Lei Feng network (search "Lei Feng network" public number concerned) : According to the author of this article to be word teacher Chen, mainly talked about how to consider a good start-up company's three criteria.
Now that artificial intelligence is very hot, it should be quite a fire. Many startup companies claim to be artificial intelligence companies whether they are really smart or not. Investors chasing artificial intelligence are desperately saving money to these companies for fear of losing profit opportunities. Fashion is indiscriminate, and it is fueling. So how do you think about an artificial intelligence startup?
Talk about personal ideas, how to look at an artificial intelligence startup company. Yes, it is a purely personal perspective and I hope it will help.
First , look at the background of the founders and the team, whether prior to the AI-related education, training, and practitioner experience and practice.
Artificial intelligence is a high technology after all. It does not believe that if you look at a few books, read several articles, and participate in several high-end conferences, you can become an expert and you can create artificial intelligence products.
Have you heard 10,000 hours of law? It is written by writer Gladwell in the book "Different": "The genius in people's eyes is remarkable. It is not supernatural, but it is a continuous effort. Ten thousand hours of exercise is anyone. From the ordinary to the extraordinary necessary conditions.". To become an expert in a field, it takes 10,000 hours. Proportional calculations are: If you work eight hours a day and work five days a week, then it takes at least five years to become a field expert.
For artificial intelligence practitioners, 10,000 hours may be a little bit more, but there is no long-term founder and team that works in the field of artificial intelligence, and it is not realistic to claim to build products with artificial intelligence. Therefore, the first point of investigation is the experience and experience of the startup company's team and founders.
Second , see if there is an artificial intelligence algorithm with its own IP, and there are innovation points that the algorithm has not mentioned .
Using deep learning or other machine learning algorithms is not necessarily artificial intelligence. It must have its own original ideas, its own innovation, and its own characteristics.
When Google first invented PageRank, other search companies were counting the number of times the link was cited, and used the number as a weight to guide rankings. However, PageRank abstracts the relationship between links into a Random Walk model rather than a simple count. This is the insight, the product came out more than competing products, rave reviews.
AlphaGo uses the same Monte Carlo search as other chess algorithms. However, it has introduced a deep learning algorithm to guide the pruning and evaluation of the chessboard, which exceeds any other algorithm and reaches a new height.
The algorithm is important, but it requires a rational look at the barriers to the algorithm. The algorithm usually exists in the human brain, and there is a risk of being compromised. At the beginning, it may have advantages. For a long time, it is not allowed to be leaked by employees and thus copied by competitors. The risk of the algorithm is also high, and it is not a once and for all thing. In addition, the accumulated experience in algorithm tuning is also very important, but it may be lost as the employees are lost.
Third , to see if there are data with advantages, the data may be a long-term accumulation of industries and a large amount of general-purpose data. It may be that many industry expert knowledge exists in the form of rules or knowledge maps.
There is no accumulated large amount of data, no expert-organized knowledge, and the data is not well cleaned, structured or even tagged. It claims to be an artificial intelligence company and does not know where the intelligence can emerge.
IBM's Watson can do precision medicine. They bought a lot of medical data companies before, accumulated and analyzed a large number of journals, papers, cases, medicines, treatment programs, and so on, so that Watson can use algorithms to create miracles.
Google, if you do not have a large number of historical user search and click data, then clean up and analyze the click data, create a user click model, and then use the model to improve the sorting of search results, Google's search results will not make many users satisfied, and brighten up .
When a company does a lot of basic data collection and accumulation, and more and more data and knowledge are accumulated, it is possible to take a step toward artificial intelligence. Moreover, these large amounts of data, such as Terabytes, Petabytes, Exabytes, Zettabytes data, are still afraid of data being stolen and stolen? This data barrier will be a long-term sustainable competitive advantage. And the more data there is, the more accurate the model will be without overfit, and the intelligence that comes out will be unexpected.
Experienced teams, unique algorithms, superior data, and especially data, use these three criteria to see if a company that claims to be an artificial intelligence is really worth it . If none of these three things are known, don't say clearly, or don't give yourself a fashionable artificial intelligence label. Of course, in the end it's not that artificial intelligence isn't important. Making a useful and amazing product can prove everything.
Lei Fengwang Note: This article is issued by Lei Feng, authorized by the Chinese character editor. If you need to reprint, please contact the original author, and indicate the author and source.
Ring And Fork Type Insulated Terminals
Ring And Fork Type Insulated Terminals,High quality insulated terminal,copper tube terminal
Taixing Longyi Terminals Co.,Ltd. , https://www.longyicopperlugs.com