Danelfin AI Investing Update #2 - This Experiment Sucks

I think I’m done with this. It was supposed to be a one year experiment. A year of living dangerously. Now it’s apparently two months of just living stupidly. What’s got me so upset? First the results, and then what it taught me about AI stock picking.

Overall, 10 out of the 11 investments I’ve made have fallen in value. All of these were Top Picks. My portfolio is down a whopping 20%, while the S&P 500 has fallen from 4450 to 4166, a drop of only 7%. Every stock currently in my portfolio is still rated at least a 9 out of 10.

One of the ‘Top Picks’, SEDG, is down an astonishing 49% in only seven weeks. Top Pick my ass.

After starting this on Sept 12, the only sell signal I've gotten is ESTA on Sept 21. It had fallen to a rating of 7 and was down 11% from where I bought it. I sold after it stayed there for two days. Since then it has fallen another 35%, so in this case I can say the Danelfin AI saved me some losses. But let’s also remember that a rating of 7 from them is a weak buy, not ‘sell’.

I replaced ESTA with MTD, a stock rated a full, perfect ‘10’ by Danelfin. It’s down 13% in the last six weeks (S&P down 3%).

My portfolio is down a whopping 20%, while the S&P 500 has fallen from 4450 to 4166, a drop of only 7%.

Here’s what I’ve learned.

First, the reason I did this experiment at all is because I had this feeling, this itch, that I was missing out on something. That itch is gone. So in that sense the experiment has been successful. Expensive, but some of the best lessons are.

I found out about Danelfin and AI stock picking in general in a newsletter I subscribe to, The Rundown AI. Danelfin was mentioned in a section of the newsletter called ‘trending AI tools’ which I mistakenly interpreted as ‘awesome new tools that are revolutionizing their field’ the way ChatGPT did. Boy was I wrong. It’s just a random collection of AI-branded websites, some of which I’m sure paid for their placement.

I’m not going to link to the AI newsletter because frankly, I don’t recommend it. The point here is that the information you choose to surround yourself with affects who you become. For example, subscribing to this newsletter will make you smarter and improve your complexion. Be careful what else you subscribe to.

Here’s the biggest learning. When I started writing this newsletter, I worried that I was entering a dying field. Wasn’t AI already writing better than humans at writing? But then I tried to use ChatGPT to write something and it sucked. Bland, unimaginative, and sometimes flat out wrong.

The magic trick generative AI plays on us is that compared to complete ignorance, it’s pretty damn good. Since we’re all close to complete ignorance on 95% of human knowledge, it appears intelligent when we ask it for a medical diagnosis or an article about reusing nuclear fuel.

But if you do know a damn thing about a field, generative AI like ChatGPT is about 70% as good as you, and it’s obvious. I don’t know why I thought this would be different for stock picking, but there you are, and you’re welcome.

I used to be an AI apologist anytime it didn’t live up to expectations. I would say, “Just wait! It’s going to get better!" But I’m waving the AI flag more slowly now. The very best outcomes in every industry are going to come from AI coupled with human expertise for quite a while yet, probably decades.

For now, I’m going to continue running the experiment for the full three months that Danelfin says it takes for its picks to outperform the market. Who knows, maybe in an up market it’ll outperform just as well as it outsucks in this down market.

1. Newsletters I actually recommend: Morning Brew for daily news, Tricycle Day for psychedelics, and 3-2-1 Thursdays for random inspiration.

2. My original AI investing post, and Update #1.