- 21 Eki 2015
- 477
- 1
What is Artificial Intelligence (AI)?
In this subject, I mentioned some of the technologies that companies use to make their business easier and reduce costs.
I'm not going to do a deep history review for AI here. You can find it in a lot of places. There is no single definition defined and decided for AI. AI means artificial intelligence, but today some groups also use it as augmented intelligence.
Andrew Moore of Carnegie Mellon University describes AI: We thought we needed the human mind to make computers using technology and engineering.
Where is AI used?
I wanted to give some examples. We can duplicate these examples and there are many varieties, but I wanted to keep it basic and simple.
1. Facial Recognition
If the resources surface is clear (nothing to interfere with), it is close to human performance. Some companies use it not just as facial recognition, but to examine people's emotions. There are even companies that use micro expressions.
2. Automatic Text Recognition
Let's say you're driving a company car from place to place in the company and you had to buy gas. You have to pay out of your own money. You also get a receipt from the gas station so that you can indicate your expenses. Instead of going and reporting that you have spent the following amount on these receipts, you just take your receipts pictures away and AI does the rest.
3. Autonomous Cars
We can say that many companies invest in this technology and contribute to its development. It does not have 100% security due to the variability of roads, equipment and weather conditions, but it is developing a little more every day and it may be one of the new technologies that enter our lives.
The photo shows the rate of error related to recall.
You can see how Artificial Intelligence has evolved over the years from the photo. We can say that it has passed even human in recent years and has achieved a very good performance.
Another use is for brands to be easily notified when they are mentioned somewhere. By scanning your company's logo or an image with its name, you can easily and quickly be informed about where it is mentioned on the internet. Thus, you will a**** the trouble of browsing the internet and searching for keywords.
Facebook, Google, Whatsapp etc. using a keyword search for your voices by getting authorization from many places, whether from the permissions on the sites you use or from the permissions you give on your phone, whether it is a product, device or game related to the subject you are talking about, it has happened to everyone to take that word and advertise it after you have finished your speech. They do this again with the help of Artificial Intelligence.
The last example is a technology developed by Blue River company. The picture on the left shows us a photo of the farmer's field. In the picture on the right, you can see the work of the artificial intelligence that decides where to dispense or not. Red squares show weeds, green squares show the plant that needs spraying. In this way, the company saves 90% as it will only throw insecticide in green squares.
So how do companies train these AIs?
Machine Learning
They use experience to teach machines to improve and make computer algorithms more precise. We can divide it into 3 parts. There are learning styles that vary according to their usage areas and the data we have.
1. Supervised Learning
They use this method to train many Artificial Intelligence. We send previously prepared and labeled data to Artificial Intelligence and we say "If you find something like this nice, if not, separate it".
This is divided into 2 parts.
Classification
Regression
a) Classification
When Artificial Intelligence encounters data, it looks at the data given to it and puts the incoming data into one of these groups. It allows you to classify unclassified data here and helps you convert this raw data into usable data in the future.
b) Regression
Regression is used for data with numbers. Let's say you are a car dealer and want to classify your cars information. In this case, it makes sense to use Regression as you will get a lot of numerical data. Because, be it the engine size, the mileage it travels or the price it will be sold, all of them are numerical data and things that can change over the years.
2. Unsupervised Learning
There is no preload data. Artificial Intelligence is expected to infer a meaning and order from the data given to itself. This process is longer, but is best suited for uncertain and unpredictable situations. Unsupervised Learning has only one branch.
Clustering
Many companies do not use this learning management, but as I said, as the areas of use change, the technique used also changes. This learning method is generally used on commercial websites. Let's say you are a company and you sell by sharing your products on your website. You can examine the paths of customers entering your site and direct them to the departments where they can buy something indirectly.
3. Reinforced Learning
They make Artificial Intelligence open to learning and throughout the process, people tell it how to learn. The best example of this is the Thai robot Microsoft left on Twitter to train in 2016. Of course, the result didnt go well, but we can see that companies are trying to develop this method as well. The company had to terminate the project as Artificial Intelligence learned from them as people wrote offensive and racist messages to Tay. Since the company doesnt want to throw away all the work it has done, it has replaced it with an Artificial Intelligence named Zo but this time they took precautions to prevent such a thing from happening again.
I talked about some technologies that companies use and are developing and how they train Artificial Intelligence to use these technologies. It excites me that it will occupy a large part of the market in the future. However, it has not yet reached the running stage, but progress is promising.
Thanks for reading!
