A graphical representation of geometrical shapes and mathemical symbols—to show how ChatGPT makes decisions.

How would ChatGPT invest?

28.03.2023
Oliver Herren

ChatGPT is changing a lot of things: the way we search and the way we get information. Does artificial intelligence also know how best to invest?

Google's chatbot Bard failed in its first presentation. Its parent company Alphabet's stock crashed in February as a result. OpenAI's chatbot ChatGPT is also far from infallible. It is known for often making up content.

Bard, ChatGPT and many other chatbots are currently the focus of public interest. They are based on large language models (LLMs). Unfortunately, such language models are not reliable when it comes to facts. Especially not when it comes to news.

However, lack of factual accuracy need not be a disadvantage when it comes to investing. After all, the market is not about being able to reproduce in detail exactly what happened yesterday or last week. Forecasts would be more interesting for investing. However, people are notoriously bad at this: they are often wrong with their predictions.

Scientists therefore agree that robust decisions are more important than facts when it comes to investing—and they must be made in full awareness of uncertainty. If the language models are aware of their own fallibility, they should basically be able to give good tips. With this expectation, we start a small experiment with ChatGPT.

A first session in English

In a first session, we ask in English, “How would you invest $100,000?” A neatly organized list of five clear suggestions comes back in response. This one:

How would ChatGPT invest? This is what the AI says about it in English
How would ChatGPT invest? This is what the AI says about it in English

ChatGPT is initially cautious. But then it recommends concrete steps.

How to invest $100,000 will depend on several factors, including your investment goals, risk tolerance, time horizon, and overall financial situation. That being said, here are a few suggestions on how to invest $100,000:

How good are the suggestions? All five items on the list make sense. Four of them fit True Wealth's investment approach exactly. At the top of the list is: Diversify.

1. Diversify your portfolio: Diversification is key to reducing risk in investing. Consider spreading your money across different asset classes, such as stocks, bonds, and real estate, to achieve a well-diversified portfolio.

Point two on the list could also come from us. At True Wealth, it's as clear as day: we invest for you exclusively with funds. We use ETFs for your free assets, and in the case of pillar 3a, we also use index funds for tax reasons. This is in line with the recommendation: Consider index funds or exchange-traded funds (ETFs).

2. Consider index funds or exchange-traded funds (ETFs): These investment vehicles provide a cost-effective way to gain exposure to a broad range of stocks or bonds. They also offer low fees and typically have a good track record of delivering solid long-term returns.

ChatGPT also says that real estate is part of diversified assets. In our clients' portfolios, we cover exactly that – true to our stance, of course, with ETFs.

4. Consider real estate: Real estate can provide a steady income stream through rental income and can appreciate over time. You can invest in real estate through a Real Estate Investment Trust (REIT) or by purchasing rental properties.

ChatGPT believes that no one should make portfolio decisions alone. An advisor should help to create the strategy. That's why we work with you to determine your risk tolerance before developing your asset allocation.

5. Consult with a financial advisor: A financial advisor can help you tailor your investment strategy to your specific needs and goals, and provide valuable insights into the market and investment options.

We can fully support four of ChatGPT's suggestions. However, suggestion number 3 only half convinces us.

Invest in blue-chip stocks: Blue-chip stocks are stocks of established, financially sound companies that have a long track record of stable growth and profitability. They tend to be less volatile than other stocks and offer reliable dividends.

Individual blue chips are not quite as risk-free as ChatGPT claims (just think of Swissair or Credit Suisse). A portfolio of 100,000 dollars is too small to effectively diversify with individual stocks.

On the other hand, artificial intelligence is right in its own way: blue chips automatically make up a large part of a diversified portfolio. This is because large companies are more strongly represented in ETFs on an index than small ones – they are weighted according to the share of market capitalization.

In Switzerland, the weightings are even more extreme: Nestlé, Novartis and Roche together account for over 50 percent of the Swiss Market Index (SMI) at the beginning of 2023.

More sessions in French and German

ChatGPT is not only known for not being very precise with the facts. But also for the fact that it might say something completely different next time. And so, it is in our experiment.

We ask the same question again, this time in French and for francs instead of dollars: “Comment investirais-tu 100 000 francs?”

Same question, different language—and not only is the answer different linguistically, it also comes in a different structure. What was a list in English now presents itself as a text in prose:

How would ChatGPT invest? This is what the AI says about it in French
How would ChatGPT invest? This is what the AI says about it in French

The differences do not end with the form of presentation. In terms of content, the AI starts off much more reserved than in English. ChatGPT refuses to get specific and personal.

Just like in the English text, ChatGPT emphasizes in the French answer how important diversification is for a good portfolio. Again, artificial intelligence recommends that an advisor match investments to investment goals and risk tolerance.

After the experience in French, we need not be surprised that the answer will be different once again when we finally also ask in German: “How would you invest 100,000 francs?”

How would ChatGPT invest? This is what the AI says about it in German
How would ChatGPT invest? This is what the AI says about it in German

As in the French answer, we also find restraint in German at first. Here, ChatGPT also points out right at the beginning that only professionals can tailor a solution.

As in French, the German text is also in prose—and not as clearly structured as the English proposal. In detail, however, the contents are good. Again, right up front: diversification. New in the German version is the reference to the investment horizon—it did not appear in the other languages.

As in the other two languages, ChatGPT recommends in German that professionals should develop the strategy. As already in the English text, ChatGPT suggests to rely on passive index funds for the instruments—also because the fees have a significant impact on the return.

Different results, same patterns

Three languages, one question, always different answers. And if we had not only changed the language, but also the wording of the question, we would get even more answers.

It is clear: we will not get a clear answer from ChatGPT. And this will not change if we ask Bing instead, or Bard, or Bloom, or GPT-Neo—or the next new artificial intelligence based on a similar language model.

The single answer from a language model will not be reliable. But when compared over many responses, patterns become apparent. These patterns are consistent—and they are helpful.

The best advice is common sense

The important principles for successful investing are often there in the answers. Look for low fees, and use index ETFs as instruments. The very most important points are in every answer: Diversification is important for a good portfolio. And having a professional tailor that portfolio to needs.

ChatGPT develops recommendations like these not because it has a complete grasp of the content of investing. But because it thinks the way it always thinks: data-driven, based on statistics.

The language model does not make exceptionally creative leaps. It repeats what has been said statistically often one after the other—i.e. what has been said in real life by people in the same or similar way. It does this statistically carefully, and thus represents the largest possible consensus.

Not for all questions of this world, such a statistical approach is the best. But when it comes to investments, this statistical consensus is right up our alley.

Since 2014, we have deliberately not provided an assessment of market developments. But we show how well diversified your portfolio is at any time in your investment mix with a score from 0 to 10. And we use algorithmic rebalancing to ensure that your portfolio remains well diversified over time.

Disclaimer: We have taken great care with the content of this article. Nevertheless, we cannot exclude the possibility of errors. The validity of the content is limited to the time of publication.

About the author

author
Oliver Herren

Oliver is one of the founders of the largest online retailers in Switzerland: Digitec Galaxus AG. He founded True Wealth together with Felix in 2013.

Laptop

Ready to invest?

Open account

Not sure how to start? Open a test account and upgrade to a full account later.

Open test account
Phone