On February 21, 2017, Numerai issued 1 million crypto-tokens to its community of anonymous data scientists as a method to incentivize the construction of an artificial intelligence hedge fund.
Motivated to revolutionize the concept of money itself, Numerai’s founder Richard Craib, a 29-year-old South African technologist, has essentially crowdsourced stock market prediction models built by machines, so that data scientists can offer sound predictions that, if correct, reward their efforts with bitcoin payments.
Built on the Ethereum blockchain and self-described as “the first decentralized smart contract powering a hedge fund,” Numerai released one million of its tokens called Numeraire (French for “coin” or “face value”) as a way to make collaboration possible between self-interested people. According to a Numerai Medium post published earlier today:
“Money was invented to solve the coincidence of wants problem and facilitate transactions. But as fiat currencies continue to lose relevance into the 21st century, cryptocurrency presents solutions far beyond money transfer. Cryptocurrency can now be used to incentivize cooperation in populations.”
Numerai takes the stance that efficiency in financial markets exists as a mere byproduct of market participants believing in the value of the money abstraction, who then seek to acquire more of it out of self-interest. The company proposes a new hedge fund model to bring synergistic network effects (or incentivized structures) to capital allocation, stating that “negative network effects are too pervasive in finance.”
The post goes on to claim that a common idea of the history of American economics is network effects, communicating their ideology in a simple statement:
“Every railroad made the railroad network more valuable, every telephone made the telephone network more valuable, and every Internet user made the Internet network more valuable.”
How It Works
Numerai already collaborates with a community of 12,000 data scientists who work together through the sharing of ideas, blog posts, tutorials, and code. Data scientists compete in sponsored science tournaments where they analyze data sets provided by Numerai, create machine-learning algorithms to find patterns in that data, and then test their proposed models by uploading their predictions to Numerai’s website. The predictions are then validated against historical data to determine how well the models performed. Numerai executes bitcoin payouts based on model performance, allowing data scientists from all over the world to earn funds anonymously while never actually losing currency.
While the company has a spirit of an open software project, they admit that the old design of their system was far from perfect, stating that Numerai had negative network effects like the financial market. Under the former model, bitcoin was used to facilitate the trade of dollars for machine intelligence. However, this transaction method essentially severed Numerai’s connection to data scientists due to the fact that bitcoin and US dollars have little to do with Numerai.
“A Numerai data scientist has no economic incentive to tell his data scientist friends about Numerai. He would only be bringing in competition and making it harder for himself to earn bitcoin. But if every data scientist could benefit from the overall network improving then collaboration would become rational and the game would shift to positive-sum.”
Recognizing this inefficiency in their system, this morning Numerai issued 1,000,000 Numeraire crypto-tokens to their existing 12,000 data scientists, distributed based on a scientist’s past performance in Numerai tournaments. Forgoing the token-sale model so common to other blockchain-based platforms, Numeraire tokens are essentially mined by data mining Numerai’s data - the submission of predictions is the proof of work.
More concerned with the live performance in their hedge fund than backtest performance, Numerai has created a staking mechanism that solves the biggest problem in quantitative finance, “overfitting” (a modeling error which occurs when a function is too closely fit to a limited set of data points).
Data scientists can now stake their Numeraire with their submitted predictions by sending Numeraire tokens to Numerai’s smart contract on the Ethereum blockchain. Once the predictions are analyzed, and if determined to be accurate, the data scientist who staked Numeraire on the prediction will earn bitcoin. If however, the predictions are poor, the staked Numeraire is permanently destroyed.
By staking Numeraire this way, scientists can express confidence in predictions, similar to how financial traders work, by deciding how much cryptocurrency to stake. Through Numerai’s proposed staking mechanism, participating data scientists stand to gain tokens backed by real world value, by building models that perform well in live data, and losing Numeraire on models that overfit the historical data patterns.
What’s In The Data?
According to Craib, Numerai is a long-short global equity strategy that utilizes various forms of data for scientists to analyze in a manner known as “supervised learning.” Different from reinforcement learning (such as with Google’s Deepmind program), supervised learning makes the most sense for the kind of datasets with which Numerai currently works. Craib told ETHNews “in the future, we’ll have more tournaments to assimilate all types of different Ai [artificial intelligence].”
Craib was not at liberty to discuss the details about Numerai’s data or its source. He did tell ETHNews that after encrypting the data, Numerai provides datasets to scientists that “preserve all the structure that you need to make great machine-learning models.”
Craib has been a fan of the Ethereum blockchain since he first read about the technology in 2015. Having contributed to the Ethereum crowdsale as well as that of Ethereum-based Augur (prediction markets), he was intrigued by smart contract functionality.
Joey Krug, founder of Augur, co-authored Numerai’s whitepaper, contributed to their smart contract code, is an advisor of the project, and also invested in Numerai’s Series A. Other investors include Union Square Ventures who contributed $6 mln in 2016, essentially eliminating the need for Numerai to hold a token sale. Because the company already possesses capital, its token model is unique in the sense that tokens represent bitcoin that data scientists use to stake their predictions and potentially earn real money.
Craib told ETHNews:
“We don’t need to do a crowdsale, because we already have the money to make this work. The tokens are decentralized because they’re on Ethereum, and the company is a real company that trades equities and has a hedge fund going for more than a year now. It’s not at all like an open source software project, but it has the spirit of that because anyone can contribute.”
Although Numerai has made their code public and all Numeraire tokens have been distributed to data scientists’ accounts, the project will not be going live until later this year. The Numerai team wants people to test and provide feedback as a way to eliminate any issues before the project launches on the Ethereum blockchain. For a copy of Numerai’s whitepaper or to see a film that provides more insight to the company’s vision, you can visit Numerai’s website.