Where the future of investment data is decided - There’s a lot of hype surrounding alternative data and AI but not a lot of information on how to utilize alternative data to generate useful insights. The objective of intelligence, human or machine, is to synthesize disparate pieces of information and understand something at a granular level for the purpose of making accurate predictions about the future. The rate at which information continues to accumulate exceeds the rate at which biological intelligence is evolving. As such, investors relying solely on traditional intelligence are struggling to understand the complex set of factors influencing a particular topic. As a result, investors are increasingly experiencing the feeling of being blindsided by idiosyncratic events despite believing that superior ‘homework’ was done.
AI can help investors cover blind-spots. AI can help humans make sense out of all of this “alternative noise”, make better decisions and generate alpha in the digital age of markets. Machines can help investors uncover hidden features and characteristics of complex subject matters by scaling semantic boundaries set by human experts. Like the famous quote attributed to Einstein, “Everybody is a genius. But if you judge a fish on its ability to climb a tree, it will live its whole life believing it is stupid.” To leverage AI successfully, humans must understand that there is no one size fits all AI. Like human intelligence, there are different types of AI with different levels of understanding, specialization and capabilities.
Accrete will be attending the NeuData Alternative Data Summit in London, bringing together the entire Alternative Data ecosystem of Vendors, Investors, FinTech and Financial Services firms. The two-day event is attended by both CEOs and Managing Directors representing the investment management community and belonging to the biggest names from the Hedge Fund and Asset Management world. Prashant Bhuyan, Accrete’s Founder and CEO, is attending the summit and shares some of his insights on key topics that will be discussed at the conference including the future of alternative data in investment decisions and the problem of the ‘black box’.
The answer depends on the way in which the asset manager is utilizing the alternative data. Good information is a critical component of making sound decisions in financial markets. However, too much information without true understanding exacerbates biases and causes underperformance. Novel data can provide insights beyond what humans can achieve independently, but an overload of data without actionable insights is of little use. Too many asset managers are drowning in mountains of alternative data and I don't see this trend stopping anytime soon. Rather, I believe that as machines get smarter, they will help parse topical nuance at a granular level in a way that furnishes users with predictive insights in a more efficient manner.
The most profound trend in algorithmic trading is the move away from manual order processing tasks to the automation of cognitive tasks. There has been a shift away from manual order processing tasks traditionally performed by human brokers toward automating cognitive tasks typically performed by highly paid front office analysts.
There are two important questions you must ask:
1) How much alpha content is there?
2) Can it produce stand alone strategies with high sharpe ratios?
The most important thing is to ensure alpha content actually exists in the data. Alternative data users can explore the data to find patterns but risk finding spurious correlations. Users must form hypotheses to test using the data and then determine whether or not the data is generated by cognitive automation or if it is traditional alternative data that is static in nature. Comparing alternative data with the labor cost of having humans generate similar data serves as a good benchmark for evaluation.
Alpha decay exists and it’s important for data providers to closely consider this factor for their clients. Providers may market the data generated by cognitive tools richest in alpha content to sell at a very high price to an exclusive group of partners. In this case, buyers take advantage of the alpha content before it decays.
Unless you have some way to look into the reasoning behind how AI outputs were generated, it is very difficult for humans to trust scores provided by AI. The issue is that most AI comes through a black box and providers won’t share with you the reasoning behind how the scores were generated. If you were to look under the hood, most AI is just hype because the machines are not truly understanding the subject matter and are unable to generate useful predictive insights. Unless you can validate the data using a combination of traditional intelligence and domain expertise, it will be difficult for you to trust the outputs of a cognitive system to make investment decisions. Accrete using the concept of a glass box where AI outputs are attributed to the source.
For more on generating alpha in the digital age, check out Nasdaq's interview with Accrete Founder and CEO, Prashant Bhuyan