It’s good to ChatGPT

Nicholas Pratt reports on the march of progress for AI-driven investing, which could have fund managers “looking nervously over their shoulders”.

In May, FundsTech reported on an investment fund created by the popular generative artificial intelligence (AI) tool ChatGPT that had outperformed the UK’s ten most popular funds.

The fictional fund was conceived by the UK-based personal finance comparison site finder.com as a “conceptual experiment” that would generate a debate about how the controversial technology may be used in the future. The main question is – will we still need fund managers in a generative AI-based future?

ChatGPT was launched in November 2022 by developer OpenAI, since when it has enjoyed a meteoric rise in usage. More than one million people signed up to use the tool within the first five days of its release. The user base grew to more than 100 million within the first three months.

The chatbot uses what is described as generative AI, an umbrella term for tools that use an enhanced version of machine learning, also known as ‘deep learning’. ChatGPT also uses what is known as reinforcement learning from human feedback, meaning that the more it is used, the more it learns.

It is also described as a large language model which employs a neural network with a massive amount of parameters. The previous version, GPT-3, used over 175 billion parameters, while the current GPT-4 model is rumoured to use more than one trillion parameters, something OpenAI has not yet confirmed.

“Big funds have increasingly been using AI for years, but the public using a rudimentary AI platform that openly says its data is patchy (...) doesn’t sound like a good idea.”

Finder.com’s fictional fund was created in March and is, in essence, a collection of 38 stock recommendations selected by ChatGPT.

In that time, the fund has generated 4.9% in returns, a significantly better performance than the then most popular funds on the UK market, which have collectively seen returns dip by -0.8% in that same time period, a difference of 5.7%. (Finder.com used data from Interactive Investor – which counts the number of “buys” – to glean the most popular funds).

The gap between the ChatGPT fund and the top-ten conventional funds, which includes offerings from the likes of Vanguard, Fidelity and HSBC, was at its widest on April 4, when it stood at 6.6%.

According to finder.com, the fund was created by following a range of investment principles offered by leading funds. Meanwhile, the chatbot’s warning that it “cannot provide investment advice” was overridden by telling the app that it was engaged in a ‘theoretical exercise’.

Finder.com also surveyed consumers for their views on the use of AI in the investment world. Almost one in five (19%) said they would consider getting financial advice from ChatGPT, while another 8% said they are already doing so.

Unsurprisingly, these figures were even higher when applied to younger generations, with 28% of millennials and 23% of Generation Z stating that they would take advice from the chatbot.

Unqualified TikTok stars

According to finder.com founder Jon Ostler, the big question is whether to use ChatGPT for investment research. “Big funds have increasingly been using AI for years, but the public using a rudimentary AI platform that openly says its data is patchy since September 2021, and lacks the intricacies of market psychology, doesn’t sound like a good idea,” he said.

“Yet a white paper we did in 2021 found that half of British investors use social media to get investment advice, and a fifth only uses social media. Would you rather get your advice from an unqualified TikTok star or AI that is capable of processing millions of data points from around the web and giving tailored advice?” he added.

While noting that ideally, investors would use primary sources or trusted advisors, Ostler added that AI seems to be something that will eventually “disrupt and revolutionise financial industries” and that “fund managers may be starting to look nervously over their shoulders”.

There are currently some very prominent barriers to the use of ChatGPT for investment advice – namely, regulation and the limits and ethics of the technology itself.

And no one has been more explicit about the technical shortcomings of ChatGPT than its creator, the chief executive of OpenAI, Sam Altman: “ChatGPT is incredibly limited but good enough at some things to create a misleading impression of greatness,” he said.

Altman has also stated that regulation of AI is “essential”. Speaking to a US Senate judiciary committee in May, he said: “We think that regulatory intervention by governments will be critical to mitigate the risks of increasingly powerful models.”

It is not just the US that is looking at regulation. In May, the European Parliament passed a new set of rules on AI use designed to enhance transparency and security. The AI Act also categorises certain AI platforms as “high-risk”, including ChatGPT, and imposes further transparency requirements upon them.

Similarly, the G7 has also agreed to promote “responsible” use of AI.

Altman still believes that the technology’s benefits outweigh its risks, including the concern that AI chatbots will put millions out of a job. “GPT-4 and tools like it are good at doing tasks, not jobs,” he said.

The trade association CFA Institute has produced a paper on how ChatGPT can be used by investment managers, complete with potential use cases. These use cases reflect the ability of generative AI to perform repetitive and natural language-based tasks that would support portfolio managers but not replace them.

The article reflects the current thinking that AI is more about augmented – not artificial – intelligence and that the principal benefit of the technology is to enable humans to get on with more value-added tasks instead of being bogged down with form-filling and other administrative tasks that provide little value to the end investor.

Productivity boost

However, one difference is that ChatGPT is able to produce content – from investment commentaries and research reports to press releases, marketing materials and, notably, personalised investment advice for clients.

“The dawn of the GenAI era marks the beginning of a transformation in how investment industry professionals and other white-collar professionals do their jobs,” states the CFA Institute paper. “Those who leverage AI as their co-pilot will boost their productivity, while those who fail to embrace this revolution risk losing their competitive edge. As various fields integrate AI, the technology will redefine the workplace and lead to new standards of efficiency and effectiveness.”

In March, API access was introduced for both ChatGPT and Whisper, OpenAI’s speech recognition software, paving the way for wealth managers and financial advisers to incorporate the technology into their offerings.

And large financial firms are starting to do just that. Morgan Stanley Wealth Management (MSWM) has run a series of tests with its financial advisers on the use of ChatGPT. MSWM will use GPT-4, the fourth multimodal large language model recently released by OpenAI, to “access, process and synthesise content” from its own library of research and data to be used by financial advisers.

Meanwhile, Bloomberg has developed BloombergGPT, an AI-based large language model (LLM) that will support natural language processing (NLP) tasks involving financial data. The tasks include sentiment analysis, named entry recognition, news classification and question/answering.

"Who leverage AI as their co-pilot will boost their productivity, while those who fail to embrace this revolution risk losing their competitive edge."

Even JP Morgan, which has restricted the use of ChatGPT among its own employees, has more than 300 use cases for AI in production, according to chief executive Jamie Dimon, who has called the technology “extraordinary and groundbreaking”.

But perhaps the biggest worry for fund managers is not how the technology may be used by financial firms but how it will be used by end investors. While tools like ChatGPT may be somewhat crude at the moment and prone to error, they are free, quick and accessible – three characteristics not commonly associated with fund managers.

While it is not advisable for investors to rely solely on a chatbot for their investment advice, the finder.com experiment shows that it is possible to override the safety mechanisms within the technology.

And the finder.com research shows that people are prepared to use it, especially the younger generations, which are set to be the beneficiaries of the largest intergenerational transfer of wealth ever known.

At the moment, there is no ChatGPT incarnation comparable to a fund manager. Instead, there are products that offer stock picks and skirt the line between market information and financial advice. And while regulated firms will be restricted in their use of the technology, fintechs and offshore firms may well use generative AI as part of their brand, thereby placing more pressure on fund managers to justify their existence, according to Ostler.

“They [fund managers] have to prove that there is more subtlety to fund management than just hard numbers – such as understanding investors’ risk appetite and anticipating market changes – but this technology will put even more demand on fund managers’ need to demonstrate their value.”