Synthetic data is poised to become a major industry within the next five to ten years. Gartner predicts that by 2024, 60% of data for AI applications will be synthetic. This type of data, along with the tools used to create it, holds significant untapped investment potential. Here’s why.
Feeding Data-Hungry AI/ML
We are on the brink of a revolution in how machine learning (ML) and artificial intelligence (AI) can grow and be applied across various sectors and industries. The demand for ML algorithms is skyrocketing, impacting everything from fun face-masking applications like Instagram or Snapchat filters to critical tools for diagnosing illnesses and recommending treatments. Opportunities abound in emotion and engagement recognition, enhanced homeland security features, and improved anomaly detection in industrial contexts.
While people and businesses are eager for ML/AI-based products, the algorithms require vast amounts of data to train on. This increasing need for diverse data makes synthetic data an essential solution.
From Grand Theft Auto to Google
Did you know self-driving cars have learned the rules of the road by playing games like Grand Theft Auto V to study virtual traffic? That was an early example of ML through synthetic data. Similarly, synthetic "scanned documents" have been used to train text recognition and data extraction models.
Banking and finance sectors already rely heavily on synthetic data for certain processes. Tech giants like Google and Facebook are also leveraging it, drawn by the extraordinary efficiency it offers to project managers and data scientists. The use of synthetic images and data points is expected to increase tenfold over the next year and by many hundred-fold in the coming years.
Overcoming Constraints of Real-World Data
Those at the forefront of ML are increasingly turning to synthetic data to bypass the numerous constraints of real-world data. For instance, a cloud-based generation platform can deliver millions of perfectly labeled and diverse images of artificial people, making data generation cheaper and more efficient. Real-world data generation can be prohibitively expensive, requiring an unimaginable amount of work to capture every conceivable angle, clothing combination, and lighting condition. Synthetic data can easily account for these endless variations.
Labeling data also becomes much easier with synthetic data. For example, pinpointing the source of light, its brightness, and its distance from an object in photos to train a shadow development algorithm would be nearly impossible with real-world data. However, synthetic data comes with these parameters by default.
Moreover, stringent regulations like GDPR make it complex and sometimes illegal for companies to share real-world data. In some cases, generating real-world data is not possible or safe. For instance, companies working on urban flying mobility use virtual worlds to train their autonomous flying cars, as there is no safe real-world environment for extensive testing.
Combating Bias in Data
Real-world data often suffers from historical biases, leading to issues like algorithms not recognizing certain demographic features properly. Even with an awareness of bias issues, it is challenging to create a real-world dataset entirely free of bias. Synthetic data can help address this by providing more balanced and diverse datasets. Additionally, data models need constant updates to avoid bias and degradation over time, creating a continuous need for fresh data.
Understanding the Opportunity
Synthetic data is still in its early stages and is not a one-size-fits-all solution. It faces technical challenges and lacks standardized tools and protocols. However, it is a powerful accelerator for ML/AI-based products as they expand into every industry and sector. We will see numerous new companies and deals in this area.
For those interested in diving deeper into synthetic data, Lean-IQ is set to become an Open Synthetic Data Community. This hub will offer synthetic datasets, papers, code, and insights from pioneers in the field, supporting the growing use of synthetic data in machine learning.
コメント