Portfolio Strategy – Betting my retirement on Generative AI
Generative AI is cool, but this is not financial advice, do your own research and/or ask your financial advisor if you’re not sure what to invest in.
Since I left my last corporate job back in early 2022 and I was traveling for the better part of the past year, I put off rolling over my 401k to an IRA until recently. It’s not a huge amount of money, but it’s meaningful enough to put some thought into where I should invest.
I’m pretty young, so my investment horizon is 30+ years before RMDs kick in and I have to withdraw the money. I’m also interested in investing responsibly and I believe an AI boom is upon us for the next couple of decades, so I would like my retirement funds to reflect these beliefs. Lastly, my funds needed to be low cost, not just high yield, since the time horizon is long.
Like any normal person, I asked my financial advisor to weigh in and recommend a mix of ETFs where to hold my $.
But since we’re in the Generative AI era, I wanted to test two of the most popular models out there, GPT-4 and Bard, and see what they would recommend as well.
Here’s what I got from each:
GPT recommended a diversified strategy, US-focused, with a small bond component.
- Vanguard Tax Managed Fund FTSE Developed Markets ETF (VEA) – 10%
- iShares ESG Aware MSCI USA ETF (ESGU) – 17.5%
- Vanguard Total Bond Market Index Fund ETF (BND) – 5%
- iShares Core US Aggregate Bond ETF (AGG) – 5%
- Vanguard ESG US Stock ETF (ESGV) – 17.5%
- Vanguard Information Technology Index Fund ETF (VGT) – 22.5%
- Schwab US Large-Cap Growth ETF (SCHG) – 22.5%
Bard was more conservative on its recommendations, while still US-focused and heavy on growth + tech stocks.
- iShares ESG Aware MSCI USA ETF (ESGU) – 10%
- Vanguard Information Technology Index Fund ETF (VGT) – 15%
- Vanguard Total Stock Market Index Fund ETF (VTI) – 50%
- iShares Core S&P 500 ETF (IVV) – 25%
My financial advisors were even more conservative and concentrated, with a non-US diversification approach.
- Vanguard Emerging Markets Stock Index Fund ETF (VWO) – 15%
- Vanguard Tax Managed Fund FTSE Developed Markets ETF (VEA) – 20%
- Schwab US Broad Market ETF (SCHB) – 35%
- iShares ESG Aware MSCI USA ETF (ESGU) – 30%
There were some common choices, where all three approaches recommended ESGU, and both GPT and Bard brought in VGT. Bard and my financial advisors chose total market indexes as the highest weight %, while GPT was the only one to include bond ETFs in the mix.
Comparing backtested returns – GPT vs Bard vs People
I used Google Finance to create playground portfolios that allow me to backdate transactions as far as I want for the ETFs in each portfolio. The backtesting isn’t perfect, since ESGU and ESGV are relatively new ETFs, and only go back to 2018 and 2020 respectively. But I held the variables to the same standards, so I trust the results.
Backtesting with two weeks of data
I started testing with two weeks of data, with transactions on June 2. I wrote this article on June 16, when we had the following results:
- Bard return: 3.05%
- Financial advisor return: 2.97%
- GPT return: 2.71%
The race is very close, 10% maximum deviation between them, but the time interval is small.
Backtesting with two years of data
I went back to 2021 data and compared the results again:
- Bard return: 4.38%
- GPT return: 3.66%
- Financial advisor return: -5.18%
I was surprised by this result, personally, since the financial advisor recommended portfolio actually lost money in the 2-year backtest versus GPT and Bard, which were still very close.
Backtesting with five years of data
This is where ESGV and ESGU will skew the results a little, since they don’t go as far back as the other ETFs. Here’s what I got:
- GPT return: 69.58%
- Bard return: 69.08%
- Financial advisor return: 37.70%
Again, GPT and Bard were neck and neck, but the financial advisor portfolio lagged behind, generating about half the returns vs the generative AI recommendations. At least in this time horizon, everyone made money.
Backtesting with ten years of data
Again, this is where ESGV and ESGU will skew the results a little, since they don’t go as far back as the other ETFs. But this is a retirement account calculation, so I wanted to go on a longer time horizon to test it.
- Bard return: 199.97%
- GPT return: 192.24%
- Financial advisor return: 73.27%
On this time horizon, the financial advisor portfolio lagged behind even more, with only a 6.2% annual growth rate, whereas GPT and Bard delivered 11–12% annual growth rates.
Conclusion
I think the financial advisor portfolio lost some returns because of the ETFs that were non-US centric in a world where the US economy is far better relative to emerging markets or the developed economies, when it comes to growth rates.
Overall, I can declare generative AI the winner here with Bard, since it took the top spot 3 out of the 4 times, and I feel confident that the allocation it provided me makes the most sense for my investment thesis and time horizon.