The artificial intelligence landscape is full of players, and not all of them are legitimate. Some are practicing something called “AI washing,” as Gary Gensler, chairman of the Securities and Exchange Commission, explained. In a video This includes “false claims to investors by those claiming to use these new technologies.”
In the financial space, the SEC charged two investment advisers, Delphia (USA) Inc. and Global Predictions Inc., fined a combined $400,000 in March for what the SEC described as marketing AI-enabled investment predictions to its clients when they actually weren't. using this technology. (The companies paid the fines and accepted the SEC's orders without admitting or denying the commission's findings.)
It's not just the financial services space that's seeing this trend. Analytics firm FactSet dug through S&P 500 companies' earnings call transcripts for the three months ending in mid-March and found that 179 companies used the term “AI,” higher than the five-year average of 73 companies. While it's unclear which of these companies are serious about the depth of AI technology in their operations, experts say that using the acronym has become common across industries.
“Putting AI into a slide deck, or even just using AI, is not that difficult because you can use it very easily through any platform without building it into a business,” said managing partner Michael Stewart. ” At M12, Microsoft's venture arm, which focuses on AI, gaming, and deep tech. “Even so, it has no sustainable competitive advantage,” he added.
In reality, what AI washing leads to is a loss of trust between vendors and their customers, enterprise partners and investors.
“The parties at risk are really the people behind who don't understand the technology,” said Timothy Bates, a professor of practice at the University of Michigan-Flint's College of Innovation and Technology who specializes in AI and other emerging technologies. Technology Officer at Lenovo and General Motors).
It's not just companies that claim to use AI when they have none. Bates says so-called button-pushing applications are also AI washing. AI learns by digesting information and receiving various inputs, or cues. “When you ask the same question over and over again, which is a button push, it doesn't build a database,” Bates said, adding that third-party applications built on common natural language processing models are long-term. are not effective for .
For example, when a law firm buys an AI assistant to replace a human, it could stop working within months if it isn't based on a unique, unbiased database that is specific to law. Be trained and actively learning.
A 'good' corporate AI litmus test
Tobey Coulthard, chief product officer at Frisee, a creative AI solution for enterprise clients (including Sephora and Macy's), says to be wary of any business that uses the term AI too broadly.
“AI is such a vague term,” Coulthard said. Instead, he says businesses should define what kind of AI they use and how they use it. have been.
Additionally, Coulthard says it's important to note when a company starts talking about its use of AI. “It's a good litmus test to see if a business was talking about AI before ChatGPT,” he said. And it's also a positive sign, he says, if they're talking about what they won't do with AI and using some kind of ethics policy.
“The more you look at a business talking about what they do with AI and the more verbose they are about what they don't do with AI, that's going to be a much more accurate representation,” Coulthard said. “
Look at the model the company uses for AI offerings, Bates says. “It's really digging into the company itself and finding out, have they built their own models? [or] Are they completely dependent on a third-party model?” And if they are dependent on a third-party model, uncover their service level agreement or key performance indicators over the next year or two, he said. These signals that are being sold as AI companies have to maintain and monitor and adjust to work.”
Microsoft's VC Approach to Investigating AI
Stewart and other AI-focused contributors at M12 evaluate startups using the four D's: Data, Dividend, Distribution, and Delight.
“If you, as a startup, don't have access to your most important customer data related to what AI will work on, neither will any other competitor,” Stewart said. That same data can be accessed,” Stewart said.
For the profitability part, it can be helpful to identify whether the output of the AI is itself a work product. In other words, is it contributing to the bottom line? When possible, the parties involved can look at the overall profitability of any type of AI business, knowing that one of the advantages of the technology is the extreme profit margin. Stewart says 80-90% gross profit is also standard for a fully AI company with limited human intervention.
Although the segmentation and happiness elements of M12's investment analysis process are not directly related to AI washing, these factors play a key role in determining a startup's sustainable longevity (understanding that young businesses develop as adults).
With thousands of AI startups in the market, winners “need strong and stable distribution channels to cut through the noise and ensure success,” says M12. Meanwhile, “creating consistently delightful user experiences is the key to building the love that keeps them coming back for more.”
AI washing is another form of jumping on the technological bandwagon. Unlike earlier AI evolutions, natural language processing technology is marketable and marketable in unique ways. Because of this, investors and consumers have a certain stake in the technology, which makes ambiguity more likely.
Until now, the SEC has focused primarily on investment advisers and broker-dealers, but there is widespread concern in the AI-washing marketplace. “The last thing we want to do with our fund, and that includes Microsoft, is use a technology that's off the guardrails,” Stewart said. “If it's a hoax, if the AI isn't really what's creating the magic, and we've missed it, then that's something we have to go back and redo our due diligence process. need.”
Already, M12 once erased the chalkboard in 2022 and created all new guidelines for analyzing AI startups after it was discovered that many computer visions were more superficial than the companies claimed.
“The hype around AI can attract investors, drive up stock prices, and attract consumer interest, giving these companies a temporary edge,” Bates said. is,” Bates said. He added that regulatory bodies are likely to step up scrutiny and enforcement, as too many resources go to superficial AI claims and not enough to make concrete progress in the field.