Is ChatGPT going to Invest for you?
AI has taken the world by storm. Amongst them, ChatGPT is taking the main limelight for its capability in having superb linguistic abilities. It can write code, pass the BAR exam, write poems and essays, and we are sure you are aware all of the wonderful things that it is able to generate.
The reason we are writing this article is also because discussion topics surrounding generative AIs are really polemic. Some would believe that ChatGPT is going to be the destroyer of the worlds, hacking away the very foundation of human civilisation. Others seem to insist that ChatGPT is a completely harmless technology that is going to have minimal impact in our society.
While we could point you to many articles and resources surrounding this discussion, let us jump into the question that we all are here for:
Is ChatGPT going to invest for you?
Perhaps a more fruitful writeup along this topic is how relevant ChatGPT is in the investing world. In our view, ChatGPT is definitely going to have an impact, but we don’t think it is going to replace fund managers entirely and make decisions for all funds in the world.
Before we guesstimate on how ChatGPT is going to impact investments in the future, let us take a brief look on how ChatGPT works.
How does ChatGPT work, intuitively
ChatGPT is a large language model (LLM) that combines statistical and mathematical innovation, large computing power and large data (the entirety of the internet) to produce a model that answers prompts in a creepily humane way. But the incredibly simplified version of how it works is as follows:
A user inputs a prompt (easy to understand)
The model will parse the prompt (still easy to understand)
When the model finishes parsing the prompt, it establishes rules of interest (what rules?)
The rules are going to be used as a mechanism on how the model would vote for the subsequent words (response) to be generated.
Once the rules of interest is generated, the model will vote on the next word based on similar phrases of the input + response generated thus far.
For example, you key in:
How are you?
ChatGPT will parse “how”, “are”, “you”, and “?”. Then it will determine that the rule is Greetings in English. Similar phrases found will be
How are you? I’m good
How are you? Don’t ask
Based on the algorithm it will give a strong vote on I’m instead of Don’t. Then it will continue to give the vote based on the phrase “How are you? I’m”, the options maybe
How are you? I’m good
How are you? I’m well
Based on the algorithm, it will give a stronger vote on good instead of well.
Finally, after the terminating condition is met, it will return the response to us.
Actually how ChatGPT works is way more complicated than the above
There are many edge cases and complicated cases to produce results that are meaningful. Nonetheless, what the previous section is trying to tell us is that the underlying infrastructure that supports the entire model is a lot of math, compute power, and data. The following excerpt from computer scientist Cal Newport when asked about ChatGPT summarises it really well:
There are, of course, mind-numbing technical terms and complex concepts lurking behind all of these basic components. The layers are actually called transformer blocks, and they combine standard feed-forward neural networks with a cutting-edge technique known as multi-headed self-attention. We also skipped over a key innovation in the move from GPT-3 to ChatGPT, in which a new reinforcement learning model was added to the training process to help the program learn to interact more naturally with people.
Full graduate theses can and will be written on any one of these topics. None of this jargon is needed, however, to grasp the basics of what’s happening inside systems like ChatGPT. A user types a prompt into a chat interface; this prompt is transformed into a big collection of numbers, which are then multiplied against the billions of numerical values that define the program’s constituent neural networks, creating a cascade of frenetic math directed toward the humble goal of predicting useful words to output next. The result of these efforts might very well be jaw-dropping in its nuance and accuracy, but behind the scenes its generation lacks majesty. The system’s brilliance turns out to be the result less of a ghost in the machine than of the relentless churning of endless multiplications.
What Kind of Mind Does ChatGPT Have? - Cal Newport - The New Yorker - April 13, 2023
So what about our question?
From having the slightest sense of an intuition of how ChatGPT works, we no longer think of it as a blackbox, let alone a diabolical alien intelligence that is formulating a plan to destroy human civilisation. Furthermore it allows us to understand a critical property of ChatGPT:
It is not original.
ChatGPT does not generate new ideas when they are asked to write a response about how Tom and Jerry can help us fight climate change. Instead, it would copy and manipulate proses that are already available to come up with something that resembles something original.
This is a significant finding to help us answer our question of interest. It means that ChatGPT is unlikely to form its own (investment) opinion about a company. It is more likely that ChatGPT gathers the opinions of people in the internet, and then comes up with something that is seemingly original.
Does that mean that all fund managers are going to be replaced by ChatGPT? Unlikely. Human intelligence is still needed to give a proper interpretation for the data that they see from companies they analyse. Besides, a strategy in investing can be highly multi-faceted.
Do we think that ChatGPT is going to be a useful tool in investing? Absolutely. We think that the information processing that ChatGPT has at a breakneck speed is going to be very, very useful to fund managers in helping them making investment decisions. They can also perhaps help execute investing strategies that are designed by fund managers in a more unbiased manner (getting rid of greed and fear).
Ultimately, our view on the advent of ChatGPT to impact investing is quite similar to how the release of Google as a search engine is to impact investing.
Before Google, it’s hard to find information on the internet about a company. After Google, it is really easy to search for information due to its ability to index information and recommend relevant pages to users.
We think it will be quite similar to ChatGPT. Before ChatGPT, it is difficult to have a sense of people’s opinions, company’s data and prospects summarised in a relatively short format (would require notable man hours). After ChatGPT, it is one prompt away.
However, as with Google (powered by AI), ChatGPT (powered by AI) perhaps will not replace fund managers to invest for you entirely. The next AI breakthrough might 👀