your123 - stock.adobe.com
With government agencies in Europe and the U.S. cracking down on the irresponsible use of AI, more organizations are becoming conscious about how they use AI technology and more careful about eliminating bias from AI algorithms and models.
One startup working with responsible AI is Mostly AI. Founded in 2017 and based in Vienna and New York, the vendor specializes in AI-generated synthetic data.
Mostly AI said on Jan. 11 that it raised $25 million in a series B round of funding led by Molten Ventures, with participation from Citi Ventures and Earlybird. The vendor also recently released the latest version of its synthetic data platform, which automatically synthesizes data.
Synthetic data is artificial data that is based on real-world events. It helps organizations and companies skirt problems that data scientists face when working with real data.
"In the past, we were saying that synthetic data is as good as real data. Now we're saying that synthetic data is actually better than real data," said Mostly AI CEO Tobias Hann. "Synthetic data can be shaped in whatever form you need it to make it even more meaningful for your purposes."
Synthetic data plays an important role in explainable AI, fairness and bias. Data scientists can create synthetic data that is modified, debiased and fair, thereby correcting the biases that real data can create, whether gender, racial or other bias, Hann said.
Interest in responsible AI
Mostly AI said it will use the funding to build on its vision of a smarter and fairer future, centered on responsible AI.
One of the best practices that organizations can use to achieve responsible AI work is synthetic data, said Sumit Agarwal, an analyst at Gartner. Gartner expects that more than half of data used for AI projects will be synthetically generated by 2024.
Sumit AgarwalAnalyst, Gartner
Interest is growing beyond the finance and healthcare markets, where regulation is forcing enterprises to build fairer models, Agarwal said. Many organizations are looking to add room in their budgets to build responsible models.
"It just helps to make the process better or make their models better," Agarwal noted.
However, challenges exist in implementing responsible AI. Chief among them is that the technology is still new.
"There're certain [integrations] where you can say there is more transparency, but there are others [that are still] an active area of research," Agarwal said.
Data and responsible AI
Data can also be a hurdle when creating a responsible AI model. Organizations must have good quality data to build models that are fair.
"It's not an easy problem to solve," Agarwal continued.
"The secret sauce is in the data," he said. "If you don't have enough data ... you can generate synthetic data to fill in that gap and balance it."
While relatively few organizations are using synthetic data now, more are looking at the technology. "This is an area where the interest is only going to increase, so startups have a good opportunity, but it has to be integrated with the overall process," Agarwal added.
Other than building out responsible AI applications, Mostly AI also plans to use the funding to grow in Europe and expand in the U.S. The vendor said it's already working with Fortune 100 banks and insurers in North America and Europe.
"We see a lot of interest and a lot of potential in the U.S. market," Hann said.