"Computers are better than humans at things humans were never very good at," is how Matt Calkins, the chief executive of cloud software provider Appian, sums up artificial intelligence and machine learning.
The remark drew appreciative laughter from a small group of journalists gathered to chat with Calkins over dinner in Manhattan a week ago. Calkins bears a vague resemblance to actor Hugo Weaving, who played "Agent Smith" in the Matrix trilogy. Coupled with a dry wit and a certain measured cadence, he has a way of delivering insightful and amusing quips that sneak up on you.
Calkins's point at that moment in the dinner was not an idle, remark, however. He was in earnest about the fact that there are very important differences between where computers can apply themselves, on the one hand, with artificial intelligence and machine learning, and where humans continue to excel and will for the foreseeable future. That has implications both for A.I. broadly speaking, and for software development, the latter of which is Calkins's focus as the leader of a software tools developer.
The proximate event for the dinner was the release this week of a survey of IT leaders, which was prepared by the company in conjunction with IDG. The survey showed that machine learning is "the most commonly deployed form of intelligent automation" in companies, with 54% of respondents using it in some form.
That is an interesting finding, given that Reston, Virginia-based Appian's focus is on how humans can use technology directly, to build increasingly complex systems with relatively little effort.
Appian, which went public a year ago in May, offers a cloud-based software development environment that emphasizes the "low-code" notion: by manipulating various modules, those with little to no programming experience can assemble enterprise applications. The technology is hosted on Amazon's AWS.
Calkins has observed human thought from some interesting vantage points. On the one hand, there is Calkins's lifetime fascination with board games. He has developed three board games of his own, the latest one being "Tin Goose," which is based on the early days of commercial aviation. A prior game involved feudal Japan and samurai. He tends to delve deeply into areas of particular cultural or historical subject matter, and to learn about the structure that can be extracted as a game. Calkins regularly attends the annual World Boardgaming Championship, and ranks well at those events.
Board games, he notes, are "higher resolution" than online games, one reason he has an affinity for them. And, they involve meeting up in "real space," which is part of why he enjoys going to the annual Championship.
Computers, says Calkins, are good at things where there is little to no hidden information, such as chess, or the strategy game Go -- venues, of course, where machine learning has had some stunning successes.
"Computers are not as good [as humans] at games where there are elements of intuition, like charades," he observes, or many types of board games where the roll of the dice dominates.
For similar reasons, Calkins believes AI will be limited to things at "the bottom of the stack," code decisions from which human beings will move away over time. With many years in the field -- he founded Appian in 1999 at age 26, originally as a consulting operation, and has lead the firm through numerous shifts in the enterprise landscape -- he concludes that, "We will still need humans to come up with solutions at higher levels of the stack, at high levels of abstraction."
"The coding landscape involves constant change," Calkins reflects. "Computers are not as good with constant change. Like, for example, computers can be good at [discerning] pictures of a cat, because a cat is a cat is a cat, or at natural natural language processing, because English doesn't change that much (irony, yes, but...)"
The view of the human role dovetails with how Calkins views his place in the market for software development. There is a "philosophical" divide in the low-code application market, he says. Appian is in what he calls the "enterprise low-code" portion, where the philosophy is to discourage programming per se, and to promote maintenance of everything at a high level of abstraction.
In contrast, what he labels the "assembly" market for low-code -- having nothing to do with assembly language -- encourages more low-level coding, and includes companies such as salesforce.com, Boston's Outsystems, and Mendix, which this week was acquired by Siemens AG for $730 million.
Appian's message is resonating, to judge by sales growth, which was almost 40% in the three months ended in June. That's off a still small base, as the annualized run rate (trailing twelve months' revenue) is just under $220 million. The company is not yet profitable. Asked about someday getting to positive cash flow, Calkins offers that he "wouldn't be surprised to see us be in a couple years' time."
"We can be cash-flow-positive at any time we want, but we acknowledge that there are opportunities that come up" for which the company needs to invest.
The stock's had a rough few months, falling from around $43 per share in mid-June to a recent $29.01, below the $35.15 offer price at which the company did a follow-on offering of stock in late August. That fall comes despite the fact the June-quarter results topped Wall Street's expectations when they were announced on August 3rd, and despite the company raising its year outlook.
Still, the stock has roughly doubled since the May, 2017 IPO. Investors, notes Calkins, care most about subscription revenue growth, which was 36% in the latest quarter. They also care about the company's "net retention rate," a measure of how much customers stick with the platform. That was "a very healthy 119%" in the most recent quarter, he says.
As far as being a hot take out, it seems unlikely. "No sane buyer would make an offer we could accept," he quips. Appian has a market capitalization of roughly $1.8 billion.
After running the company for 19 years, moreover, Calkins shows little inclination to walk away. "We really see a lot more room to go in this phenomenon" of low-code.
But then, "we weren't in this for the money," he says in deadpan fashion, then, after a brief pause, delivers the punchline, "I was in L.A. recently, and I told people, We were in it for the fame!"
- There is no one role for AI or data science: this is a team effort
- Startup Kindred brings sliver of hope for AI in robotics
- AI: The view from the Chief Data Science Office
- Salesforce intros Einstein Voice, an AI voice assistant for enterprises
- It's not the jobs AI is destroying that bother me, it's the ones that are growing
- How Facebook scales AI