34 – Math Model Mania



Print Article

Wall Street Quant: “Ugh. This is so frustrating! I don’t understand why I didn’t see this coming!” 

Artificial Intelligence Expert: “See what coming?”

WS Quant: “The market pullback. The stock market’s plunged over 25%, but my model predicted it would keep going up.” 

AI Expert: “Hmm. For an expert quant on Wall Street, that’s not a good look. As an expert in Artificial Intelligence, I suggest you look into employing the tools of my trade. You’re behind the times. At any rate, that’s too bad about your failed prediction. Did you backtest your model?” 

WS Quant: “Of course! I’ve built a remarkably accurate mathematical model that correctly predicted the last five recessions.” 

AI Expert: “You mean it accurately predicted the last five recessions after they happened, correct?” 

WS Quant: “Well, yes. I used historical data to backtest my model, and it worked perfectly! I even did some scenario testing by making up my own data with various hypothetical situations that might occur.”

AI Expert: “Sounds like you weren’t imaginative enough.” 

WS Quant: “That’s for sure.” 

AI Expert: “You should have used an A.I. machine learning model instead.”

WS Quant: “Really? How would that have helped?” 

AI Expert: “The machine learning model teaches itself.” 

WS Quant: “Teaches itself? How does it do that?”

AI Expert: “Well, first there’s an initial supervised learning stage where you train the model. Then, after it’s trained, the model is able to adjust its own parameter weights using gradient descent.”

WS Quant: “But how does it know what direction to adjust the weights? And which weights?”

AI Expert: “Based on experience. As it encounters new data, it determines how much its predictions vary from the observed data, and then it adjusts the weights for any deviations from its predicted outcome.” 

WS Quant: “So, let me get this straight. Your model screws up by making a bad prediction, and then incorporates the new information into its model moving forward?” 

AI Expert: “Exactly!” 

WS Quant: “And how does your A.I. model get trained in the first place?” 

AI Expert: “Using historical data.”

WS Quant: “The same historical data that my model predicted perfectly?” 

AI Expert: “Correct.” 

WS Quant: “Then I’m guessing that your model would have also screwed up and made the same wrong prediction about the stock market that I did?” 

AI Expert: “Right, but it will now incorporate the new data into the model, and adjust it accordingly.” 

WS Quant: “So, in essence, your model used the same historical data that my model did, made the same incorrect prediction that my model did, and will now incorporate the new data into the model, just like I will now do. Doesn’t seem to be anything different going on here.”

AI Expert: “Well, there is one very important difference. My model uses Artificial Intelligence. This results in my model getting billions of dollars in funding from thousands of clueless investors, because A.I. is the wave of the future, soon to take over the world.” 

WS Quant: “Wow! I like the ‘billions of dollars in funding’ part! Where do I sign up?” 

Share This Article: