The most popular and comprehensive Open Source ECM platform
Let us know more
Data-Centric AI: Shifting the Focus from Big Data to Good Data
Quality of data may be the number one requirement for data analytics and AI project success, but not far behind in the list of requirements is the volume of the data that is analyzed or used for AI training. Curernt AI algorithms are often successful only when massive amounts of data are used.
John McQuaid, journalist at the Wilson Center, wrote that “to achieve better and better results, deep learning algorithms need bigger and bigger datasets — the bigger, the better. Computer scientists found that in many cases deep learning scales: The more data they use, the more accurate it gets.”
But the need to have large data sets in order to achieve good results can be a roadblock. Unless you work for a large tech company or the Chinese government, your access to data may be limited. Trying to build large datasets is hard. Often the data related to a problem simply is not available or there are data privacy restrictions that limit how and how much data can be collected.
The need for AI techniques that don’t require massive data sets is growing.
Andrew Ng, Stanford professor and early promoter of AI and pioneered the use of GPUs in AI, said that “we know that in consumer software companies, you may have a billion users in a giant data set. But when you go to other industries, the sizes are often much smaller. From where I’m sitting, I think AI–machine learning, deep learning–has transformed the consumer software Internet. But in many other industries, I think it’s frankly not yet there.”
Ng gave the example of AI image recognition to explain the importance of the value of small carefully curated data sets. “Architectures built for hundreds of millions of images don’t work with only 50 images. But it turns out, if you have 50 really good examples, you can build something valuable, like a defect-inspection system. In many industries where giant data sets simply don’t exist, I think the focus has to shift from big data to good data. Having 50 thoughtfully engineered examples can be sufficient to explain to the neural network what you want it to learn.”
