Will You Trust Deepfake Technology to Make Real Business Decisions?

A new artificial intelligence (AI) and machine learning (ML) startup is betting you will. Founded by former Google, AWS and Uber execs, RealityEngines.ai emerged from stealth with an interesting offering—use the technology that gave us doctored videos and images of people doing and saying things they’ve never done (think revenge porn and fake news) and using it for good...business. 

The technology that underlies Deepfakes is Generative Adversarial Networks (GANs). Ominous as it sounds, GANs simply describes neural networks that are pitted against one another in a game of “Real or Fake” where one network (the generator) tries to generate images, for example, that are so realistic that the other (the discriminator), though trained using the real data set, can’t discern the difference. An algorithm is applied to both networks to improve the accuracy of its generative/discriminative function until the newly generated data set is virtually indistinguishable from the real data. 

RealityEngines uses the resulting synthetic dataset to create a sort of amplification or augmentation of a company’s real or original data set into the much larger ones required for AI to deliver accurate predictions, identifications and diagnoses. According to the company’s CEO, Bindu Reddy, their approach creates models that are up to 15% more accurate than models trained without the synthetic data (small and large, noisy data sets)—even without domain-specific expertise. Their results jibe with those found in other studies involving use of synthetic datasets where data was limited due to privacy concerns or because generating real data would be too expensive and time consuming.

RealityEngines first set of products address standard enterprise use cases for ML, including lead scoring, churn predictions, threat and fraud detection, personalized recommendations, and cloud spend optimization in hopes of “making machine learning easier for enterprises...and giving businesses an easy entry into machine learning, even if they don’t have staff data scientists.”

The company further claims that they’ve landed upon a more effective, efficient way to reduce bias via a measure-agnostic optimization approach.

Will you trust Deepfake technology to make decisions for real customers, patients, employees?

See how real Deepfakes can be with this open source GAN, StyleGAN.With this, you’ll never need to hire another model for that ad...maybe?