Artificial intelligence today is not just a joke about the enslavement of humanity. Thousands of businesses are moving to AI at an incredible speed, bringing production turnover to a new level and making our lives much more comfortable. But how can AI scientists teach the algorithm not just to produce a ready-made emulsion but to adapt to changing circumstances? The principles of machine learning have come to the rescue. Check this explanation of ML basics by an expert team, and we’ll move to the practical cases which has helped many companies.
Case 1: CompanionPro
As you know, obedient dogs are very likely to find their owners. Shelters are well aware of this, but few find the funds to pay for a large number of trainers. Michael Wang, director of the company, Companion, looking at the problem from the owner of the two cats, realized that this frightening omission requires an immediate solution. CompanionPro is an adaptive algorithm based on Tensor Flow, which can recognize the behavior of animals and reward them in the case of the successful command. Before the commercial release, the algorithm went through a humourous database of shelters, where it learned to determine what condition the dog is now, in “standing” or ” seated” as well as to select the voice timbre that will be the most appropriate. The cherry on the cake was the reward system, which empirically determined the amount of “goodies” for your favorite pet.
Case 2: Vitek “EYESYSTEM 1.0
I think everyone has ever had the urge to drink a cold yogurt in the morning. Workers in their production are no exception. How good that Vitek has developed an automated visual control system that has helped in such a difficult situation. The yogurt goes through two major stages, such as: checking the presence of the lid and the permissible level of the product in the bottle. Previously, the first stage was extremely uneconomical in time, but it was not immune from human error. Now the algorithm detected the loss with the highest accuracy, informing the entire plant about it. After analyzing a thousand lids, he knew each one by sight. But how would he know that I had not been top-up? Yogurt is yogurt, little or much of it. EYESYSTEM 1.0 could argue with you because looking at this damn bottle for the hundredth time, even you would be able to determine the compact differences in volume, right?
Case 3: 3D fitting and analysis of Wildberries
Imagine you come in after an exhausting day at work, and there are three dirty shirts in the wardrobe, one of which has a large brown stain from an invigorating coffee drink. The chief wants to see you at the parade tomorrow as if to spite you. What will you do? Run to the store? What a pity, you have an urgent task at work. Can you order a charming costume online? But how do you know your size when you last went to the store six months ago? If you at least slightly recognized yourself in this situation, then the 3D fitting from Wildberries will appeal to you. You need to point the camera at the desired part of the body, and a clever algorithm will adjust to you. What kind of magic is this? Not magic, but the thousands of bodies that the AI had painstakingly processed to produce such an impressive result. Many of you may say: “Well, 3D fitting is certainly good, but I’m not so lazy that I haven’t walked to the store yet. I’d rather buy cheaper clothes”. Great! And for this task, an all-powerful AI would do just fine. Just take a picture of your favorite thing, and you will see hundreds of analogs that will not be inferior to the original.
Case 4: Holos Servorobot
How many articles have been written, how many scientists have talked about the importance of careful treatment of the world around us, but people continue to harm themselves. If a person can’t reason with himself, then let a robot do it! The company Holos seriously took up this issue and presented a machine that is able not only to anticipate trouble but also to deliver dangerous cargo. Servorobot is equipped with a variety of sensors that can analyze the environment with high accuracy, determining the danger. Especially important in oil fields, where an unfortunate gas leak can not only leave you without cheap gasoline but also flatten the nearest neighborhood. In addition, this big guy is educated by the bitter experience of gigabytes of data, so if there is a war, your cargo will be safe and sound.
Case 5: SAT Logistics Optimization
Have you ever wondered how your new phone moves from the store to you? Of course, you know, there is a conditional point A, where you start the path, and point B, where it ends. A simple math problem, but why do you wait so long, sometimes even overpaying for delivery? The answer will be found if you look at this process through the eyes of the logistics department. Your parcel can generally make eight circles around the city until the courier still knocks on your door. A logical question would be: “Can this be avoided?» Can. This is done by logisticians, who, however, do not always perform their work in good faith. So it turns out that there is corruption and deception everywhere? Corruption is not quite, but the deception is successfully fought by SAT, which has implemented logistics optimization using geolocation. The Google Maps API, which conveniently allowed them to mark the place where the goods will be delivered on the map even without knowing the address helped this. Their architecture was able to analyze the routes of the company and choose the optimal chain of delivery of goods. Moreover, the cunning logisticians could no longer leave the best for their friends, because now big brother was watching them.
Case 6: Ecovacs Robotics Deebot
Perhaps I won’t surprise anyone if I say that vacuum cleaners have learned to clean your house themselves, but it’s also not surprising to hear that their intelligence wants the best. What annoyed you the most? Oh, yes, I know what you’re thinking. This “beautiful” moment, when this silly boy once again gets entangled in a neatly lying wire, gives the impression that he is your restless pet, and not the crown of technological progress. The head of the AI department of Ecovacs Robotics has faced this problem more than once since he decided to teach the robot to recognize the result of your sloppiness. Using a huge image data set, the robot learned to analyze small objects and masterfully drive around them. Now such insidious things as discarded socks, a black wire, or a red lego cube are left behind, saving you another nerve cell.
To sum up, I would like to say that this is not all that machine learning is capable of business. Many areas need the development of this technology more than ever. It can turn a loss-making enterprise into a gold mine, and in principle change the concept of business. The main thing is not to be a conservative and think again: “Maybe a call center of fifty people is not the best idea for a pizzeria?”