.png){kind=logo}
It is a well-known fact that an AI model is only as good as the data it is fed. Model-based AI is judged against benchmark datasets and is compared on specific parameters. Most researchers depend on algorithmic improvement rather than on data quality. Most companies hesitate to adopt AI due to this very reason – lack of good data. Andrew Ng, the prominent AI researcher believes data-centric AI to turn this game on its head. And there are pretty many good reasons for this proposition.
An AI cycle involves, for most of its part involves preparing high-quality data, which makes up at least 80% of the AI lifecycle. Even though training the AI model constitutes 20% of training the model, it has been the center of focus for researchers and businesses as well.
The benefits of a data-centric approach go beyond this single advantage. For example, labeling data which is basically done by humans comes with certain inconsistencies. It introduces noise in the data which can prove detrimental to the AI cycle. This happens because of differences in interpreting guidelines. Andrew says, by identifying the points of disagreement for labeling, discrepancies can be reduced to a great extent. When a model is fine-tuned based on faulty data, it is bound to have lower accuracy. In a data-centric approach, as the data is pre-engineered for designing a model, one can expect data accuracy apart from achieving model optimization.
Data engineering has a lot to do with fixing noisy data. In most cases, AI projects fail to see light due to a lack of sufficient data. When a small dataset is used for a model, it makes room for a lot of noise and therefore errors. When you have pre modeled dataset, the noise level can be reduced significantly. Data-centric AI addresses all the fringe cases with clearly defined data focusing on the model architecture.
Join our WhatsApp Channel to get the latest news, exclusives and videos on WhatsApp
_____________
Disclaimer: Analytics Insight does not provide financial advice or guidance. Also note that the cryptocurrencies mentioned/listed on the website could potentially be scams, i.e. designed to induce you to invest financial resources that may be lost forever and not be recoverable once investments are made. You are responsible for conducting your own research (DYOR) before making any investments. Read more here.
