Synthetic data will help build better AI and the metaverse

Synthetic data – the generation of artificial images to train AI and computer vision – will be essential in building a future metaverse.

Why is this important: AI has long been trained on images – including human faces – captured in the real world, but it can create serious privacy concerns.

  • Using synthetic data instead can help work around this problem, although it raises new concerns about accuracy and authenticity.

Driving the news: Facebook announced on Tuesday that it plans to shut down its ten-year-old facial recognition system and remove facial scans from more than a billion users, for privacy concerns.

Between the lines: Increasingly, privacy concerns will lead companies to shift from capturing real faces and other images to training AI when transitioning to using synthetically generated data.

  • Tel Aviv-based synthetic data company Datagen performs high-quality digital analysis and motion capture of real people and objects, then uses AI to generate realistic but not real versions.
  • Gartner recently predicted that by 2024, 60% of the data used for the development of AI and analytics projects will be synthetically generated.

The big picture: Since images of real people are not used directly, privacy and bias are less of a concern.

  • Early computer vision systems were often trained on data sets sourced from the Internet that were disproportionately white and masculine, which meant they were less accurate at recognizing the faces of other races and genders.
  • With synthetic data, “you can incorporate actual distributions from the real world, so there is no bias between age, gender and more,” says Gil Elbaz, co-founder and CTO of Datagen.

The trap : Some experts worry that the synthetic data may not be as valid as the real thing, which could damage the performance of AI models trained on it.

  • Many of the same tools used to generate synthetic faces for AI training could also be used to create compelling deepfakes, although Elbaz notes that technical tools such as smart contracts could be used to separate synthetics from. false.

And after: Synthetic data will be key to creating a more realistic version of future AR and VR called the Metaverse.

  • “The metaverse is going to have a hardware and software component,” Elbaz explains. “Synthetic data will be part of the software that enables the right kind of hardware.”

Leave a Comment