Editor's note: With artificial intelligence making major breakthroughs in image recognition, Go, and poker, AI has now been hacked. Many new products and services are introduced using AI technology. However, many of them are fake AIs that sell sheep's meat. How can we tell whether it is true or false? Wallarm’s CEO Ivan Novikov provided his advice.
Artificial intelligence has become a popular word. People who sell now are always talking about it, and you often see the advent of AI on new electronic products and apps. This thing is also the most important tool in the field of cyber security.
This article aims to show you how to identify whether service providers really use artificial intelligence in their products and have enough data to help them form such technologies, instead of just using AI as a gimmick to attract new customers. This can also help investors understand the true state of the AI ​​at the supplier/startup side.
The following are five questions that need to be asked when evaluating AI products.
1. Can the company provide you with an independent presentation?
To avoid having someone manipulate your data in the cloud, you can ask for a stand-alone demo so that their software can process your data. If suppliers can only process data in the cloud, you have to evaluate the product with large-scale data to ensure that it is impossible for analysts to handle it.
Can you use your own data?
When suppliers use their own data for presentations, their systems often work very well. But this does not mean that the same system can handle your data perfectly. My advice is to use your data to test their systems in real time to ensure that the system works in real-world scenarios.
3. What are their data sources? How big is it?
Without big data, AI can't be a real AI; it's just like people can't survive without oxygen. For AI to work and develop effectively, it should provide real, large-scale data, and suppliers should provide exact numbers, parameters, and existing capabilities. It is really important to understand and validate data sources.
Common sources: In most cases, suppliers will use historical data such as stock market data, government data, and open source data sets. Here are some good sources:
Https://
Https://deeplearning4j.org/opendata
Http://archive.ics.uci.edu/ml/datasets.html
What are the details of the algorithm?
Ask for implementation details of the vendor algorithm. Ask how the data is encoded and decoded in the end. For example, find out how a recursive neural network is implemented in the product. These methods are not trading secrets, so your supplier can discuss the details in this area. If the supplier is not willing to discuss the technical method of implementation, it is a red warning signal.
5, ask them to see them without reference to the customer?
Talk to people who are already using their products to see how they are used. For example, for security, there is a huge difference between shielding and monitoring. Only in the case of false positives can the screening mode affect the business. Try to find customers who have similar needs with you.
Customers who actually use AI and have high standards of quality should be able to easily answer these questions.
We are still in the early stages of AI development, so it remains to be seen how this will advance in the future.
Laptop With Touch Screen,2-In-1 Laptop,Laptop For Education,2 In1 Windows Tablet
C&Q Technology (Guangzhou) Co.,Ltd. , https://www.gzcqteq.com