In this week’s blog article Matthew Humphreys discusses the do's and don'ts when using Optuma with the cloud.
The world was introduced to ChatGPT, the next generation of Artificial Intelligence (A.I.) platforms, in November of 2022. This natural language model using A.I. machine learning algorithms revolutionised the way people interacted with their computers. Within two months of its initial release, ChatGPT became the fastest-growing site in history. It quickly hit over 100 million daily users, faster than any other platform before it. In the same way Google redefined the search engine platform (to the point Googling entered the zeitgeist), ChatGPT is doing the same for A.I. based user-interactions.
What is ChatGPT and why should I care?
Good question, I’ll ask ChatGPT to describe what it is and how it works:
Hello! I’m ChatGPT, a language model created using Artificial Intelligence (AI) techniques. Put simply, I’m a computer program that can understand and generate human-like language.
To create me, my creators fed me with a massive amount of text from various sources such as books, articles, and websites. I used this data to learn patterns in language and gain an understanding of the way words, sentences, and paragraphs are used to convey meaning.
So now, when you ask me a question or give me a prompt, I use the knowledge I’ve gained from that data to generate a response that’s hopefully helpful and makes sense. I can handle a wide range of topics and can converse with you on various subjects.
In a nutshell, I’m a computer program that can understand and generate human-like language, and I work by using the vast amount of data that I’ve been trained on to generate responses to your questions and prompts.
Based on my own use of the system over the last few months I can tell you it’s being modest here. The scope of the things ChatGPT can do is astounding. It can review text and provide suggested improvements. It can learn your writing style and provide additional content. It can code in popular languages such as python or java. It will find and give several sources on a topic or write entire assessments. It’s not unreasonable to say this technology will see many areas (workplaces, schools, universities, etc.) having to adapt quickly to ensure it’s used effectively and ethically.
Think of ChatGPT as an artificial brain with perfect recall, but it’s still learning what to do with all that information.
This is an Optuma blog, what does this have to do with me?
A.I. is still in its early stages of development. What we’re seeing now is a beta version of A.I. from a single developer. Google has begun to roll out its own A.I. platform called Bard in a closed beta, and there are more advanced versions being developed right now.
We don’t yet know what this will mean for people trading the markets, but it would be crazy to assume A.I. won’t be used or have an effect in our industry. We’re already starting to see the early stages of this such as the A.I.-managed ETF AIEQ.
AIEQ, an AI-powered stock market fund, has had an interesting journey since its launch a few years ago. Initially the fund underperformed against the US market, but it gradually caught up and finally surpassed it in 2020 by at least seven percentage points. This was an impressive achievement for the fund. It demonstrated that AI technology can be effective in the stock market.
However, AIEQ has also faced challenges along the way. For example, its higher weightings in the tech and healthcare sectors saw a temporary dip in performance in Q2 2021. Despite these setbacks, AIEQ has had a good start to 2023, with an increase of 6.56% YTD, compared to 3.69% for the S&P500.
AIEQ’s performance vs the S&P500 Index
There have been examples of others using ChatGPT to summarise company financials or provide rebalancing suggestions based on an established portfolio.
Where A.I. will really shine will be its ability to search large swathes of information (the Optuma forum and Knowledge Base for example) and give you a detailed response. There’s also great potential for an A.I. platform like this to learn Optuma scripting language and assist in building a script based on a set of requirements listed in plain text. There’s a lot of work to do before this is possible, but our initial investigations have been exciting and we hope to be incorporating this in 2023.
Can managing a portfolio really be boiled down to a series of 1’s and 0’s?
Can I retire now and let A.I. trade my Portfolio?
Well, you can, but it’s not something I’d be doing anytime soon. As amazing as this new technology is, there are still gaping flaws that have yet to be overcome. In the example of ChatGPT, the responses are always well-written, and provided with confidence, even when the information is completely and utterly incorrect. It’s still very much a “user beware” situation, and any information provided needs to be verified before being acted on.
The developers of ChatGPT (OpenAI) are also aware of the dangers associated with stock picking and financial advice. Any response in that area always comes with the note that a professional should be consulted for any investment decisions.
How far are we from being able to relax while A.I. does everything for us? I’m not sure, but all the non-chart images used in this article are originals, generated by an A.I. service called Dall-E with a few simple sets of directions to work from. Just like ChatGPT, it’s impressive but misses the mark in some areas. Another A.I. driven tech I’m excited to see evolve over the coming years.
This hasn’t stopped people from building Portfolios from ChatGPT interactions. Here’s an example of an experiment I ran where I provided ChatGPT with the following directions:
Using the stocks listed on the Australian stock market, build a portfolio of approximately 20 stocks that have low risk but offer good returns, take dividends into account, avoid stocks in the finance sector especially Buy Now Pay Later companies, and have a higher weighting towards emerging technologies and renewables.
Provide each stock with a weighting based on historical volatility, dividend yield, and ensure one sector is not overrepresented.
List the constituents and weightings of each stock in a table using an initial portfolio balance of $85K.
Matthew Humphreys prompting ChatGPT
Within seconds ChatGPT was happy to provide an example of 20 ASX stocks and their suggested weighting in the portfolio. However, as it was based on a 2021 database, the information was stale, and some BNPL companies were still included. Other suggested stocks were no longer listed.
The fact an A.I. is capable of producing results based on a set of conversational instructions like that is the part that’s impressive. The final product still needs a lot of work, and there’s no way I’d be using it in day-to-day analysis at this point in time.
For now, as exciting as the world of A.I. is, it’s still no match for an educated analyst working off a tested and proven trading plan.