There are many in the AI ââworld who are concerned about its implications. One of those people is Navrina Singh, former product manager for Qualcomm and then Microsoft, who saw firsthand at Microsoft how a Twitter bot developed in 2016 as a “conversational understanding” experiment was, to quote. The Verge, “Taught To Be A Racist Asshole In Less Than A Day.”
The Microsoft crash is just one of many examples of AI gone awry. In 2019, an algorithm sold by health services company Optum to predict which patients will receive additional medical care was discovered by researchers as a poor assessment of the health needs of the sickest black patients. AI credit scoring systems have repeatedly proven to be sexist.
While many large companies have assembled teams to tackle the ethical issues arising from the massive amounts of data they collect and then use to train their machine learning models, progress on this front has not been without clashes. In the meantime, small AI-powered businesses that can’t afford teams of individuals are largely taking flight.
Enter Singh’s company, Credo AI, a SaaS company that is today winning $ 5.5 million in funding that it raised from Decibel, Village Global and AI Fund.
The company’s promise is pretty straightforward as she explains, even though she manages the complex. Singh and his current team of 15 employees have developed a risk framework that gives companies a window into their own governance. It’s not so much that the startup’s technology is revolutionary (according to our understanding) but rather that Credo AI resolves what is often a lack of accountability within organizations by giving them a control panel with the tools to manage all of it. kinds of data they collect, like as well as suggest controls that they might not be using, like IEEE standards that they can integrate to provide stronger safeguards for their machine learning models.
“What a lot of companies haven’t really understood is that there is a lack of common language and alignment on what ‘good’ AI governance looks like, so organizations are looking for really help with this standardization, âshe said.
Credo AI software is not a one-size-fits-all offering, Singh notes. Different organizations see different impacts from their models, and even within what are called verticals, individual companies often have different goals. âFairness is not defined in different sectors,â says Singh, who gives the example of financial services, where much of it is constantly being redefined by federal banking agencies. âWhat does fairness in fraud mean? What is equity for underwriting credit? “
Rather than waiting for answers, says Singh, Credo AI works with companies to align with their own values, then gives them the tools to manage accordingly, including adding additional metrics when they want, and different. stakeholders. âWe want to allow your data science team to collaborate with your compliance team, your executive to collaborate with your person who wins machines, your product manager to collaborate with your risk manager,â says Singh.
Credo AI wants to help businesses avoid these tough times – or worse.
Certainly, this is a great market opportunity. According to data released earlier this year by the International Data Corporation (IDC), global AI market revenue, including software, hardware and services, is expected to grow 16.4% year on year on the other in 2021, to reach 327.5 billion dollars. By 2024, IDC said, the market is expected to cross the $ 500 billion mark.
As businesses spend more on AI, they’ll likely need more help to make sure it works the way they expect it to and doesn’t cause damage. Indeed, if Singh is successful, working with Credo AI will someday serve as a sort of cachet that companies will use to publicize their interest in ethical AI.
âIf we do our job well,â Singh says, âI want anyone who builds good AI to be associated with Credo AI. It is certainly our aspiration.