Snehal Shinde, co-founder of Mezi, a chat robot shopping app, recently wrote an article to share his experiences with artificial intelligence startups. The artificial intelligence company Mezi, founded by Schend, has obtained A-round financing. Its entrepreneurial experience is of great reference value.
About Mezi
Mezi combines artificial intelligence with machine intelligence to understand user shopping preferences to create a seamless, personalized shopping experience. The company provides users with a free application that simulates the experience of shopping with best friends through smart, user-friendly chat conversations. This app plays the role of user shopping assistant, mainly for consumers who do not have time to shop.
When shopping is required, users only need to send a text message to Mezi, and then Mezi can filter through thousands of product options based on the user’s target price, preferences, and shopping habits. This helps users place orders and simplifies the user’s The entire shopping process. In addition, the Mezi robot can accept and understand a wide range of user needs, and can even apply conscious design and artificial intelligence to learn the user's conversational style, thereby simulating the user's way of speaking.
Mezi was founded in 2015 and is headquartered in California. In July 2016, Mezi announced the completion of a $9 million Series A round of financing. Investors included Nexus Venture Partners, Saama Capital, American Ventures and Angel Investors Amit Singhal and Gokul Rajaram. . The following is the original content:
Mezi co-founder Snajar Sindh
Artificial intelligence ushered in the outbreak period. Although it was only a scene in science fiction, today, all sensible entrepreneurs seem to be chasing this technology. As the famous editor and futurologist Kevin Kelly said: "The business plans of the 10,000 startup companies in the future are easy to predict: Integrate X with artificial intelligence."
However, while there are many subversive opportunities around artificial intelligence, this technology still faces many challenges. Anyone who is trying to train a new type of artificial intelligence system will tell you how to remove people from it is a key obstacle.
As we all know, large-scale artificial intelligence systems often need to be equipped with people as safety nets. They are responsible for standing by at any time in the background, and once the system fails, they are responsible for taking over control. Even Facebook, the pioneer of artificial intelligence and smart chat bots, will occasionally be staffed to ensure a high-quality user experience.
Although this model effectively helps artificial intelligence to achieve development, automated artificial intelligence systems are the ultimate goal. How can we achieve this goal? From the experience of personally developing Mezi, we should first find a suitable vertical area, and then we must also grasp the promotion rhythm.
Choose the right vertical area
Many people have a lot of misunderstandings about artificial intelligence and believe that it is possible to develop a kind of "universal artificial intelligence." But in fact, the artificial intelligence industry is more likely to disperse countless highly specialized artificial intelligence systems, and is unlikely to be dominated by a universal artificial intelligence.
Because of this, choosing the right vertical area is the most important. When we first launched Mezi, we focused on shopping and tested a few vertical areas such as clothing, gifts, and travel. We soon discovered that the travel industry has key information to train artificial intelligence systems and expand its scale.
First of all, the travel industry is highly fragmented. There is no single application that can take care of ticketing and a comprehensive travel experience. The industry is also highly commoditized because metadata on flights, hotels, cars, and travel have been displayed in a structured way that facilitates AI learning.
Artificial intelligence can develop in such an environment. By deploying natural language processing, deep learning, and neural network technology, we can more easily understand the user's intentions, actions, locations, and other characteristics, and have high accuracy. This structured environment also facilitates us to understand user preferences and runs a series of machine learning algorithms to provide highly personalized recommendations.
Another property that needs to be considered when choosing a vertical realm for artificial intelligence is a business use case. Specific to the travel industry, a considerable part of the market is driven by business travelers. This not only creates an opportunity for repeated use - it is a key goal of any software application - and there is already a precedent for personal assistants. Business travelers have relied on personal assistants to help them book tickets, which also makes it easier to adapt to technical solutions that provide a better experience.
Hold the promotion rhythm
It must be borne in mind that the transition from a human-supported machine to a fully-automated machine is not a day's work, but rather requires a step-by-step process.
When Mezi was first introduced, almost all user activities were integrated into the background with human factors. We hired cool and experienced experts and then developed a set of artificial intelligence to observe and learn their interactions with users, including their intonations, choices, and emojis.
Soon, we can hand over some of these tasks to Mezi - initially a simple task, and then gradually transition to a complex task. Every time artificial intelligence is tasked with a new task, humanity can be liberated, focusing on more complex challenges and identifying new areas for artificial intelligence to learn.
When we started a pilot cooperation with a global top financial service in early 2017, we experienced a dramatic increase in traffic, and our already-performing artificial intelligence began to suffer problems. Our natural language processing system is not ready for such diverse requests. Fortunately, we have a lot of people on hand to deal with this situation while also training natural language processing systems. We continue to iterate every hour to make the artificial intelligence system more intelligent.
In the end, that pilot became a useful stress test that helped us find the strategic areas for the future and directions for improvement. This also gave us an important lesson: It should start small and then gradually expand the scale.
When expanding the scope of application of artificial intelligence, do not write code for all interactions from the beginning. Instead, use humans for guidance and then gradually implement automation. This model can provide important artificial scaffolding for your artificial intelligence system, allowing it to learn its own development and create a set of coded systems for rapid iteration. Later, as the scale of the artificial intelligence system is expanded and the intelligence is improved, it is only necessary to withdraw one set of auxiliary facilities and then proceed to the next stage.
Axial Fan,Axial Flow Fan,Axial Blower,Axial Exhaust Fan
Hangzhou Jinjiu Electric Appliance Co Ltd. , https://www.jinjiufanmotor.com