Deus ex Machina ~ The implications of generative AI for hedge funds
Imagine a world where artificial intelligence could generate intricate financial models, predict market trends, and even develop groundbreaking investment strategies - all with the click of a button. Brace yourselves, because that future is now, thanks to the astonishing power of generative AI.
Generative AI, a cutting-edge technology harnessing the prowess of artificial intelligence, is set to redefine the landscape of finance as we know it. Gone are the days of tedious number-crunching and gut-driven decision-making.
The innovation of gen AI, underpinned by large language models, will unleash a tsunami of possibilities, enabling intelligent algorithms to autonomously create complex financial models, simulate market scenarios, and make predictions that would leave even the most seasoned Wall Street expert awe-struck.
Picture this: investment portfolios optimised with unrivalled precision, risk assessments made in a fraction of the time, and trading strategies that adapt and evolve with lightning speed. Generative AI has the potential to revolutionise every aspect of the financial industry, from algorithmic trading to risk management, wealth management to fraud detection.
As generative AI technology evolves, it could democratise access to financial services, providing individuals and businesses with unprecedented insights and opportunities. As Microsoft, Google, Amazon, Baidu and Nvidia battle it out to develop the next best tool, the world is entering an exponential growth phase helped by reinforced learning with human feedback (RLHF).
Who knows what level of nuance and sophistication the sixth, seventh, or 10th iteration of ChatGPT might display, as AI systems learn how to interpret irony, sarcasm, and fake news from factual statements as they scour the web?
That they could begin to develop original content – a science fiction novel or perhaps a new piece of algorithmic code - is a tantalising prospect; bringing AI one step closer to achieving human-level intelligence, as it becomes more adept at processing and interpreting data in large language models.
Hyperscalers are well positioned
With this in mind, what could the implications be to the hedge fund industry? Citadel’s Ken Griffin has already begun to explore options for obtaining an enterprise-wide CHATGPT license to help Citadel’s employees become more productive.
From an investment perspective, the cost of developing LLM infrastructure is going to be significant. According to a new Jefferies equity research paper on AI, this will benefit hyperscalers – Google, Microsoft – and data management vendors and platforms such as Snowflake. Legacy businesses who are slow to develop their tech stack and talent pool will be most at risk.
Mike Merritt-Holmes is the co-founder and COO of Kvasir Technologies, a systematic hedge fund that uses machine learning techniques to power its investment strategy. Commenting on the CHATGPT hype, he says:
“Being a company whose strategy is driven by AI, it is not necessarily a big surprise. Hedge fund allocators, in a way, forgot about AI as the hype around crypto took off. Now, with ChatGPT and other generative AI tools, it has come back into the fore.”
In his view, generative AI could benefit discretionary fund managers as a tool for collecting more information on companies – beyond merely financial data - and lead to highly original insights that only an AI-driven system could be capable of. Think of the famous “move 37” that AlphaGo made in game 2 when it beat Lee Sedol in 2016.
That flash of genius – of super-human intelligence – is likely to ignite into a fire of possibilities.
“Hedge fund allocators, in a way, forgot about AI as the hype around crypto took off.”
Mike Merritt-Holmes, Co-founder & COO, Kvasir Technologies
A golden oracle..?
With the hype, however, come caveats. One only needs to scratch under the surface of LLM-based AI applications to realise that they are only as good as the data behind them. Without that fuel, which AI can use to create fake photos of the Pope or resurrect the Notorious B.I.G. delivering Nas’s N.Y. State of Mind, there is no possibility of it acting like a golden oracle. Unless, that is, the appropriate machine learning models are used to predict future patterns and outcomes from the data using neural networks and deep learning algorithms.
As WorldQuant highlighted in a recent article, it was an American psychologist, Frank Rosenblatt, who developed the first machine learning system in 1957, named the “Perceptron”.
“In its current form, generative AI represents an exciting new era for deep learning,” says Merritt-Holmes. “You can see how, in 10 years time, we will be relying on these tools for a lot of decision-making. At present, though, I wouldn’t try to use generative AI for financial decision-making. These tools have not been built with financial models in mind, and are not backed by the data needed to make such decisions.”
He says that one beneficial application already being deployed within the hedge fund industry is as a co-pilot, “giving it discrete tasks to speed up engineering work (i.e. checking code before implementation), or building out the basic framework for a piece of code, which you can then take and edit. That’s quite exciting as it will speed up productivity for engineers and data scientists.”
Amsterdam-based Theta Capital, best known for its Legends fund-of-hedge-funds vehicle, which provides investors with unparalleled access to some of the industry’s best hedge fund managers, has been using AI to streamline various operational processes including fund due diligence.
It has also, via its Blockchain Ventures Fund (now its third vintage), sought out investments in generative AI, with Marc de Kloe, Partner, Risk Management & Compliance, confirming: “We have exposure to ChatGPT’s rival, Stability AI, which we invested in at an early stage.”
“The team at Renaissance Capital are pulling in something like 40 terrabytes of data every week.”
Marc de Kloe, Partner, Risk Management & Compliance, Theta Capital
AI has become humanised
He believes generative AI will open up the space for stock pickers, over time. For now, however, he sees it being more the preserve for the biggest hedge funds, “who have the infrastructure for sourcing and pulling in the data”.
“The team at Renaissance Capital are pulling in something like 40 terrabytes of data every week. They have 120 PhDs working on ways to process all of that data,” he says.
What ChatGPT has done, in de Kloe’s view, is ‘humanise’ AI by its ability to give human-like answers, thanks to the huge computing power that now exists.
“If you compare it to something like cleantech, AI has raised a fraction of the money – but one of the things we are seeing, even in our Blockchain Ventures Fund, is how generative AI could impact blockchain investing.
“With new technology, because all the pipes and the plumbing already there, the scalability of it is so much faster. Even though ChatGPT is a brand new technology, within months 100 million people were using it. You can drop that equation into so many other things as well; i.e. a new protocol for investing on blockchain. It begs the question, ‘What’s next?’”
Merritt-Holmes offers a final thought:
“What ChatGPT is doing is showing that it’s possible to start to have machines talk to you like human beings; that’s where people get scared but also quite excited by generative AI’s potential. We are just at the start of this new journey. ChatGPT-10…who knows what that could be?”
Disclosure: the first four paragraphs of this article were written by my ChatGPT assistant editor.