Let’s address the question of AI in finance…
More specifically, we can think about the different outcomes we might see in the future and apply them to what we do in modern banking and elsewhere.
Lisa Huang has some interesting thoughts on this. After starting at Goldman Sachs in 2008, she worked at Betterment on a robo-advisor program, and now works at Fidelity in the area of AI for asset and wealth management.
Meeting Huang at a recent event, we can see some of his expertise in using AI for presentations, as well as some interesting insights into where things are going.
I found it very instructive how she uses the image of a dandelion to illustrate the beginning of her career, when it seemed like everything was floating in a mist.
I also thought it was a brilliant idea to use AI to design your own slides. In particular, she used beautiful.ai to write a disclaimer stating that her opinions are her own and not those of the company.
She also provides us with this aphorism, which really makes sense and is worth thinking about.
“I think we need to start thinking that the way you speak is a bit like a code in itself,” she says.
Regarding Huang’s view on AI in finance, she actually has a four-point projection on a formulated timeline. This includes the present, which she describes as “human and machine automation with a pinch of AI” and the near future, which she describes as “autonomous wealth management powered by AI.”
Then there are two other categories: the not-too-distant future and the very distant or “more distant” future.
In the not-so-distant future, she mentions decentralized finance as a major force in the financial world.
“Defi has the ability to split any asset,” she says, emphasizing the practical usefulness of this method for forming liquidity in the markets.
In the more distant future, she says, quantum computing could become paramount.
Next, we follow Lisa in a more granular examination of points in this timeline, starting with the present.
She answers her own whimsical question: “Why only a pinch?” talk about the high cost of error, a small data problem that many parties face, and the need for explainable models.
If you’re funding a trillion dollars, she said, you need models that are fairly interpretable and explainable.
It breaks down an investment cycle into three basic elements: what, how much and when.
“AI is starting to affect this whole investment cycle,” she says, also referring to a project aimed at asset allocation through collective intelligence.
“We looked at the collective intelligence of funds and trading, then derived an implicit reward function using inverse reinforcement learning,” she explains. “Once we have a reward function, we can then use RL technology to optimize transactions.”
She also mentions a behavioral gap that must be corrected when working on behalf of a client. She says the gap is a potential difference between benchmarking and returns.
“We try to do right by the customer and create value for them,” she says. “In fact, you have to understand their behavior. … It just depends on how you trade… you see people following the market in a bad way – we need to find ways to mitigate that.
She then discusses optimizing a tax-smart withdrawal strategy through reinforcement learning and presents a case study demonstrating a savings of $18,000 in tax burden.
Here’s another idea that came out of this conference: LLMs can provide financial advice on a large scale.
“I actually tested it as soon as it came out,” she says, revealing that the program kept telling her it couldn’t offer financial advice because it wasn’t a human being . Then, she says, she figured out how to better elicit it, and what she got looked a lot like portfolios she’d designed personally.
“It did a really good job,” she said. “I think people will use it in that form (for advice).”
In terms of personalized finance, she suggests the technology is already available to do this for people.
She mentions a series of four criteria for intelligent trading tools:
· Risk profiling
· Personalized information
· Monitoring and alerts
· Cash management
“You have to have supervision,” she said. “Are they misbehaving?” Are they buying when they shouldn’t? Do they sell when they panic? And then, you have to understand their entire financial situation, in order to optimize cash flow. So, in doing so, we will create the platform of the future.
Huang also reviews aspects of using an “AI planner.”
“You can have many goals,” she says, calling it a “mathematically solvable problem.” “You have different backgrounds.”
The AI engine of the future will be an investor, a planner, a therapist, an educator and a coach.
“He knows your risk profile,” she says. “He will give you advice in times of volatility – he will guide you to success.”
Her presentation also ends with a neat conclusion: instead of concluding with a statement, she shows us an AI-generated example of a fictional company called Futurawealth, and by watching the video you can see for yourself what that the program offers – something that looks very much like your general business page attracting potential customers!