Justin Wang received a bachelor’s degree from Murdoch University in 2012 with a grade of 83.7% and a master’s degree in Information Technology Management from the University of Sydney in 2016 with a grade of 82.5%. In January 2017, he founded a Singapore-based company with the mysteriously cool name Scry in order to “manage information technology to achieve business goals, as well as – and perhaps more importantly – how it can be wielded to disrupt existing value networks.”
What’s behind the mystery and jargon? It turns out that Scry is a “social forecasting platform.” Users join for free and can enter their personal estimates of the probabilities that certain events will happen, with Scry calculating the average probability. For example, one question is, “Will Apple launch a commercial self-driving electric vehicle before the end of 2024?” As I write this, there are 18 responses, entered up to six months ago. Eight answers are 50-50 and two are 100% yes. Overall, the average probability is 57%. This seems like harmless fun, inspired by the fickle wisdom of crowds. One glaring weakness is that users see the probabilities others have chosen (and the average value) before they enter their own probabilities. The wisdom-of-crowds argument assumes that each individual assessment is made independently. That assumption is sometimes untrue and is surely false here.
Soon, Wang decided to monetize these predictions by creating Scry Vantage:
We turn any questions you have about the future into measurable and accurate predictions, made through our proprietary AI.
No details are given beyond the price: “starting at $400/prediction.” I suspect that if a question is already in the Scry database, then the user probabilities are the basis of the $400+ answer, but I wasn’t about to spend $400+ to confirm this. In any case, the calculation of an average (or, perhaps, a weighted average) is hardly AI.
If a user pays $400+ and asks a new question, then the question might be forwarded to Scry users, although that would take an embarrassing long time. Perhaps Wang makes a semi-educated guess? Maybe he asks a handful of friends to make semi-educated guesses? Maybe he Googles the question and sees what turns up?
It is almost certainly the case that there is no true AI algorithm churning data for an answer, since computers don’t understand what words mean and have no way of processing what questions mean. (Ransacking Wikipedia to find the names of companies that make electric vehicles is qualitatively different from estimating the probability that a certain company will make a certain kind of vehicle by a certain date.)
The Scry website gives absolutely no information about their purported “proprietary AI.” The attitude seems to be, “We call it AI; that’s all you need to know.”
The word “AI” was selected by the Association of National Advertisers as the Marketing Word of the Year in 2017 and, too often, it seems that AI has become just a marketing ploy, like “.com” was during the dot-com bubble.
Scry is a small, but clear, example of the presumption by too many businesses that people will be impressed by anything that advertises itself as AI. Larger scale examples are OpenAI’s CLIP image-recognition algorithm which claims to use AI to understand words, but doesn’t, and OpenAI’s GPT-3 text generating algorithm which claims to use AI to write meaningful prose, but doesn’t.
I am reminded of Long-Term Capital Management, a hedge fund launched in 1994 by Solomon Brothers superstar John Meriwether. He put together a dream team that included several MIT PhDs who had worked for Meriwether at Solomon, Myron Scholes and Robert C. Merton (who would win Nobel prizes in 1997), and David Mullins (another MIT PhD, who left his position as vice-chairman of the Federal Reserve Board, where he was expected to succeed Alan Greenspan as Fed Chair). What could go wrong?
The minimum investment was $10 million and the only thing investors were told about Long-Term’s strategy was that the management fees would be 2 percent of assets plus 25 percent of profits. A legendary portfolio manager (“Dave”) told me that he had been pitched by Long-Term but chose not to invest because the only thing he knew for certain was that they were greedy. Other investors were not so cautious. Long-Term raised more than $1 billion.
After a few stellar years, Long-Term crashed in 1998. After the collapse, Meriwether promptly launched a new hedge fund, named JWM Partners. JWM crashed in 2009, and Meriwether started yet another hedge fund called JM Advisors Management.
Fool me once, shame on you. Fool me twice, shame on me. Fool me three times, you have no shame.
Warren Buffett has often warned, “Never invest in a business you don’t understand.” Dave heeded that advice when Long-Term came looking for his money. We should all heed that advice when a person or company says that they are using AI and doesn’t give a clear explanation of exactly what that means. Don’t trust an AI algorithm you don’t understand.