Part of the genius of the Internet is its ability to coordinate the actions of many disparate and geographically diverse people by eliminating the marginal cost of sharing information. In the last few decades alone, we’ve seen this happen again and again in areas such as education, science and commerce with remarkable results.
One question we have asked ourselves over the years is how best to apply the network model to the business of allocating capital. What other components or technologies would be necessary for a global scale community of people making investment decisions?
Numerai is a hedge fund managed by an anonymous community of data scientists. It encrypts its data and allows anyone in the world to continuously apply machine intelligence to the set and anonymously submit price predictions back. Numerai turns these predictions into trades and compensates the best performing models with bitcoin.
Today, we’re excited to announce that USV is leading the Series A round of financing in Numerai’s management company.
For an activity so heavily dependent on the efficient transmission and interpretation of information, the business of allocating capital has been slow to adopt the network model. Meanwhile, the pace at which the field of machine learning is advancing is rapidly accelerating. Between breakthroughs in our understanding of the science, platforms such as Kaggle, the Netflix Prize and a wealth of free online learning tools, there is an increasing supply of talent tackling all aspects of computing and data analysis. But these thousands of data scientists around the world with expert knowledge in machine learning are unable to apply that expertise to finance for lack of high quality data and trading capital.
Numerai attempts to fill this gap by acting as an interface between the machine intelligence community and global capital markets with an open-access, open-participation model. Anyone with an email and a bitcoin address can download the company’s data for free and train machine learning algorithms on it.
Every participant approaches the data set in their own unique way, producing many different solutions to the same problem. Numerai then combines each of these approaches into a single meta model, which dictates how to allocate the assets in its investment fund. In return, users are compensated in bitcoin in proportion to how much they help improve the meta model.
By encrypting its data set before releasing it to the public, Numerai turns the challenge of price prediction into a purely mathematical problem by removing the influence of human bias upon the results. Participants don’t know which securities they’re modeling nor what their predictions mean; only whether their model is performant or not. At the same time, Numerai itself does not know what algorithms the data scientists are using; their code and intellectual property remains theirs. Richard Craib, founder and CEO of Numerai, calls this a trustless relationship between Numerai and the data scientists, facilitated by encryption and anonymity.
Numerai is thus not a search for the best model; it is a platform to synthesize many different models, an invisible collaboration to build the meta model. At scale, Numerai’s fund is exposed to every model and a diversified portfolio without the risk of relying on a single and imperfect model.
This is a new kind type of capital allocation business that also has network effects – from the community, from their individual models and from the collective meta model. These network effects result in an open access fund that will generate more intelligence than a closed system built on a pre-internet organizational design.
In the year since its launch, this appears to be working: 7,500 data scientists have created over 500,000 models representing 28 billion predictions. This may be the largest ensemble of stock market machine learning models in the world.
Numerai describes this directly and succinctly: “The world doesn’t need another hedge fund, it needs exactly one hedge fund that’s powered by every artificial intelligence.”
More about Numerai can be found at https://numer.ai/about