger semiconductors. As a result, the role of quantum effects increases near those dots. Changing their size and form, we can control its conduction and luminescent features.
Putting aside all this complex physics, let’s draw a conclusion: theoretically chips based on this technology will be several times less (maybe even thousand times), but the same number of times more powerful than current silicon analogous, which are almost reached their limit of the computing power.
But let’s get back to AI. At the current moment there are a number of diverse options of constructing neuronet quantum algorithms. Although practical hardware was developed only for Hopfield quantum network. For its creation SQUID technology was used. Due to it neuronets can be scaled, what allows using them for commercial purposes.
The weak point of such a development is low operating temperature of 30–80 mK which requires significant cooling costs. This drawback prevents compact implementation.
AI of quantum type: usage options
One of the usage options for quantum type neuronet is the model that E. Berman and his co-authors came up with. Quantum neuronet in it presented in the form of a one-dimensional array of quantum dots based on GaAs which interacts with the thermal reservoir of the substrate.
This model has one significant drawback — the impossibility of a local influence on a particular quantum dot, as well as the absence of a controlled interaction potential between two neighboring points.
There is also a version of quantum neuronet that functioning on the base of two-dimensional array of quantum dots. Here, the interaction of individual quantum dots with the substrate through its photons is possible. Along with it between two neighboring points there is a controlled type of dipole-dipole interaction. The control of this interaction is accomplished through a change in the transfer lines of charge carrier concentration by electrostatic means or by means of plasmons.
The quantum neural networks themselves are represented in several types of computing devices:
- Quantum computers. Here operations with quantum systems, which exists in two different states, are performed by means of quantum gates. Those network computers can resolve different types of tasks;
- Quantum simulators. They are analogues of computers. But here models the behavior of the complex system which is researched through more simple system that is controlled by simulators;
- Quantum neuronets (adiabatic quantum systems). Considered the most successful in terms of commerce. It is something in between of simulators and network computers.
Currently the most claimed neuronets are quantum ones.
The cause of interest
Quantum systems raise interest in terms of presence of two significant effects:
- Long-range interactions of the tunnel type. They create hopping conduction format. If the system is constructed from such dots than quantum resistance will be much less than multiple. If the voltage is applied to such a system, the charge carriers will be able to move not only to neighboring nanocrystals, but also “jump” to other points within a radius of up to 100 nanometers;
- Blinking current. This effect means that after a while nanoparts can be turned on and off.
The usage of such effects will allow to create more advanced nueronet. Although those technologies are not public yet and modern developers are forced to look for new solutions, combining available tools.
For example, the NeuroSeed project chose another approach — inclusion of artificial intelligence based on nueronet (Deep Learning) in blockchain. This symbiosis forms decentralized, transparent and generally accessible multi-level platform in the format of a global supercomputer. As a result there should be brand new digital product which will allow to develop more effective AI for resolving user’s problems.
