IBM has just announced that its quantum computer has achieved 64 quantum volumes, reaching a new milestone in quantum computing. Honeywell, the mask factory, has lost its advantage in ion trap.
IBM has just announced that it has reached a new milestone in quantum computing, the highest 64 quantum volume at present. Compared with last year, the performance of its quantum computer has doubled.
as early as March this year, Honeywell said, “in the next three months, we will launch the highest performance quantum computer in the world.
the quantum volume score of its latest quantum computer is 64, twice that of IBM and Google competitors. In 2019, IBM’s quantum volume will reach 32.
the concept of “quantum volume” was first proposed by IBM. Quantum volume is an index used to measure the performance of quantum computers, rather than just the number of quantum bits as the measurement standard. This index is proposed to get rid of the relatively outdated evaluation method in the industry.
the influence factors of quantum volume include quantum bit number, gate and measurement error, cross communication of devices, device connection and circuit compilation efficiency. The larger the quantum volume, the more powerful the quantum computer is, and the better it can handle the complexity of time and space.
to better understand the advantages and disadvantages of IBM and Honeywell’s quantum computing, we must first know that they adopt two completely different technical routes.
at present, quantum computing is mainly divided into two categories: solid state devices and optical routes. Google, IBM and Intel belong to the “solid state device route school”, while Honeywell’s ion trap technology route belongs to the “optical device school”.
at present, more than 80% of quantum computers in the world adopt the route of solid-state devices. Solid state devices have obvious advantages in compatibility with classical computing, and optical routes such as ion trap are more often used in scientific research.
the energy level of superconducting quantum circuit can be intervened by external electromagnetic field, so it is easier to realize customized development of circuit. Moreover, the current integrated circuit technology is very mature, and the scalability advantage of superconducting quantum circuit is obvious.
but there are also some problems. Due to the uncloseability of quantum system, environmental noise, flux bias noise and other uncontrollable factors, quantum dissipation and coherence weakening often occur. In addition, the physical environment of superconducting quantum systems is extremely demanding. For example, ultra-low temperature is an inevitable problem in the process of superconducting quantum computing.
in addition to IBM, companies such as Google and Intel are also actively carrying out superconducting quantum research. The bristlecone quantum chip released by Google quantum artificial intelligence laboratory can realize single bit gate manipulation on 72 qubit length, and the optimal fidelity of single qubit gate reaches 99.9%.
previously, some experts said that Honeywell’s claim to have the world’s fastest quantum computer is suspected of speculation, because it has not solved any problems in the world, it is just a parameter improvement.
Honeywell’s ion trap uses the interaction between electric charge and electromagnetic field to restrain the movement of charged particles, and takes the ground state and excited state of the restricted ion as quantum bits.
quantum states are stored in a single ion trap and information is read from it. Qubits can interact with each other directly through their movement in the well, or through the emission and absorption of light and microwave.
although the development of ion trap technology can be traced back to 1980, the realization of quantum computation by using ion trap technology was first proposed by Austrian quantum scientists circa and zoller in 1995.
in 2003, the laboratory realized the use of detuning laser beam irradiation and laser cooling to control the non gate. In the same year, the laboratory successfully implemented Deutsch Jozsa algorithm using ion trap technology for the first time.
However, the ion trap technology also faces many problems. Due to the instability of external laser and electromagnetic noise, the coherence of quantum bits is weakened, and the ion trap is difficult to coexist with multiple ion chains and has poor scalability.
Honeywell’s leap in quantum computing in ion wells is partly due to its breakthrough in a key technology project in 2015, which is able to capture charged particles in superposition state by using laser.
IBM’s quantum volume has doubled every year since 2017. In 2019, IBM said its quantum computer, Raleigh, had a quantum volume of 32, up from 16 a year ago.
although quantum computing is still in its early stage, there have been many innovations and breakthroughs. What role will it play in the field of artificial intelligence?
how does this help us improve the classical machine learning model? If you try to use a small number of face data sets to train the classical face detection model, but the performance is not very good, we can use quantum enhanced generation model to enhance the data set, so as to significantly improve the performance of the model.
“sense perception” refers to that computers can actually understand whole sentences, not just words, but also extend their perception ability to whole phrases and even whole articles. The further improvement of language models such as Bert depends on the powerful computational power of quantum computing.
for example, telling a computer what a cat is can be a bit challenging because it can’t learn. If you train the neural network with enough “cat slices”, the computer will be able to recognize other cats correctly.
for some algorithms, even exponential growth can be achieved. Quantum computers are not only faster to perform tasks, but they can perform tasks that were previously impossible.
quantum physics is probability theory, and financial market prediction is also to some extent. Quantum computers can establish candidate models without restriction, and have the potential to better predict the distribution, so as to obtain more accurate answers.
the basic idea is that some problems require AI to generate new data to make decisions. To solve this problem, we may need to propose a potential model to solve the problem of probability distribution under unknown conditions, which is a strong point of quantum computing.
as far as machine learning is concerned, classical computing and quantum computing have the potential to work together. We can use the flexibility of cloud computing and the powerful computing power of quantum computing to work together.
both classical and quantum computing have advantages, and the current development of quantum computing makes it a part of the solution. With the passage of time, both methods of calculation will continue to develop.
the ability to speed up the workload on traditional GPU and ASIC, while also using the ability of quantum computing, can make quantum computing faster and more reliable. Science Discovery