This article is one of the articles in a series ‘Kiwicode startup stories’, and you can read the last article ‘Start a business right after graduation and achieve nothing in a year. (startup story 1)’ at 'https://www.indiehackers.com/post/start-a-business-right-after-graduation-and-achieve-nothing-in-a-year-startup-story-1-d352a6c159'.
Within a week of Google’s TensorFlow release, we found this new project. Individual developers and small teams like ours can now use machine learning to solve problems like image identification, voice recognition, and text recognition. We understand that this could be a huge opportunity for us. Artificial intelligence (AI) has game-changing potential. It is clearly a business opportunity with greater prospects when compared to handheld smart devices we considered before.
Adrian presented me with a number of AI application scenarios. Machine learning, in my opinion, is a method of processing a large amount of data and making a prediction of a result, with the prediction accuracy gradually improving as the model is deepened. Machine learning is officially used to retrieve this value from data. The black box state, according to Adrian, is a disadvantage of AI technology. We don’t know the internal parameters, and there is no logical reason for them. We have a basic understanding of AI technology as a result of our discussion, and we quickly came to an agreement to work on some specific AI applications.
We were prepared to fight, but we ran into a problem. AI technology applies in many ways, but what kind of problem can be solved best for AI technology? We don’t have any answers to this question. I don’t think I can stay in the room with my head, I need to go out to meet potential customers, sell AI Technology to them, and see whether someone would like to pay the bill or ask certain specific questions. It happened that one of our university classmates, Dong told us that he knew a university professor in Yangzhou and that he might also have some needs. Why don’t we go to Yangzhou? (Yangzhou is a small city hundreds of miles from Shanghai). Adrian and I agreed to go to Yangzhou with Dong to visit potential customers.
Adrian, Dong, and I arrived in Yangzhou and immediately met a man, only to discover that he was not a professor, but rather a member of the university’s staff. He had no idea why we were here, but he was patient and listened as we talked about AI technology. He expressed his admiration for the entrepreneurial spirit of some of our lads after a few hours of embarrassing conversation, but he had no idea what AI was or what he had to do with AI. Finally, he asked us to stay here for a few days while he inquired of other professors about the need for such tech.
Dong is optimistic that there may be a large market waiting for us to explore; Adrian is pessimistic; I believe we can wait and see what happens. A few days later, the university’s staff contacted us and informed us that a university professor wanted to meet with us and that we would need to give a presentation. This time, Adrian carefully prepared some small demos, while I prepared a slider. We are confident that we will be able to persuade this university professor to work with us.
We received the professor’s email, which contained some problem descriptions and instructions, as well as a 443kb excel about wheat form attachment, after the evening celebration. They hope that by using historical data, we will be able to forecast this year’s wheat production. We feed the information into the machine learning model and wait for the results. Because the sample data is too small, Adrian believes it will be unable to predict successfully. Dong believes the outcome will be favorable. As a result, we failed. It’s as if our model is a guessing game. It is unable to predict future outcomes at all, and its accuracy varies greatly. Adrian claims that the model cannot be successfully trained with such a small number of data samples. The only way to do so is to increase the sample size. I inquired of the professor, who stated, “These are all the samples.” Even after attempting to transform a large number of parameters, the desired result remains elusive.
Everyone was a little letdown. This project was a colossal flop. Machine learning is not suitable for solving this customer’s problem. I think we should be more honest and tell the university professor directly that we have failed. This is a necessary responsibility. When you face customers, you need to be realistic and don’t exaggerate the function of your product. If your product can’t bring good results to customers, don’t lie, admit your shortcomings frankly and work hard to improve.
We failed this AI technology attempt because we were unable to determine which questions you can assist customers with. Chinese market’s understanding of AI technology is still in its infancy in 2016. We’ll have to devote a significant amount of time to persuade customers that this technology is revolutionary. We did not gain any customers, however, due to a lack of specific usage scenarios.
The time has come to the summer of 2016. It has been a year since I graduated from university. We have no income or make decent products. It is really too risky to start a business without any experience.
If you have just graduated from university and have a dream to change the world, you can join any company and begin with a job that will be more beneficial to you.
Although I have achieved nothing this year, I have also gained a lot of experience and knowledge. We need to make some income the next time.
The next article will be published on Friday at a weekly interval. Simply send me some claps and feedback if you enjoyed my novel.
The next article will be published on Friday at a weekly interval on my medium