Advice to AI Startups: Lift Veil on Algorithms

The hardheaded folks who advise investors on what technologies show promise and which are mostly smoke and mirrors are beginning to weigh in on much-hyped approaches such as machine learning. While venture capital keeps flowing into machine learning startups, some technology consultants note that the investment craze around AI is slowing as deployments come up against market realities such as murky predictive models and emerging data protection regulations.

This observation by consulting firm SDR Ventures provides some insight into where AI stands today and where it might be headed: “…as long as a company spelled ‘machine learning’ correctly, it could expect to raise a round of capital around the concept of predictive analytics or artificial intelligence.”

Put another way, there has been a lot of “dumb money” flying around, much of it invested in AI technologies.

SDR notes in a new assessment that investments in AI startups slowed during the second quarter. Indeed, market tracker CBI Insights reported that AI investments declined 19 percent compared to the first quarter of this year.

“Investors may be trending toward more than a modicum of discernment,” SDR asserts, which sounds a bit like an Alan Greenspan-esque way of saying the “dumb money” has gone elsewhere.

The technology consultant further notes that one area where machine-learning startups are showing results is creating a measure transparency about how AI algorithms actually work and what they can accomplish. The ability to understand the inner workings of modeling techniques is seen as plus, especially for decision makers who previously had to trust analytics models they didn’t understand.

“Very few management teams or boards have been willing to accept the ‘because the model told me so’ defense,” SDR notes.

Another potential drag on machine learning investments and deployments are new data protection regulations such as the European Union’s General Data Protection Regulation scheduled to take effect in May 2018. GDPR establishes stringent new rules on data handling along with transparent usage policies and consumer-friendly privacy policies within the EU.

Tighter data regulations “will restrict automated individual decision-making,” the tech consultant notes. That prospect has encouraged a growing list of information management vendors to offer frameworks for meeting the EU’s compliance deadline. Among them is Veritas Technologies, which commissioned a study this spring that found 47 percent of enterprises survey doubt they will meet the compliance deadline.

All this is not to say that some machine learning startups are making headway in the market and continue to attract bullish investors. Those attracting the most venture capital are directly addressing “transparency hurdles” within machine learning models so users can figure out how they work.

“For companies specializing in predictive analytics to gain wide acceptance and achieve scale within a given sector, and especially around specific data-based decisions, they will need to demonstrate the ability to easily let users ‘under the hood’ in order to let decision makers get comfortable with their algorithms’ explanatory power,” the consulting firm concluded.