Artificial intelligence is a new and very attractive field in which machines and smart devices are manufactured so that they can be used instead of humans. In the 1950s, pioneers in the field, Minsky and McCarthy, described artificial intelligence as a machine or program that performs the tasks assigned to it, using humans as its own intelligence.
This definition is very broad. It is why we sometimes hear arguments as to whether a product really uses artificial intelligence.
All AI systems perform at least a number of operations related to human intelligence. Planning, learning, reasoning, problem solving, knowledge, perception, movement, and manipulation, and to a lesser extent, social intelligence and creativity.
What are the uses of artificial intelligence?
Artificial intelligence is everywhere today, such as: for online shopping, to recognize the voice commands you give to virtual assistants like Amazon’s Alexa or Apple’s Siri, to detect what or who is in a photo, to detect spam And identify fraudulent credit cards.
Types of artificial intelligence
High-level artificial intelligence can be broadly into two broad types: limited artificial intelligence and general artificial intelligence
Limited artificial intelligence is what we see today as the kind of computers around us. These devices are intelligent systems that have been for, how to do certain things. But on the other hand, these devices are not able to program themselves, but they must have a plan.
These smart systems can be in Apple’s virtual assistant (Siri). Which has voice recognition. Visual systems in auto machines and search engines that offer products on what you have already bought. Unlike humans, these systems are for doing only what you have taught them. That is why these systems are called limited artificial intelligence.
What can limited artificial intelligence do?
A large number of applications related to limited artificial intelligence are emerging, such as: interpreting video feeds through drones that can penetrate into infrastructures such as oil pipelines, organizing business and personal calendars, responding to Simple customer questions, work with other intelligent systems to do things like book a hotel at the right time and place, help radiologists diagnose possible X-ray tumors, prevent inappropriate online content, detect wear and tear in elevators from data Collected by LOT devices. This list could go on and on.
What can general artificial intelligence do?
General artificial intelligence is completely different and is a kind of device adapted to human intelligence. These systems are a flexible form of intelligence that can learn how to do very different things. Everything from hairstyles to making statistical tables or arguing over a wide range of topics based on the accumulated experience of AI is more common in movies like HAL in 2001 or Skynet in The Terminator. But today this artificial intelligence does not exist and artificial intelligence experts are increasing rapidly to make this a reality soon.
How to implement artificial intelligence and machine learning
A survey of four groups of experts in the field in 2012. In which Vincent C. Müller and Nick Bostrom, researchers and philosophers in the field of artificial intelligence. They stated that general artificial intelligence (AGI) between 2040 to 2050 will progress by 50% and by 2075 will reach a ceiling of 90%. The group went even further, stating that 30 years after achieving AGI, intelligence called “Super intelligence” will be created that will go beyond any intelligence, including humans and other artificial intelligence.
According to the report, some experts are very optimistic about such predictions for our low understanding of the human brain. They believe that we are still centuries away from achieving AGI.
What is machine learning?
Extensive research has been done in the field of artificial intelligence. So that each of them complements the other, in other words, they complement each other. Using growing knowledge, machine learning is a process that uses a large amount of information into a system to learn to do something specific, such as voice recognition or captioning.
What is a neural network?
The key to the machine learning process are neural networks. These networks are brain-inspired systems in which several layers of algorithms are interconnected, called neurons, and can make changes to the input data as they pass through the layers. Learn a specific activity. During the training of these neural networks, constant changes are for this data until the output of the neural network is closer to what we expected. In this case we say that the machine has learned to do a certain job.
The subset of machine learning is deep learning. Where neural networks are transformed into large networks with a large number of layers that are trained by large amounts of data. It is these deep neural networks that are causing the current leap in computers’ ability to perform tasks such as speech recognition and computer vision.
There are different types of neural networks. With different strengths and weaknesses. Linear neural networks are a type of neural network that is particularly suitable for language processing and speech recognition. While ring neural networks are commonly used in image recognition. The design of neural networks is also evolving. With researchers recently designing a more effective form of deep neural network called short-term memory, or LSTM. It allows them to act fast enough to Use high-demand systems such as Google Translate.
Another area in which artificial intelligence is used is in evolutionary computing. It is borrowed from the famous theory of natural selection. In this theory, genetic algorithms undergo successive mutations and combinations over successive generations in an attempt to find a suitable solution to a problem.
This view has even been helpful in designing artificial intelligence; Especially the construction of an artificial intelligence by another artificial intelligence. The use of evolutionary algorithms to optimize neural networks is called neuro-evolution and can play an important role in helping the efficient design of artificial intelligence as the demand for artificial intelligence design by IT scientists is increasing day by day. The technique was recently demonstrated by Uber Artificial Intelligence Lab, which presented papers on the use of genetic algorithms that can be used to teach deep neural networks to enhance their problem-solving ability.
Finally, there are expert systems that are programmed with rules that allow them to make a set of decisions based on a large number of inputs that allow the device to handle the behavior of an expert. Imitate a human being in a particular domain. An example of such knowledge-based technology is autopilots, which automatically balance and propel an aircraft.
What are the power supplies of artificial intelligence?
The greatest advancement in the field of artificial intelligence in recent years has been related to machine learning, especially in the field of deep learning.
This is done to some extent with easy access to data; This development has intensified since GPU computing systems were used for machine learning. As they were able to perform very powerful parallel computing with this technology.
GPU systems not only provide much more powerful systems for teaching machine learning models. there are also now widely available as cloud services over the Internet. Over time, large technology companies such as Google and Microsoft have begun to use these specialized chips so that in addition to performing the work of artificial intelligence. They can also use it in deep machine learning.
For example, the Google Tensor Processing Unit (TPU) is one of these custom chips that, in its latest version, speeds up machine learning. For example, Google’s TensorFlow software library uses this chip, which not only extracts information through input data, but also speeds up machine learning.
These chips are not only for teaching models such as DeepMind and Google Brain systems. But also the basis of Google Translate and image recognition in Google Photo, as well as services provided to the public through Google’s TensorFlow Research Cloud to use They also make machine learning models. The second generation of these chips at Google’s I / O conference in May last year. This model with a set of new TPUs can build machine learning models that can be in Google Translate. Data processing time can be reduced by half when using the GPU.
What are the elements of machine learning?
As mentioned, machine learning is a subset of artificial intelligence and is generally into two main categories: supervised and unsupervised learning.
1- Supervised learning
A common way to teach artificial intelligence systems is to teach them using a large number of labeled examples. These machine learning systems are powered by huge data that has been annotated to highlight favorite features. These may be pictures that show whether they have a dog or not, or a series of footnotes to indicate whether the word “Base” refers to music or fish. After training, the system can apply these tags to new data; For example, when we upload a new dog photo to the device, it automatically tags the dog.
This process is supervised learning. The task of tagging data is the responsibility of online users or employees of platforms such as Amazon Mechanical Turk.
Training these systems typically requires extensive data. For example, in some systems, millions of data must be into a machine to learn a particular task; However, now that we are in the data age and dealing with an unlimited amount of data, this is possible. Educational datasets are very large and increasing in size. For example, Google’s image database is about 9 million images, while the YouTube video repository is about 7 million links. ImageNet One of the primary databases, it has more than 14 million classified images. All of this image data was over 50 years by 50,000 people. Most of whom were Amazon Mechanical Turk, who reviewed, tagged and sorted nearly one billion selected images.
2- Learning without supervision
In contrast to supervised learning, unsupervised learning uses a different approach. In a way that algorithms try to identify data patterns and look for similarities that they can use to classify that data.
For example, fruits that have the same weight or cars that have the same engine size can be together.
This algorithm is not pre-configured to select specific types of data.But rather searches for data that can be on its similarities. Google News, for example, collects stories on similar topics every day.
A relatively crude analogy for reinforcement learning is that when a pet performs an interesting trick, we reward it, and this reinforces that behavior in it.
In reinforcement learning, the system tries to maximize the reward based on its input data. So that it basically goes through a process of trial and error until it achieves the best possible result.
An example of amplifier learning is the Google DeepMind Deep Q-Network, which has been for the best human performance in a variety of classic video games. The system feeds the pixels in each game and determines various information such as the distance between objects.
The system determines a particular move as the best move and gives you points according to the difference between your move and the set move. You will get the perfect score and the best performance when you do the same thing that the system has set. For example, in Breakout, you have to place the paddle in the best possible place so that the ball does not get out of your reach.
Which companies are leading in the field of artificial intelligence?
With the growing role of artificial intelligence in modern life today. Each of the major technology companies is trying to provide more powerful machine learning technology for home use to sell them using cloud services over the Internet. Bring them.
Each of them regularly receives titles in a specific field of artificial intelligence; However, with DeepMind AI AlphaGo, Google has probably had the greatest impact on public awareness of artificial intelligence.
Which AI services are available?
All major cloud operating systems (Amazon Web Services, Microsoft Azure, Google Cloud Platform) provide access to the GPU suite for machine learning. Google is also equipping its system to allow users to use Google’s Tensor processing units. These processing units are operating in a way that can be for optimizing, learn, and implement machine learning models.
All infrastructure and related services are available to use these systems, including Internet-based stores, the ability to hold large volumes of data to build machine learning models, and data conversion services for parsing. And they analyzed visual tools to clearly show results and software that makes it easy to build models.
This cloud operating system even simplifies the construction of custom machine learning models. Recently, Google introduced a service that automatically builds artificial intelligence called Cloud AutoML. In this service, you can easily grab a tool in the software and drop it in a certain direction (drag and drop). It allows the user to be able to make image recognition models without any specialization in making machine learning models. Create.
Cloud computing and machine learning services are constantly evolving. In 2018, Amazon introduced a new AWS host designed to simplify the learning process of machine learning models.
For those companies that do not want to build their own machine learning models. But instead want to use artificial intelligence-based on-demand services – such as voice, vision and language recognition – Microsoft Azure offers a wide range of services. After that, the services provided by Google Cloud Platform and then AWS are pioneers. In addition to offering custom products, IBM is also trying to sell specific artificial intelligence with the goal of doing everything from medical care to minor tasks. The system has all the offerings under its IBM umbrella and has recently invested $ 2 billion in the purchase of a weather channel to complement all of its AI capabilities.
Which countries are leaders in artificial intelligence?
It is a great mistake to think that the United States tech giants are at the forefront of artificial intelligence. Chinese companies Alibaba, Baidu, Lenovo have invested heavily in artificial intelligence in areas such as e-commerce and automated cars. As a country, China is pursuing a three-step plan to make artificial intelligence the core of its industry. The program will be worth 150 billion yuan ($ 22 billion) by 2020.
The Chinese company Baidu has made huge investments in the construction of automatic machines. In the construction of which it uses a kind of deep learning algorithms called Baidu Auto Brain. After several years of testing, it intends to use these fully automatic cars in Unveiled in 2018 and mass-produced by 2021.
Baidu has also partnered with NVidia to use artificial intelligence to build a standalone car platform so that it can easily convert a car from a cloud to a real model, and they plan to use this model to build cars by manufacturers around the world. Use the world.
A combination of weak privacy laws, large investments, consistent data collection and data analysis by large Chinese companies such as Baidu, Alibaba and Tencent all mean that China in the not-too-distant future in the field of research in The field of artificial intelligence will be more superior to the United States. An analyst spoke about China’s chances of being superior to the United States in the future. He stated that China’s chances were 500 to 1 in China’s favor.
How can i start the AI?
All major tech companies offer a variety of AI services, including infrastructure to build your own machine learning model and work with it through web services that allow you to Access AI-based tools such as speech, language, vision, and emotion when needed.
How will artificial intelligence change the world?
1- Robots and drones
The desire to build robots that can operate independently and perceive and navigate the world around them means that there will be a natural overlap between the robotics industry and artificial intelligence. Although artificial intelligence is only one of the technologies in the robotics industry. The use of artificial intelligence is helping robots to enter newer fields. Such as automobiles (robots that deliver mail packages).
In addition, these conditions help robots to acquire new skills. GM has said it will build a drone, steering wheel and pedals by 2019, while Ford has pledged to do so by 2021. Wimo (a carmaker under the auspices of Google) has also stated that it will soon offer unmanned taxi services in Phoenix.
2- Fake news
We are on the verge of having neural networks that will be able to create real images of a person or replicate a person’s voice in a very precise way. Despite this technology, which will have its destructive social effects. It is no longer possible to look at movies or audio films as a valid document. There are also concerns about how to use this technology to misuse people’s images. As we see today, tools are for linking images of famous actors to adult films.
3- Speech and language recognition
Machine learning systems have helped computers recognize what people are saying with almost 95% accuracy. Recently, Microsoft AI and their research team reported that they have developed a system that can speak English in human registries.
Researchers are pursuing a goal with 99% accuracy to make human-machine interaction a fairly common norm.
4- Face recognition and monitoring
In recent years, the accuracy of face recognition systems has taken a huge leap forward. With one Chinese company called Baidu claiming that it can connect different faces with 99% accuracy, provided that Its face is clear enough in the video. While police forces in Western countries only attend major events using face recognition systems. Chinese officials are looking to implement a nationwide program to install CCTV cameras using artificial intelligence. Be able to identify suspects and suspicious movements.
Police are also using face recognition goggles. Although privacy regulations vary from country to country, it is likely that this intrusive use of artificial intelligence (such as the use of artificial intelligence to detect emotions) will spread to other countries over time.
5- Health care
Artificial intelligence can ultimately have a significant impact on health care. They help radiologists diagnose tumors using X-rays and help researchers identify disease-related genetic sequences or identify specific molecules that could be useful in making the drug. .
Tests on artificial intelligence technology have been in hospitals around the world. These include the IBM Watson Support Tool, which helps physicians make clinical diagnoses. The AI was at the Memorial Sloan Kettering Cancer Center and uses Google DeepMind technology, which is currently in UK National Health Centers. This system helps diagnose eye diseases and simplify the process of diagnosing patients with head and neck cancer.
Could AI take your job?
Although AI cannot replace all jobs, what is certain is that AI will change the nature of work. The only question is how quickly and deeply automation will change the work environment.
It is difficult to find areas in which humans can work but artificial intelligence does not have the potential. According to Andrew Neg, an artificial intelligence expert, “a lot of people are doing everyday, repetitive tasks.” Unfortunately, new technologies work well in everyday and repetitive tasks. “I can imagine a lot of unemployment in the next few decades,” he said.
There is evidence that there are already jobs that are over to artificial intelligence. Amazon recently launched Amazon Seattle, a cashless supermarket in which customers can just take what they want off the shelves. We need to see what happens to the 3 million people who work as cashiers in the United States!
Amazon has also facilitated the use of robots to improve performance within its warehouses. These robots move the shelves of products to human vans, which they select items to send to them. Amazon has more than 100,000 robots in its realization centers, and plans to add more. But Amazon also emphasizes that with the increase in the number of robots. The number of human workers in these warehouses has also increased. However, Amazon and other small robotics companies are looking to automate the rest of the jobs left in the warehouses. It means that over time, the increase in robots will result in a drastic reduction in manpower. It will have warehouses.
Fully automatic vehicles have not yet become a reality. But forecasts show that in the coming decades. The automotive industry will have about 1.7 million jobs. Regardless of the impact it has on taxis and couriers.
What do the robots do?
However, robots still do not have some of the simplest tasks that can be easily automated. Millions are now in managing, importing and copying data between systems, tracking and booking appointments for companies. As software advances in updating automated systems, the need for manpower will decrease.
For some, artificial intelligence is a technology that will increase the number of employees instead of replacing them. Not only that, but it’s a business necessity. Every job requires a manpower. And these robots can help him as a manpower assistant. For example, when a customer refers to your company’s workforce, the AI, which is into your employee’s ear as a headset, tells him or her exactly what the customer needs, even before the customer starts talking. To have done. These conditions are more effective than when artificial intelligence works alone.
There is also disagreement among scientists as to how quickly artificial intelligence can surpass human intelligence. The Future of Humanity Institute at the University of Oxford has asked several hundred mechanical experts to predict the ability of artificial intelligence for decades to come.
The dates they have shown show that by 2026, artificial intelligence will replace humans in writing articles. By 2027, truck drivers will be out of work. In 2031, it will surpass the human ability in retail. By 2049, the writing of best-selling articles, and by 2053, surgeries.
They estimate that by the next 45 years, artificial intelligence will be able to defeat all the work that humans do. And by the next 120 years, robots do all human jobs.