The Future of Investing

How ChatGPT is changing Wall Street

Hi there! I’m excited to share this article with you – it’s one of my longest and most detailed articles yet and goes deep into the field of AI and investing. Hope you enjoy reading it. You can subscribe below if you haven’t already:

In 2016, neuroscientist David Eagleman presented a fascinating idea in a Ted Talk: We could use technology to substitute missing senses in people. He said the brain doesn’t care where it gets its data from as long as it gets some data to work with. A bat is practically blind but can “see” the world through its hearing. A snake “sees” the world through its temperature. David took this idea one step further by creating a vest that would pick up sound in the room and vibrate according to the sound. Someone with no hearing ability could wear this vest, and even without hearing a single sound, they would be able to associate spoken words with vibrations in the vest and learn to “hear” through their body. David even demonstrated this with someone who had been deaf since childhood but learned to hear through the vest in just four days! He went on to start a company that now develops technology to help people hear.

But there was another idea in his talk that went unexplored – David claimed that humans could also develop new senses which have never existed. The example he gave was a guy getting signals through his vest and pressing a green or red button based on the feedback the vest gave him. What the subject didn’t know was that he was getting real-time stock market data as input, and he was making buy or sell decisions for which he was getting feedback. It would have been fascinating to see whether he developed the ability to “see patterns” in the stock market! Sadly, I couldn’t find any additional information about the experiment.

But the idea of using Artificial Intelligence and technology to beat the stock market has been around for a long time. After all, the entire basis of AI is to give computers access to data and help them build their own model – a very similar idea to David Eagleman’s project. Sometimes, it looked like things worked very well, but the problem was that we had no clue why the computer picked the stocks that it chose, whether it was “right” or whether it was a guesswork-based system.

All that has changed now. ChatGPT made waves earlier this year with its ability to answer any question you could ask it – but it’s now unlocked a possibility that we never had before. It can create investing strategies and also explain its decisions. The obvious thing to do would be to see if ChatGPT could beat the market and teach us how it’s doing so – and people have already started doing this.

In this deep-dive, we’ll see:

• How ChatGPT is putting active fund managers to shame.
• The earlier attempts at AI-based investing and how this time is different.
• The effects that AI could have on the market in the long term.
• How technology disrupted the market last time when High-Frequency Trading made its entry, and what’s likely to happen next with AI investing.

And most important of all – How is all of this likely to affect your investments, and how can you make money with this information? Let’s get started.

The smartest AI ever

ChatGPT is probably the biggest thing that’s happened to the knowledge search space since Google. It took Instagram 2.5 months to hit 100 million users, but ChatGPT reached that number in just five days! Now, its user base has grown to over 100 million. It’s not just an impressive statistic – it has massive implications for the effect ChatGPT can have on the market, and we’ll come back to that later.

Just to get us all on the same page: ChatGPT is a computer program that lets you ask it almost any question and then spits back an answer using Artificial Intelligence. It’s like your own robotic helper (which is definitely not sentient) who reads through an entire database of information, weeds out the unnecessary stuff, and summarizes it into an answer that saves you time – within a matter of seconds.

We’ve had Chatbots before, but ChatGPT was so effective and eerily good that it triggered alarms among all tech giants to compete by introducing their own projects. Google hastily launched and botched up the demo of “Bard”, its own competitor while Microsoft thought it was smarter to just buy into the company that made ChatGPT. Elon Musk and many others warned that unchecked AI would lead to a Terminator scenario – and YouTubers, including me, made videos on how you could make the most of ChatGPT.

But that was just the beginning. On March 14th, OpenAI released ChatGPT-4 and they said that the new one was so powerful compared to the old one, that it was like comparing the size of the sun to the Earth. In this context, it made ChatGPT-4 a superintelligent computer the likes of which we’ve never seen before.

These were not empty claims either: ChatGPT went on to successfully pass exams that take humans years to prepare for – like acing the SAT, AP exams, Sommelier (wine expert) tests, 4 medical exams, and even the State Bar – on the first try! But you know where the best minds of the country are eventually funneled into, be it Harvard or Stanford or Yale – they all end up at Wall Street. And why should ChatGPT be any exception? Introducing:

ChatGPT investing

Asking ChatGPT to summarize books and give you “how-to” guides is like asking Einstein to play the violin. It’s amusing and relatable, but there’s so much more it can do – and the investing community was quick to catch on. Analysts at Finder.com created a dummy portfolio of 38 stocks with ChatGPT’s help. According to CNN, the portfolio gained 4.9% while its benchmark clocked an average loss of 0.8%!

Now, this isn’t a magic pill – ChatGPT won’t give you the solution to make up your losses after listening to basement-dwelling degens on WallStreetBets who were YOLOing their savings to enter the hall of fame. The dummy fund was put together with some ​strict criteria to create a portfolio of stocks from high-quality businesses that would compete against the leading funds in its category. The criteria included:

• Low levels of debt
• Sustained growth
• Assets that generate a competitive advantage

The benchmark was set as the 10 most popular funds in the UK. 9 weeks later, the Finder.com fund is still going strong – outperforming the overall market by 5%. The fund has created a sensation, and now people want to create their own portfolios with ChatGPT to make more money. With ChatGPT being software that’s technically accessible to every person in the world, this is something that’s really possible! And this is where the ChatGPT fund comes in.

First of all, nine weeks is a really short time when it comes to investing benchmarks. Investing is a long-term game, and returns have to be measured over decades – and we still don’t have enough data on whether ChatGPT can consistently beat the benchmark. The second thing is that ChatGPT beats fund managers, but that isn’t saying much, considering that professional fund managers have historically been terrible at beating the market – It’s been shown that throwing random darts blindfolded outperforms hedge funds, fish beat the market by 14.5% over 3 years, a cat outperformed by 8%, and a chimpanzee beat 94% of Russian bankers.

In fact, the most successful investors are the ones who do nothing – like literally nothing, because they’re dead. In a study run by Fidelity, it turns out that the people with the best returns were dead people whose portfolios hadn’t been liquidated, just ahead of people who had forgotten their password.

So to test if ChatGPT beats the market over the long term, a new challenge has been started: ChatGPT is given a simple prompt as follows:

It turns out that the bot is in, and ChatGPTFund was started with this concept. Based on a research paper that successfully used ChatGPT to analyze stock headlines, the first stage is to figure out whether the headline is positive, negative, or neutral for a stock. Then ChatGPT picks the headlines with the largest movements in stock price – and it turns out that it could learn pretty fast what sort of headline would cause a stock to fall the most. The program found that negative articles had a strong chance of leading to a large, immediate move in price. The program then focused on looking for such moves among smaller stocks where the potential for upside was higher. By shorting them (betting against them), it was able to make profits.

ChatGPT was given six parameters to beat the market: 3 of them lost money, while the best dataset increased by 500%! If you had followed all of them, you would still have broken even. Since this was a one-off experiment, we’re not sure if this strategy is repeatable or if it’s a fluke. One of the major problems with ChatGPT was that it had information only till 2021, but you can feed it custom information to train it on new information. The real challenge is getting ChatGPT to give stock picks because it does not give financial advice – but people have figured out how to “jailbreak” this and get around the problem. Now so many people trust it that more than $2 million has been committed to the fund to follow along and see how it does!

The perfect storm

This isn’t the first time someone has tried to automate investing or bring technology into the picture:

• A YouTuber built a trading bot and gave it $30,000 to trade automatically. It did a stunning job.
• One of the few people who have beaten Warren Buffett’s stunning long-term record is Jim Simons with his Medallion Fund – and it was run by him completely hands-off. He had no clue about the trades his program was making. He trusted it to learn by itself, and ended up making a 66% annual return for a straight 20 years!​
• ​Numerai is a crowdsourced fund that runs completely on AI – people submit their programs and stake money in it, and if their models work, they are rewarded for it. The best program eventually finds its way to the top!
• And of course, you know what WallStreetBets did with the power of coordinated social media, giving hedge funds a run for their money and causing Robinhood to pull some controversial moves.

The thing is, ChatGPT is the perfect storm that could combine all of these features together. It takes the power of funds with high-tech computers (like the Medallion Fund) and makes that technology accessible to any talented programmers – and if the programmers have a winning program, anybody can invest with them.

Sometimes, programmers or fund managers need not even enter the picture. People could be given the power to play with their own portfolios with the help of an AI assistant. Here are some examples:

• ​Portfolio Pilot lets users “copy and paste their portfolios into the tool for analysis, receive investment recommendations, search for market information - and ask the bot questions – all for free.” If you think “Who on earth would trust an bot?”, the answer is: High-Income Americans. 77% of them reported that they’ve utilized it for stock picks, or, the way I see it: They have the disposable income to give it a shot for all of our amusement.
• Or, you also have the investment AI “Magnifi,” which “uses computer programs to provide personalized, data-driven investment advice” while “democratizing intelligence in a personalized way.” Essentially, the publisher on CNBC tested it out – and found it to be an effective way to quickly answer questions, give up-to-date answers on the portfolio, and recommend specific companies depending on what you ask it.

This could lead to some interesting effects: As more and more people start using a certain AI tool, it could create self-fulfilling prophecies that control the market. You don’t think this could happen? It already does. Jim Cramer’s stock picks are… shall we say, controversial. And yet, because so many people watch his show, they buy into his recommendations and his stock picks reliably go up in the short term. It’s called the Cramer Bounce.​

A similar thing happened with Robinhood, where a large portion of its users were just investing in the “Top Movers” – those stocks traded 14% higher than they would have seen otherwise. If ChatGPT starts attracting the attention of a lot of people, it could start moving the market in a massive way because it can explain the decisions it’s making. With the right marketing, this herding mentality could be channeled into something big.

But that leads to the question – who takes the blame when things go wrong? Let’s rewind the clock a little bit.

When things fall apart

High Frequency Trading revolutionized trading in 2010 – new developments in technology allowed 240,000 shares to be traded before you could blink your eye. Trading went to a realm where humans just could not compete with computers trading millions of shares every minute. The race was on to write better and better programs that would beat the market.

And then on 6th May 2010, the Dow Jones crashed by more than 9% in less than 40 minutes – more than a trillion dollars in worth, and rebounded just as quickly. The market was thrown into chaos. After scrambling for answers, it was found that Navinder Sarao, a programmer, had “spoofed” the stock market’s programs to place and revoke fake orders and manipulate the market. At least, that’s whom the blame was pinned on finally – because the real problem was that there was no security in place for such a possibility. Imagine a single person being able to move trillion dollars in minutes!

That was before the era of AI – now with ChatGPT entering the picture, things become even more complicated. There are a lot of questions we have no answers for:

• Are companies that give personalized advice using ChatGPT liable if their customers lose money?
• Can ChatGPT or other AI programs manipulate the market? If they can, who is liable for it?
• What happens when we combine AI with automated trading? Who is responsible if mistakes happen in that space?

Normally, when you ask ChatGPT about investing, it emphasizes that “when you build a portfolio, you should incorporate your investment objectives, timeframe, and risk tolerance.” That makes sense on a broad level, but every situation is different, and what works for me might not work for you. Some people want to take way more risk, while others want to protect the money they have at the cost of lower returns. Also, investing is highly emotional and not “logical” like it seems on the surface. How would ChatGPT account for all these?

It’ll be interesting to watch how this pans out over time – so the question is, what should you do with your money? Is AI worth the risk?

Let’s assume that AI does figure out the ultimate investing strategy: But what happens if everyone piles on? The Efficient Market Hypothesis states that “when new information comes into the market, it is immediately reflected in stock prices and thus cannot generate excess returns.” So, if everyone has access to something while AI begins recommending the same stock to the entire planet, it no longer works.

With investing, risk is proportional to reward – and I think if you’re planning to invest like you’re going to be around for a while, it doesn’t make sense to take on that extra risk. I personally feel investing should be dumb. If dead investors make the most returns, it’s because they don’t change their investing plan. An S&P500 index fund might sound boring, especially in comparison to Artificial Intelligence, but for most people, it’s the right solution.

Where you can use AI is to improve your investing knowledge and learn faster: ChatGPT is extraordinary at simplifying things and making it digestible for the average person who doesn’t have time to sift through charts and earning reports. If that’s how it “takes over the world”, I’m all for it.

Stay safe, stay invested, and I’ll see you next time – Graham Stephan.

If you liked this article, you can forward it to one person who might find it useful :)

Disclaimer: This is not financial advice. This information is intended to supplement your knowledge in the field of investing and personal finance. Please do your own research carefully.

113 Cherry St #92768, Seattle, WA 98104-2205
​Unsubscribe · Preferences​