Where artificial intelligence is used successfully

Artificial intelligence has already shown itself in many areas of activity: with its help, hearing aids filter out extraneous noise, and navigators help to find the most convenient and shortest path, online stores analyze customer preferences based on previous purchases, and medical programs help to diagnose and prescribe treatment. 

Modern speech recognition systems allow virtual assistants to answer simple questions and fulfill simple requests. Optical recognition of typed and handwritten text helps sort e-mail and digitize old documents. Face recognition is widely used at state border checkpoints of different countries.

The US State Department's facial recognition system contains over 75 million photographs of those who applied for a US visa.

Computer translation is still far from perfect, but it can be used for certain needs. The new translation software is based on statistical self-learning systems that automatically analyze language constructs. The programmers working on these systems don't even know the languages ​​they work with.

Intelligent dispatching is another area in which artificial intelligence has been successfully used for several decades.

In 1991, during Operation Desert Storm, the Defense Advanced Research Projects Agency's automatic planning and operations program saved the Department of Defense the equivalent of a thirty-year investment in artificial intelligence.

Robotics is another area where artificial intelligence is used. Today, our planet is inhabited by over 10 million robots: pets, cleaners, rescuers, surgeons and industrial robots.

The future of artificial intelligence

In recent years, interest in artificial intelligence has increased significantly, which is likely to give impetus to work on powerful hardware, next-generation software, and developments in related industries such as neuroinformatics.

One indicator of interest in machine intelligence is the fact that 160,000 students from all over the world signed up for the free online course on artificial intelligence taught by Sebastian Thrun and Peter Norvig from Stanford University in 2011, and 23,000 of them received certificates. about its end.

The path to the supermind

Scientists identify several possible ways to create a superintelligence:

Improving artificial intelligence.  It has already become obvious today that in order to achieve a superintelligent state, the system must be self-learning and capable of constant self-renewal and improvement: the early version will create its own improved version, which will then be continuously updated and improved.

Full-scale simulation of the human brain. Through scanning and subsequent modeling of the structure of the human brain, a powerful intelligent program can be created, which requires the following steps to be performed sequentially:

creating an extremely detailed scan of the human brain;

transfer of primary data from a scanner to a computer => subsequent automated image processing => reconstruction of a three-dimensional neural network that provides the thinking process of the human brain;

running the reconstructed brain structure on a powerful computer.

It will be possible to speak about a positive result of modeling if it is possible to reproduce the intellect, memory and personality traits of thinking of the “owner” of the modeled brain. The created model of the brain can exist either in virtual reality, or be realized in real reality as a separate "organ" for the robot.

Strengthening the human brain.  The third way to surpass modern human intelligence is to improve the human brain through genetic selection. However, this is too long a journey - a generational change occurs every 20-30 years, that is, the results can be judged only in the next century. This process can be artificially accelerated by using the iterative embryonic selection method, which includes four stages:

study of the genotype and selection of several embryos with the desired genetic characteristics;

extraction of stem cells from selected embryos and their transformation into sperm and eggs, which will mature for about six months;

crossing of sperm and eggs and the appearance of new embryos;

repeating the previous steps until significant genetic changes are obtained.

Thus, it is possible to complete the process of genetic selection, which in natural conditions would have lasted for more than one century, for several years.

Development of human-machine interfaces.  If the human brain is connected to a computer through the implantation of an implant, it can become a superintelligence.

Such interfaces are already being created for medical purposes: a device is implanted into the brains of patients with Parkinson's disease, which stimulates a specific part of the brain using an electric current.

Despite the fact that implants can optimize the thought process, provide fast and accurate solutions to complex problems and high-speed information transfer, it is unlikely that such interfaces will be widely used in the future. Potential complications are the main obstacle: infections, electrode misalignments, bleeding, and cognitive decline.

Building a collective mind.  Another possibility of creating a superintelligence is the gradual strengthening of organizations and information networks. However, it is more likely that this path, like genetic selection, will lead only to a slight improvement in human abilities, but will not make people superintelligent in comparison with our contemporaries.

Nick Bostrom is convinced that the existence of multiple possibilities for creating a superintelligence increases the likelihood of its occurrence. Moreover, it is most likely that the superintelligence will become a product of the development of artificial intelligence.

Strengths of machines

Even subtle differences in brain volume and the number of neural connections can provide significant benefits. The best illustration of this statement is the intellectual and technological achievements of man in comparison with the ape.

Machines are already many times superior to humans in a number of parameters:

Frequency. The speed of computing in the human brain is determined by the frequency, which in humans is 7 orders of magnitude lower than that of a microprocessor: 200 Hz versus 2 GHz.

Intercom speed.  The speed of brain axons is no more than 120 m / s, while the processor cores can transmit information at the speed of light - 300,000,000 m / s.

The number of computing elements.  The human brain contains about 100 billion neurons. This amount is due to the size of the cranium and metabolic processes. The machine can be huge, for example the size of a house, or it can use additional power through high speed cables.

Memory.  Human memory does not retain more than 4–5 pieces of information; moreover, long-term memory can malfunction. The capabilities of computer memory are incomparably higher.

Other advantages of the equipment: type of engineers and salaries

Transistors are more reliable than biological neurons.

The human brain gets tired after several hours of work, while microprocessors can work without time constraints.

The machine can be optimized for different tasks, but the architecture of the human brain is unchanged from birth, and any changes require significant time investment.