Following the launch of a competition to test AI-driven predictions about investment and markets, we speak to an organisation seeking to promote the benefits of an "ensemble" approach to data predictions.
When it comes to using AI to guide investment, it pays to harness
the “wisdom of crowds” rather than hope that a single engine
delivers the goods, so a figure in the tech world
This publication recently spoke to Arnaud Castillo, founder and chief executive of CrunchDAO, a research team of data scientists using the power of “collective intelligence,” competition, and Web 3.0 tech to produce financial insights. The firm has launched the ADIA Lab Market Prediction Competition, a machine learning contest.
(ADIA Lab is an independent research lab supported by the
Abu Dhabi Investment Authority, the sovereign wealth fund.)
With wealth managers wondering whether AI can solve some of their investment and risk management challenges, there’s real ferment in the tech space to come up with tools. Tests, competitions – even injecting an element of fun into the process – appear to be a smart way to approach harnessing different ideas from thousands of people.
The ADIA Lab competition, launched in early May, has a total
prize pool of $100,000. Running competitions as a way of
unlocking ideas is a passion for Castillo, who is based in
“If you want to diversify risks you need to find assets that aren’t correlated. Machine learning is good at that because it can tell you what the correlation is. This [insight] is already used a lot in the Modern Portfolio Theory,” he told WealthBriefing.
The benefits of the solutions [via AI] improve when a range of “answers” from AI are drawn together to create an “ensemble,” enabling the real signals to be drawn from market noise, Castillo said.
(The use of the word “ensemble” is significant. As explained in one definition, “in statistics and machine learning, ensemble methods use multiple learning algorithms to obtain better predictive performance than could be obtained from any of the constituent learning algorithms alone.”)
“This is about bringing together thousands of predictions,” Castillo continued.
Castillo has worked as an M&A Analyst at Lehman Brothers on Wall Street and as mergers and acquisition vice president at Deutsche Bank.
Wall Street experience
During his time on Wall Street, Castillo came to realise that AI in its current form did not possess the skills required to accurately predict stock and bond prices. What he found was that by trading their "predictions" algorithms, AI instantly corrects the anomalies they have detected, thus changing the initial parameters of the prediction and rendering it obsolete.
Castillo found that financial markets' quantitative research has been limited for many years due to using too few parameters in their mathematical models, where successful tools like ChatGPT are trained on more than 175 billion parameters. It was then that he decided to take it upon himself to solve this problem, leading him to co-found CrunchDAO. Castillo thinks it is better to design models that take into account the complex and changing phenomena of the market, rather than using convenient but inappropriate mathematical models for financial data.
“AI is obviously a big trend but it’s not new. Algorithms have been around for a while,” Castillo said, citing the creation of the MSCI Barra Risk Model some years ago.
Work has been done to look at non-linear factors and release the value of machine learning.
“The wealth management market is a `winner-takes-all’…we have to compete to get the most out of the market,” he said.
The CrunchDAO team of researchers can make predictions for different clients. “This is bringing an additional value of quality and predictions that you can get.”
Castillo commented on how the term “artificial intelligence” can
be misused and misunderstood: “I have always thought AI is
not the right word..it is not artificial and it is not
intelligent. It is real as it is fed with historical data and
learning to it to replicate signals it finds. There is no
creativity or imagination – it is about replicating history.
People are realising that ChatGPT is extremely powerful and in
some ways more powerful than Google.”
Explaining AI to the client is a big task.
AI can also be used in areas such as behavioural finance, he said. AI acts without sentiment and is free of biases.
(See views by this news service about AI and what it brings, or does not, to wealth management.)