AI won’t leave people unemployed, but improve work effectiveness – interview with Pig's Big Brother startup founders
Being disappointed in the absence of IT innovations in the Russian production, Sergey Strelnikov and Sergey Kantserov decided to act on their own.
Sergey Strelnikov worked hard in the meat industry of Russia. He was the Chairman of the Board of Directors of Perm Pig Farm and Chairman of the Board of the National Union of meat processors of the Russian Federation, and saw problems of the industry from inside. It was then that Sergey joined forces with Sergey Kantserov, who for seven years helped companies in businesses development and participated in the development of three residents of Skolkovo.
So Pig's Big Brother was created. It is a startup that uses artificial intelligence to optimize fattening of pigs.
In an interview for the Internet of Things forum, the founders told why now we eat pork of poor quality, how Pig's Big Brother fights against this problem and what people who are building a startup in Russia should do.
“Pigs are excellent objects for computer vision systems, since they are exactly the same”
In October 2017, one of the organizers of the Open Innovations forum in Skolkovo, the Russian Economic School, invited us to the event. It made a double impression on us. On the one hand, we saw that the development level of modern IT innovations is quite high and that the technologies of artificial intelligence and the Internet of Things have moved so far forward, but on the other hand, practical ideas and examples of the introduction of these technologies were almost absent.
It caused dissonance in our minds, but paved the way for a lively discussion. As a result, we decided to come up with something innovative and absolutely practical, with a visible and tangible economic effect.
We were especially interested in the computer vision. Projecting this topic to the previous experience in agricultural business, we decided that computer vision systems are the ideal way to monitor the development and care for animals. First of all, for pigs.
Pigs are excellent objects for computer vision systems, since they are exactly the same. There are only a few basic breeds of pigs, so it's enough to create a database for one of the breeds and project it onto a new generation.
We immediately shared this idea with our friend, a specialist in deep learning – Sergey Nikolenko, who was going to publish his book Deep learning. We discussed with Sergey the problems in the development of the pig breeding industry and the activities where computer vision can be used.
Sergey said that computer vision, such a Big Brother for pigs, is quite a realistic idea. So that the startup appeared.
“The meat sector has always lacked structured data about what happens to animals”
First of all, we had to find a development manager. This position was given to Sergey Nikolenko, since it was he, who confirmed that such software can be developed for reasonable sum.
Later on, in mid-summer of 2018, Sergey Kantserov, who has 7 years of experience in local and international development of startups, including Skolkovo residents, joined the startup.
Now the team has been formed, all the organizational challenges and ways of solving them have also been adjusted. The startup has developed Proof-of-Concept, which means that such a system is possible and will work. We just have to get Proof-of-Value: confirmation of sufficient economic effect and scalability of the solution.
To analyze the market before launching the project, we used Sergey Strelnikov's previous experience. He always noted that the meat sector has always lacked structured data for analyzing what happens to animals and how they grow.
It was not difficult to analyze the market, calculate the number of potential customers and see the economic effect that can be achieved using our project. Of course, we assessed the possibilities for the project development not only within Russia, but all over the world, since Russia accounts for only 3% of the world's pig breeding.
“There is no traditional opportunities to measure the weight of animals every day and track the behavior of each animal”
The complexity of such a startup development wasn’t really visible at the first stage. The fact that the idea itself is absolutely understandable for the pig breeding industry representatives. All of them understand that now pig breeding industry has no possibilities and tools for operation management of the growth and breeding of animals.
Currently, there is no traditional opportunities to measure the weight of animals every day and track the behavior of each animal. As a result, we don’t see what is happening to them and cannot detect problems with their health in advance. However, Pig's Big Brother can help here.
The first function of PBB is the identification of each animal in the group. The second is timing of animal behavior through tracking poses and analyzing behavior patterns. The third and most anticipated function is the ability to estimate animals remotely every day, without any interference in their peaceful breeding. This allows breeders to avoid stress of animals when assessing the most important indicator – their weight.
Now we are focused on the stage of fattening animals, but we’re planning to expand the capabilities of the computer vision system to earlier stages of breeding, for example, when piglets live with the sow or growing.
We also plan to expand the system to other types of animals: sows, boars, replacement stock, etc. To ensure that the entire pig population of the business process is under supervision and the system guarantee the absence of any emergency situations.
“The first country, where we are going to launch the project is the Russian Federation”
Naturally, the first country, where we are going to launch the project and already developing a pilot project is the Russian Federation. It was chosen because of the understanding of the system's capabilities. In general, the world situation is as follows: more than half of the total number of pigs is bred in China. Many pigs are bred in South-East Asia. About a quarter of all pigs are in America and about 15-20% in Europe.
In terms of the enterprises size, Europe has quite small farms located in different countries. At the same time, such countries as China, Russia, Canada, the United States, Brazil, Thailand, followed the path of creating very large pig breeding complexes.
Enlargement is economically justified, and it is large enterprises preferring small-scale maintenance of animals, where the computer vision system is most in demand. It is absolutely impossible for a person to keep track of hundreds of thousands of animals that are fattening simultaneously.
“If we avoid needling and feeding animals with hormones, pork will be better”
The main problem in the production of meat is the lack of ability to control the weight of animals and optimally manage the processes of feeding and maintaining the microclimate. Now we do not have feedback from animals. They show their contentment and discontent only through their own behavior. To analyze these changes, one need a trained computer vision system.
To equalize the growth of animals, pig breeders use hormones. Often, to equalize the health and minimize the risks of diseases, all animals are needled with antibiotics.
In such a way they pull up the laggards in the growth to an average level, but thereby reduce the level of quality of meat healthy pigs.
Modern technologies, which primarily use biological and chemical preparations, average the quality of meat to a certain safe level, which affects the taste of the product itself. It's no wonder that now pork from large producers has practically lost its taste.
The product has reached an averaged quality, without any taste characteristics. It can be easily seen if buying a similar product from a familiar farmer and a small producer for comparison.
If we avoid needling and feeding healthy animals with hormones, pork will be better. Our startup takes pork production to a new level and thereby significantly improves the quality and taste of meat.
“We achieve high accuracy in weight estimation due to a huge database of images of animals of known breeds”
The most memorable moment was when we realized the true value of our product. The idea of all the project functions, which we are talking about, was developing for some time. At first, we thought that to increase the effectiveness of fattening pigs, three components are enough:
- monitor animal behavior;
- prevent extraordinary situations timely;
- build deeper analytics due to the dynamics of shifting different poses.
Only then we came to the idea already directly related to our project, namely, to create a database for deep machine learning, so that the computer vision system can continuously estimate weight of each animal. It is a very important aspect, because compared to conventional weighing, determination of weight using the computer vision system is more accurate.
We achieve high accuracy due to a huge database of images of animals of known breeds with pre-known weights and linear dimensions. For example, conventional weighing-machines reflect not only the mass of the animal, but also the mass of what is inside it: that is, what they were eaten and drunk. Therefore, traditional weighing will show results that are highly distorted by the daily intake of water and food.
The computer vision system is aimed at a more representative indicator: the change in linear dimensions. This is achieved by using data of animals with empty stomach in the database we created for machine learning.
We analyze their behavior, linear dimensions and thus create video of animals of different weights, ages, born at different times of the year with relevant marks. The more extensive this physical parameters database is, the more accurate the results will be presented by the system.
“AI is designed not to leave people unemployed, but to improve work effectiveness”
The project is guided by two main principles.
The first principle is that artificial intelligence, which we implement in business practices, is designed not to leave people unemployed, but rather to help employees to perform their work more correctly and optimally in order to improve economic results.
The second principle is the increase of public welfare. We believe that artificial intelligence is an instrument that isn’t meant to redistribute public welfare, but to improve it.
The food that animals eat doesn’t go into the ‘atmosphere’, ‘heat’ or ‘landfill’, but it increases animal weight. Everything turns into a product – pork. Currently, about a quarter of the food that pigs eat, goes nowhere and disappears from public goods. In other words, what are pigs, in fact, doing? They ‘warm up’ the atmosphere.
Pig's Big Brother will look after every animal so that it doesn’t worry about anything, and a person rarely comes in its sight. Thanks to this, animals will grow more calmly, they won’t have stress, and the food will go more directly to meat and the expected weight gain.
“When you start working on a project, the hardest thing to do is to convince yourself that the project is a promising one”
To get bitten by the bug and start to create your own startup, first of all you need to critically evaluate the level and prospects of the idea that you are going to develop. Unfortunately, today there are a lot of half-baked ideas that don’t cover relevant problems in business. In fact, these are just additives to already existing, narrowly popular systems.
To do things that simply help a person to solve some minor problems at the end-user level is not a significant idea a worthy startup.
If you are sure that the idea can really optimize business processes in enterprises, which can improve public welfare, it can be developed elaborately and endlessly.
After all, what is the most difficult thing to do in the beginning of the project development? It's difficult to convince yourself and even a few closest friends that your project is a promising one. If you receive confirmation from a sufficient number of people, then the idea is worthy.
What should beginners in the market do first of all? It seems to us that first of all it is necessary to take on the role of the investor, who will be ready to invest in the project without confirmed studies, prototype or MVP.
You need to build your business plan so that even at the maximum level of risk within three to five years, investors could see a multiple increase in their capital share.
Don’t forget that the main factor of decision-making for the investor is risk/profitability balance. For example, a socially important health project will attract less funds than a project for optimization of drilling and production processes even if the first one has 10% higher implementation perspective and yield.
To meet with the founders of Pig's Big Brother startup, come to the Internet of Things conference, which will be held in Moscow on September 25.