Skip to main content

"Relationship between AI, machine learning, and deep learning-how to reproduce human "intelligence""


The differences and relationships between "artificial intelligence (AI)," "machine learning," and "deep learning" can be understood smoothly by examining the history of AI.

"Human "intelligence" can be artificially reproduced by machines." --In 1956, the term "artificial intelligence (AI)" was born from the ideal of such researchers.

Since then, research has continued for more than half a century.

　During this period, it starts from solving mazes, puzzles, games such as chess and shogi (searching and reasoning), registers human knowledge as a dictionary or rules in a computer, and

tries to derive answers like experts ( Rule base and expert system) have been carried out.

　However, since humans make dictionaries and rules, it is not possible to register all the phenomena in the world.

In addition, there was a problem that it would not be possible to process it given the contradictory rules that are common in the world.

As a result, we were able to achieve results in a narrow and limited field, but it was far from "human "intelligence"" that could be widely applied in various

fields, and we did not achieve great results.

　After that, "machine learning", a method of analyzing the data in a specific field to find out the regularity and rules for classification, distinction, judgment and prediction, will appear.

　Although the concept of machine learning has been around for a long time, the performance of computers is insufficient, and it has not reached its full potential, and it has

improved its capabilities as computer performance has improved and methods have evolved.

Also, the spread of the Internet has made it possible to collect a large amount of learning data at low cost, which has also accelerated this research.

Humans have to specify what kind of characteristics AI should be able to focus on when performing classification, distinction, and judgment, that is, "selection and combination of characteristics (feature amount)"

Machine learning is a method that finds out the distribution, pattern, and regularity of this feature quantity by analyzing a large amount of learning data, and uses the result

　However, the feature amount had to be designed and registered by humans, and its skill greatly affected the result.

　After that, research on the function of the brain when humans recognize images proceeded, and "deep learning", a method of machine learning that applies the results, will appear.

　This technology can be created by analyzing the data for selecting and combining feature quantities.

Therefore, the performance can be improved as the amount of data increases, without depending on human ability.

　It can be said that this has given rise to the possibility of arbitrarily classifying the world's forests.

Labels:

Comments

Post a Comment

Popular posts from this blog

Why Apple Logo Is Half Bitten?

Why Apple Logo Is Half Bitten? Today we want to take you closer to the development of one of the world's most famous logos: the Apple logo. Steve Jobs, who grew up as a social orphan in Mountain View, California, was a big fan of the electronics industry while still at school. After he dropped out of college in Portland, he worked for several years at the computer company Atari. In 1976, Jobs finally founded Apple Computer Coporation with Steve Wozniak and Ronald Wayne. Since Jobs was temporarily a Frutarian in the 1970s, i.e. based on fruit, the apple in the company name was also derived from its own statements. But there are other theories for the fruity company name: Steve Jobs is said to have been on vacation shortly before the name was given, and to have been inspired by an apple harvest at the resort. Or else, the idea for the name came from one of his favorite bands, the Beatles, which was under contract with Apple Records. There is also a rumor that Jobs and...
Post a Comment
Read more

Top 6 Android Apps That You Can Make Money With

Top 6 Android Apps That You Can Make Money With When you look at your phone, you probably think it's just a reason to spend money. Your favorite widget will almost certainly cost you a lot of money, whether it be for your initial expenses or your current monthly payments. Fortunately, however, there are many apps that can be found on Google Play that can help you offset some of that cost. Thanks to Android Pit we can find out which 6 Android apps we can make money from. 1. Google Opinion Rewards With Google Opinion Rewards, you can earn Google Play credits by completing surveys. You will be notified when a new survey is taking place. Just take a few minutes to answer the questions. Once you've completed the survey, you earn credits that you can use on Google Play to buy apps, music, and more. 2. Rewardable Rewardable is being announced as an app for US users, but in reality it can be used by people all over the world. All so-called "virtual tasks" c...
Post a Comment
Read more

Big Data

Data collection and interpretation are very important to the seed breeding program. Without data and its interpretation, progress is impossible. Technology will create the ability to collect new data, and the effective innovation of a successful company will result. In addition, as technology evolves, the cost of technology decreases. For example, the cost per gigabyte of storage in 1980 was $ 500,000. Today the price is less than $ 0.03. This is a huge expense for the amount of data collected. In pig breeds, there is a wave of data collection. Basic data collection includes weight, number of births, birth year, age group, wean to service time, door count, and more. These features do not require sophisticated technology. Ultrasonic balance, depth, and intramuscular oil are being collected because of advances in ultrasound technology. Improvements in body and food quality have emerged over the years, including tools to improve pH, maripling, color, and digestion. All of these factors ar...
Post a Comment
Read more