- Investors include Lupa Systems, Betaworks Ventures and TitletownTech
- Funding will help Synthetaic grow team, and scale-up modeling tech
- It uses generative adversarial networks to train AI models
- AI can detect poachers with weapons, endangered animals in the wild
Synthetaic was founded by Corey Jaskolski to tackle data creation using AI. It focuses on training artificial intelligence to generate images. The company concluded a $4.5M funding, where $1M is funded in pre-seed, and $3.5M is received in the seed round. Investors included Lupa Systems, Betaworks Ventures, and TitletownTech (based in Wisconsin). TitletownTech funding is a joint venture of Microsoft and Green Bay Packer.
Synthetaic combines high-fidelity 3D modeling and novel, generative AI to grow large and high-quality datasets. These data sets are used for machine learning. The AI-enabled industries are now able to overcome data constraints that caused hindrance in the application of artificial intelligence. Synthetaic focuses on synthetic image data that is more difficult to produce than columnar data. It uses Synthetic datasets to resolve the high stakes use cases.
The company will benefit from the seed funding, says Synthetaic’s CEO, Corey. They will use the funding to grow their team, scale the modeling, and generative AI technologies. It will also provide support to their technology strategic partners.
What led to Synthetaic?
Corey has been awarded the Explorer of the year award by National Geographic. While working on his project for NatGeo, he got conscious of the need for data sets in conservation. The project aimed to identify poachers and endangered animals from camera footage automatically. But, the major obstruction was the unavailability of images of either the poachers or the endangered animals in the wild. This made it challenging for them to train AI for detection.
Synthetaic uses generative adversarial networks combining it with the work of 3D artists and modelers. This makes it much more scalable and affordable as compared to its competitors that employ 3D worldbuilding. In 3D worldbuilding, a replica of the world you want the AI to learn is generated. This synthetic creation of AI training data is expensive.
Corey illustrates Synthetaic’s model with the example of identifying poachers. The 3D team created a photorealistic model of weapons and applied adversarial networks on them. This generated an abundance of images depicting that model against several backgrounds.
Synthetaic then tested that AI simulation on real data and validated the results. Corey adds that customers do not require to possess any AI expertise, as the company provides an ‘end-to-end’ solution.
