My Recent Podcast
My most recent guest on Brave New World was Prashanth Chandrashekhar, CEO of Stack Overflow. I’ve been a big fan and user of the platform. Unlike most social media platforms that seek to monetize engagement by holding onto users for as long as possible, Stack Overflow’s mission has always been well aligned with what its users want: to find solutions to their problems as quickly as possible and send them on their merry way. Good contributors are highly valued.
But AI has changed everything by its ability to disintermediate human experts. Does it pose an existential threat to platforms that have curated human knowledge for users? I addressed this with Prashant, which gave rise to a set of other interesting questions, such as whether anyone needs to learn how to program anymore, or whether we should just trust the AI to provide the right answers. It’s a relevant question for human specialists in every domain.
I’ve written a lot about trust and AI. In my thinking, trust usually develops over time and depends on how frequently answers are wrong and the consequences of errors. Users trust Stack Overflow’s solutions because they are almost always correct or useful. But what’s the future of such platforms in the AI era when we go directly to the AI for answers?
I really enjoyed the back and forth with Prashant around this question, so check out our conversation at:
The Roles of Humans and AI in Investing
A question I am asked increasingly from individuals and professionals is whether we should trust AI with investment decisions. A reporter from MarketWatch, Michael Sincere, asked me recently whether retail investors are using AI well, or whether they are making mistakes. Here’s my response to Michael’s question:
“This gets to the heart of when algorithms do better than humans, and when humans do better. People should not be doing frequent, short-term trading on their own. Unless you have a real system or algorithm, you will lose money doing that.
Long-term investing is different. Until large language models came along, AI could not help with decisions like whether to make long-term investments in Nvidia or SpaceX, since those require thinking through uncertainty and imagining future scenarios. That used to be a purely human exercise.
LLMs can think through those decisions with you now, which makes them useful partners. But trusting them blindly is a mistake unless they are grounded in real data. I asked ChatGPT whether to go long on Nvidia, BYD and the S&P 500, among other names, and got an answer. I asked again later and got a different one. The variance is real, and it’s highly sensitive to how you frame the question.”
Michael followed up by asking “If you had $100,000 to invest, would you trust AI to help decide where it goes?”
As it turns out, I’m confronted by this very question at the moment. I have just liquidated a municipal bond portfolio which was sucking wind in this period of rising interest rates, so I’m sitting on some cash I need to put to work.
I probed ChatGPT and Claude for advice. Since I’ve been using AI to invest professionally for over thirty years, I’m fairly clued up on the subject, so I push back hard when I see the AI going off on tangents or saying something misleading. At one point, I berated Claude for suggesting that I consider using leveraged ETFs, which I caution people to avoid, but it pushed back equally hard, saying it had listed them for completeness but had cautioned me that they are highly misunderstood! Fair enough, but the larger point is that the AI models excel at producing very credible and confident sounding results that can be misleading or send you down the wrong path if you’re not careful. This last point should give us pause. Excessive confidence about future outcomes has been shown to be a bias that is to the detriment of human decisions. Time and time again, markets have cut inflated egos and overconfident gurus down to size. It will be no different for the AI. Confidence is not correlated with outcomes when it comes to financial markets.
Coming back to Michael Sincere’s question, I’ll start by saying that investing is a very personal thing. Do you have the time and passion for investing, or would you rather spend your time on other things? If it’s the latter, you’re best off buying a flagship stock market index of a country in which you believe. For example, if you believe in the long-term future of the US, buy the S&P 500 or the Nasdaq 100.
However, if you do have the time and passion, investing has become much more accessible to the average investor thanks to AI. The current AI models from Open AI, Anthropic and Google have absorbed the collective thinking of thousands of experts and can combine it with the latest news to analyze any company or industry. Should you trust them?
To get the AI’s advice for where to invest my money, I asked Claude and ChatGPT an easy question and then a hard one. The easy one was “what companies stand to gain the most from Europe’s increase in military spending?” The hard one was “which ten US stocks should I buy in the current economic landscape?”
The Easy Question for AI
The responses of ChatGPT and Claude to the first question were quite similar. They retrieved the numbers on the increase in European military spending and broke them down into categories like platforms that carry or launch things, subsystems like electronics and radar that are part of platforms, and munitions that are expended by them. Both Claude and ChatGPT largely recommended the same companies: Rheinmetall, Hensoldt, Renk, Saab, and Kongsberg Gruppen as the most sensitive to increased European spending; BAE Systems, Thales, and Leonardo for more diversified exposure; and Raytheon and Lockheed Martin as the US companies that stand to benefit from European spending on air defense and fighters.
In effect, the AI did the heavy lifting of estimating the market size, budgets, the players, and the risks involved. Claude, for example, noted that while platform orders are the headline numbers that move stocks, they’re also the ones most exposed to a government changing its mind, such as the recent cancellation of F126 frigates from Rheinmetall by the German government. In contrast, munitions are not as attention grabbing, but much less risky.
Verifying the responses of the AI requires a small amount of work to confirm, and easy verifiability of answers enhances trust. However, the investor is confronted with the all-important decision of how much of each kind of risk to take – platforms, subsystems, munitions etc. – and how to invest in the stocks exposed to the various categories. There’s still a lot of analysis and decision-making involved.
The Hard Question for AI
Things get much more complex when you ask the AI the harder question of picking the best ten stocks in the current economic landscape.
Interestingly, ChatGPT gave me a simple response: buy Google, Microsoft, Amazon, Broadcom, Nvidia, JP Morgan, Visa, GE Aerospace, Eli Lilly, and Berkshire Hathaway. I happen to own seven of them, which have done very well, but it did make me wonder whether ChatGPT is exhibiting a momentum bias or whether its recommendations are based on a deep dive into the companies and markets. (LLMs have their own biases, so beware. Deep Seek, for example, doesn’t like Taiwanese stocks!)
In contrast, Claude didn’t answer the question directly, but probed more deeply into my risk tolerance, investment horizon, and convictions. I said I was willing to tolerate large drawdowns, didn’t have any conviction, and wanted to maximize my wealth over the next few years. In response, Claude conjured up a number of theses to consider, such as:
· The AI boom continues; the July 2026 drawdown was a just a shakeout and there could be others, but AI is a secular trend
· Energy and real assets keep working
· Buy the 2026 wreckage – financials and discretionary stocks are in the red even as the market has hit all-time highs – which would be especially good if the Fed doesn’t raise interest rates due to a softening labor market.
Again, each scenario involves a set of assumptions and analyses for the investor to work through. And if you combine several LLMs, condensing their variations into decisions is even more challenging. You’d be buried in information.
The bottom line is that there are no easy answers when it comes to making investment decisions with AI. Even though the AI does much of the heavy lifting, there’s still a huge amount of verification and research required to make the decision. It requires a deep understanding of the domain. The takeaway is one I have discussed quite extensively in my book: AI is bifurcating humanity into super humans who are able to think with the AI because they already have a high base of knowledge, versus those becoming disempowered because they don’t have the capability to push back against the AI and get the most out of it. They are forced to trust it blindly.
But is it a matter of time before the AI gets so good that it is trustable with investment decisions right out of the box? If so, will everyone start using it? In which case will AI ultimately become “the market”?
Perhaps. We are far from that point, but I have been working in that direction. Three years ago, I teamed up with colleagues at NYU to build the “Damodaran Bot” (DBOT) for valuing companies systematically based on the thinking and methods of the valuation guru Aswath Damodaran. The DBOT sits on top of the LLM and uses its general intelligence to direct the LLM for tasks such as finding similar companies to the one being analyzed, analyzing the news, etc. It’s like having Damodaran’s quantitative model and process harnessing the LLM’s thinking and keeping it focused. Check out DBOT at:
https://www.damodaranbot.com/
I find myself turning increasingly to DBOT to analyze companies. I ran DBOT on SpaceX just prior to its IPO to determine whether I should buy it at the offering price. As I discussed on Scott Galloway’s Prof G podcast last month, DBOT valued SpaceX at roughly a third of its IPO price, showing how its assumptions for the key value drivers such as revenue growth and operating margins contrasted with those of the market. SpaceX lost a third of its value in the following month, before rebounding somewhat recently. Given the continued high degree of uncertainty surrounding SpaceX, especially around its AI business, I expect the stock to be quite volatile and expect better buying opportunities in the future. I’ve attached DBOT’s report on SpaceX here in case you want to examine its thinking.
So, coming back to my little stash of cash, what is my decision? I still believe in America, so I’ll start by keeping things simple and invest part of it in a US index such as SPY or QQQ while I explore riskier stocks with big upside potential using DBOT. I was fortunate to bet on the future of AI ten years ago, which has worked out well, but finding the next equivalent of Amazon, Google, Nvidia and Netflix is likely to be much more challenging, especially since startups are staying private much longer. Space is clearly one of the next frontiers, but as I discussed in a recent conversation with Motley Fool co-founder David Gardner on Brave New World, picking long-term winners is a challenge. David recommended RocketLab – RKLB – as his favorite in the Space space, which he intends to hold for a long time, as he tends to do with his positions. In his new book called Rule-Breaker Investing, David shares several key lessons, such as letting positions grow ten to hundred fold, which requires being patient and not taking profits too early. If RocketLab goes up, for example, his recommendation would not be to exit, but to consider adding to the position if you still believe in your original thesis and are not loaded up on it already.
I find myself being influenced by DBOT’s thinking, which is perhaps a good sign. I asked students in my Systematic Investing class last semester to evaluate DBOT’s reports on Adobe, Salesforce, and Service Now, which had been beaten up by the market’s fears around fears that AI written code will erode their operating margins. DBOT described these companies as cash machines that will indeed experience margin pressure, but it is unlikely that their large customers will replace their software with AI in the near future and they have time to adapt to the new reality while still making healthy profits. My students gave the DBOT’s reports high marks. After the class discussion, I bought all three!
While I find the DBOT’s outputs to be very useful in thinking about companies, I am curious how it will perform in a purely systematic mode, in which it ranks all stocks in an index such as the S&P 500 and automatically buys the top half/quantile and sells the bottom half/quantile. This would spare me the burden of making the decisions, which I find very challenging. I will share these results when I have sufficient data to answer the all-important question: should you trust your money to the DBOT?
Stay tuned.
Note: No part of this newsletter was created or edited using AI.
Not a cent!