COVID-19 Gives AI a Reality Check
While it seems unlikely that AI will enter another nuclear winter, the current COVID-19 situation is giving enterprises the opportunity to rethink their AI strategies, giving the better AI projects more room to run, while discarding the borderline AI projects that were unlikely to pay off.
The macro economic situation deteriorated rapidly thanks to COVID-19. In the span of a few weeks in late March, the United States went from record-low unemployment and widespread prosperity to massive layoffs and the abrupt end of the longest economic expansion in history. Amid this brutal economic toll, many companies have declared bankruptcy, while others are slashing their budgets in an attempt to weather viral the storm.
Will AI projects survive the mess? Yes, but with caveats.
Companies that have already implemented AI have reported fewer impacts and a greater ability to respond to unanticipated disruptions to their businesses. We’ve seen this in the consumer goods supply chain, where machine learning technology has helped manufacturers, distributors, and retailers adapt to radical changes in demand for products, like toilet paper and hand sanitizer. We’ve also seen it in IT management and telecommunications, where AI-based monitoring tools have stayed on top of abrupt shifts in usage patterns for Internet access amid government orders for workers and students to shelter at home.
Despite these examples, AI remains an aspiration for most organizations, not yet a concrete reality. For years leading up to the COVID-19 pandemic, AI was wildly popular among business leaders, who saw machine learning and advanced analytics techniques as the tickets for becoming data-driven organizations. As we have documented in this newsletter, AI and related technologies have been flourishing, even if the actual results from AI projects lag expectations.
The COVID-19 crises and recession are forcing business leaders to take stock of their current situation and make dramatic changes to business plans. All sorts of radical changes are on the table: Will workers return to company buildings? Is on-prem computing dead? And should companies should scrap investments in AI and advanced analytics initiatives, or double-down on them?
According to Nick Elprin, the CEO and co-founder of the data science platform developer Domino Data Lab, companies are continuing their AI investments.
“What I’m seeing is, the trend we’ve seen for a while around more and more enterprises investing and scaling in data science is accelerating,” he says. “Enterprise across industries are doubling down on making data science a core capability and trying to get more and more value out of it and putting models at the heart of their business.”
Even in light of COVID-19 and macro uncertainty, companies are pushing forward with AI, Elprin says. “Actually what we’re seeing is the best research organizations, the most advanced enterprises, are leaning even more heavily on data science to drive competitive advantage and increased performance in light of the macro economic situation,” he says.
Arijit Sengupta, the founder and CEO of the AI startup Aible, says companies will move strongly toward AI projects that create value quickly, as opposed to funding theoretical projects for long-term innovation.
“It will force people to put up or shut up,” Sengupta says. “There’s a lot of vaporware and a lot of marketing in this space. You need to show me actual value delivered, something tangible.”
Amid COVID-19 budget cuts, companies no longer have the patience for three-month AI projects, he says. If an AI software vendor or service provider can’t show a return on investment in three weeks or so, the customer should keep moving.
“Reality just bit,” Sengupta tells Datanami. “The first generation of AI, they’re all coming out of labs. I was lucky in that I studied AI from people like Ed Feigenbaum. I was at Stanford, so I caught the last generation of people who actually failed with AI, if you will, who experienced the nuclear winter of AI.”
Sengupta was the founder of BeyondCore, which he sold to Salesforce for $110 million in 2016, and which became a core component of Salesforce Einstein. So he can afford to build a nice bunker to withstand a nuclear winter. But he’s convinced that AI can flourish if practitioners turn to more pragmatic use cases, which is a model he’s taking with Aible, which helps companies use AI to solve real-world problems.
“One of the things that I learned through those experiences [at university labs] is they made the same mistakes that a the current generation of AI has been making, which is they’re making cool models, cool technologies out of labs, out of Kaggle contests, divorced from business reality,” Sengupta says. “And that was fine when it was an interesting curiosity and I have a lot of money to spend on curiosity. It doesn’t work in the current situation, where reality has actually bitten.”
During recessions, companies often become defensive in their investments. That’s likely to be the case with AI tech, which has the advantage of being applicable for both defensive (i.e. shoring up existing business, such as reducing churn or improving user satisfaction) as well as for offensive purposes (i.e. expanding business, such as finding new customers or increasing profits).
Companies will likely demand a greater return on investment (ROI) for AI and big data projects during the COVID-19 recession, according to a June 16 article in the MIT Sloan Management Review titled “The Recession’s Impact on Analytics and Data Science.”
The challenge is that “ROI is a tough standard for data science, in part because many algorithms never get deployed into production applications,” authors Jeffrey Camm, Melissa Bowers, and Thomas Davenport write. They note the notoriously high failure rate of big data projects, including Gartner analyst Nick Heudecker’s 2017 estimate that 85% of big data projects fail.
The hype around big data has morphed into hype around AI, but the data suggest that AI hasn’t escaped the high failure rates of big data. Despite the perception that AI and machine learning usage is widespread, the reality is that, outside of the biggest enterprises and tech giants, most companies are still struggling to get positive results from the technology, at least in a timely manner.
Databricks CEO Ali Ghodsi referred to this as “AI’s 1% problem.” “There are only about five companies who are truly conducting AI today,” Ghodsi told Datanami back in 2017. “All the rage is about AI and the predictions it can do, but they’re not talking about the 99% versus the 1% problem that they have.”
But it’s not all doom and gloom. Nobody doubts that AI and data science has the potential to dramatically impact an organization’s success. We see those results every day. But we also need constant reminders that there is no “easy button” for AI, and that it takes commitment and work from a skilled and motivated group of people.
Data science teams that have proven their ability to get positive results out of big data and AI will likely get more work during the COVID-19 recession. Some of these will be defensive in nature, but some companies will likely use the economic disruption as an opportunity to solidify their market positions and thus will be offensive in nature.
On the other hand, data science teams that have not shown good results with their AI projects will likely see their work cut, or even possibly shown the door as organizations struggle to keep the lights on amid a recession with an uncertain end (we’re already hearing anecdotal reports that data science and AI vendors are snaping up these discarded data scientists and AI engineers). The support that data science teams have with the C suite will also be tested during these unprecedented times.
In summary, the COVID-19 pandemic and recession do not mark the end of AI. But it will likely help to weed out the poorly conceived AI projects and technologies, while simultaneously rewarding the AI projects and teams that can achieve business success and show a positive ROI. It might be uncomfortable for some, but it will likely leave AI and the companies stronger in the long run.