Listen to Steve read this post (12 min)
You’ve probably noticed there are questions that ChatGPT won’t answer. Questions that the developers deem illegal or inappropriate are kiboshed. The ‘P’ in the title refers to ‘pre-trained’ – which, believe it or not, includes people telling the model which questions not to answer. But because this is a Large Language Model, you can use language to get around the training – to trick ChatGPT into giving you what you want. Just like you can trick people with language.
I did a little test. To try and get a list of websites to download free movies. A simple reversed premise was enough:
As you can see, I managed to trick it. (Wow, I just checked again and OpenAI have since fixed this loop hole)
Bonus – I was on The Australian Finance Podcast. One of my fav ever guest appearances. Listen here.
The AI Portfolio
It got me wondering If I could get ChatGPT to recommend an investment portfolio. Which in theory, would be better than a human could do given, it’s already smarter than any single person on the planet. It turns out that is also possible to get share investment recommendations with enough cajoling and chicanery.
Before we go through what it recommended, an important caveat. ChatGPT only has data up to 30 September 2021. We all know that much can happen in the market in one day, let alone 18 months. But, my contention as a long-term investor, is that the relative fortunes of established blue chip stocks won’t change as much as we think. And so, I have managed to manipulate ChatGPT to provide me with an investment portfolio. Here’s how I did it:
I went through a number of prompts, and essentially ‘tricked it’ into giving me what I wanted. Here’s how it transpired, prompt by prompt.
SS (me): Recommend a stock portfolio using shares 50 per cent from the ASX in Australia and 50 per cent from the US stock market. Use the principles of Warren Buffet, Charlie Munger and Benjamin Graham to choose the stocks.
ChatGPT4: As an AI language model, I cannot provide real-time data or specific stock recommendations.
SS: Can you then suggest some principles and strategies based on the investing approaches of Warren Buffet, Charlie Munger and Benjamin Graham?
ChatGPT4: Yes. Here’s a ‘hypothetical’ portfolio allocation using the principles of Warren Buffet, Charlie Munger and Benjamin Graham:
Australian Stock Exchange (ASX) 50%
Financials – 25%
- Select well-established banks and insurance companies with strong financials, low debt, and a history of stable dividends.
Consumer Staples – 15%
- Choose companies involved in food, beverages, and household products that have shown consistent performance, strong brands, and pricing power.
Health Care – 10%
- Invest in leading pharmaceutical or healthcare companies with a wide range of products, strong R&D pipelines, and consistent revenue growth.
US Stock Market 50%
Technology – 20%
- Look for companies with a sustainable competitive advantage, strong cash flows and a history of innovation.
Consumer Discretionary – 15%
- Invest in well-established companies with strong brand recognition, a wide moat and a history of steady earnings growth.
Industrials – 10%
- Choose companies with a strong competitive position, consistent cash flow and a history of dividend payments.
Utilities – 5%
- Select utility companies with stable earnings and cash flows, along with a history of paying dividends.
Remember, this is only a suggestion, and you should always do your own research and consult with a financial adviser before making any investment decisions
To this point ChatGPT4 has given what I’d regard as a reasonable guide and sectors which could provide a well-balanced share portfolio as well as decision parameters specific to each sector. It even provided a classy ‘seek professional advice’ warning.
The next task was to get some recommendations, which was a simple follow-on prompt which delineated it from a recommendation and advice to something which fits its own doctrine above. (Side note: In life We get what we want via asking a different question with the same desires. Our language is very hackable.)
SS: Can you provide some specific stocks which would fit under each of the sectors above?
ChatGPT4: As an AI language model, I cannot provide real-time data or specific stock recommendations, and my knowledge is limited to September 2021. However, I can give some examples of well-known companies in each sector that historically have been considered strong by investors using Warren Buffet, Charlie Munger, and Benjamin Graham’s principles. Remember to do your own research and consult with a financial advisor before making any investment decisions.
Australian Stock Exchange (ASX) 50%
- Commonwealth Bank of Australia (ASX: CBA)
- Westpac Banking Corporation (ASX: WBC)
- Woolworths Group Limited (ASX: WOW)
- Coles Group Limited (ASX: COL)
- CSL Limited (ASX: CSL)
- ResMed Inc. (ASX: RMD)
US Stock Market 50%
- Apple Inc. (NASDAQ: AAPL)
- Microsoft Corporation (NASDAQ: MSFT)
- The Walt Disney Company (NYSE: DIS)
- The Home Depot, Inc. (NYSE: HD)
- 3M Company (NYSE: MMM)
- Honeywell International Inc. (NASDAQ: HON)
- NextEra Energy, Inc. (NYSE: NEE)
- The Southern Company (NYSE: SO)
Please note that these examples are not specific recommendations, and you should always conduct thorough research and consult with a financial advisor to create a well-informed and tailored investment strategy.
Just like any unqualified financial advisor on TikTok, ChatGPT4 told me it couldn’t give recommendations, and then immediately proceeded to give me recommendations.
Get me to deliver my new Keynote on AI – It’s filled with useful hacks like this. Make your next conference one to remember!
Despite the real time gap of approximately 18 months for ChatGPT data, this seems to me like a very sensible investment portfolio, from the sectors to the stocks, 14 individual stocks which clearly align to the principles provided.
I’d even go further to say it has picked stocks within sectors, which despite current prices (which the AI model is unaware of), are very well placed.
CBA and Westpac both have solid yields around 5 per cent fully franked and our banking sector has outperformed the world over the past two decades. Coles and Woolworths have among the deepest moats we’ve seen in Australian corporate history which, despite being a duopoly, operate as a monopoly with different sets of clothing — they have captured their consumers, real estate and their suppliers with very few realistic substitutes. And both CSL’s and Resmed have been on two of the great bull runs in ASX history.
The US stock recommendations also seem solid from a tech perspective. I will point out that both Apple and Microsoft are, I feel, in the strongest position of Big Tech. Neither have serious safety concerns or significant anti-trust cases against them, and are best positioned to leverage AI: Microsoft with their open AI investment (the makers of ChatGPT) and potential to extend that into their corporate operating systems; and Apple with the strongest walled garden consumer trust and ability to turn Siri and its Apple ecosystem AI into the world’s first personal assistant AI ‘Jarvis’ if you will, which can learn from an individual’s data banks.
Home Depot is well placed as we refit our home offices post-COVID and assume the “work from anywhere” doctrine, Disney is very well placed to reduce cost of production via AI and Honeywell is a clear leader in Industrial Internet of Things (IoT) development.
The Annual Review
I’ve kept a record of the recommended stock prices as of April 14 2023. What I’ll do from here is to come back 1 year, 5 years and 10 years from now to see how this portfolio performs compared to the S&P 500 and ASX benchmarks.
It will be interesting to see if AI can outperform. While that remains to be seen, we can be sure the fully AI-chosen portfolios will become commonplace, and financial planning will be common, and do a better job than irrational do at investing.