Programming/Coding: Reigniting Excitement (using chatGPT or Claude.ai)

tldr

#AI. It's here!

Anyone who can verbalize and analyze requirements can now "create" Python code; they can create SQL or NoSQL databases on the fly, and anything they can dream up... all with #NLP and a chatbot. Currently, the codebase is simple and limited by copy buffer lengths; however, ...

By tomorrow, the codebase created by #chatbots could solve any problem... including

1. Create a HA/DR load balanced globally scaled infrastructure that supports millions of active users. 
2. Install #1 into AWS for me. 
3. Design the platform in a way to keep the price of ongoing operations to <= $0.00258 per hour of compute power for normal operations.        

After coming back to this article weeks later, the so-what I've discovered is... I am re-excited about coding. Claude.ai gets me started with a code snippet, and then I take off and enjoy problem solving (again). I ?? Python coding!!

Yeah... I think Mr. Huang was right. 英伟达

Let me jump to the end (the so-what) so you that you may have a reason to read the entire 5 minute article...

I typed an idea into OpenAI #chatGPT and Anthropic Claude.ai for deriative trading. I got the idea from another article. The result: I used the chatbot's code for my real-life deriative trading and assessing a company's value based on 2 Stage Discounted Cash Flows... and I

• Didn't lift a finger to code it.
• Didn't pay a dime to run it-- sometimes up to 10M model runs -- in < 45 seconds!
• Didn't need to know where my code was stored and in what repo...
• I just bookmark openai... and I bookmark colab... and I'm done!

After the first iteration

1. "I" extended the program's functionality to save my trading ideas into a database. #chatGPT suggested a SQL database.
2. Then "I" added error-handling and checked for one (1) edge case.
3. By "I" ... I mean #chatGPT and #claude.ai did the work.

If anyone's interested, I could share how the two #llms created slightly different codebases, and focused on different elements of the functionality.          

FWIW, initially I altered the code on my own in a Python code editor (e.g., changing prompts, formatting outputs, changing if statements). Then a ?? went off: Why not ask the chatbot to alter the code based on my new requirements | my words? It worked!

Sure, I typed my requirements, but I could have easily used 苹果 's #Siri or 亚马逊 's #Alexa to convert speech to text. I could have even tried to load an audio file in #claude.ai and tested out how that worked.

In a recent interview, Jensen Huang CEO of 英伟达 said future learners should spend more time on natural language processing #NLP than learning to code. I doubted him: I threw a virtual brick at my screen!!

Coding is eternal. Also, someone has to code the coding language that converts natural language to 0s and 1s. My senior Computer Science project at Drexel University was to create a compiler. That's code to create code. That will always be needed! Also, coding isn't about software engineering per se; coding is about

• Solving complex problems... logically and thoroughly
• Addressing edge cases | unique scenarios | being fool-proof, and
• Empowering users to get stuff done... faster and more efficiently.

Our approach to doing that is to ... engineer software

Engineers are always solving complex problems! So, I doubted and vehemently disagreed with Jensen Huang . He's wrong!! That's right I said it!! ??

I know... I know he's a super-star visionary | trillionaire | the next Steve Jobs, etc, but sometimes those CXOs talk game and try to be bombastic and polarizing.

But, Mr. Huang may have been right. Mr. Huang may be right. Truly, he may be the smartest person in the room.

Read a little further as I began to understand the use case Mr Huang was considering:

A world where we verbalize code.         

# Super engineers will be still be needed;
# but only super engineers will be needed.

if super_Engineer() == "Y":
    # enter and create codebase
else:
    # verbalize it, baby! 
    # talk to the chatbot, my friend!         

Over the last few months, I've been exploring how to use statistical models to identify the best stock market trades. Using (old school) Monte Carlo simulations and newer Black–Scholes models, Fidelity and other brokerages will offer you the probability of profit for your deriative trades. You present a deriative and Fidelity tells you the probability of profit. It's a trial and error approach.

I wanted to flip that model, though: instead of guessing and having Fidelity tell me; why not use these models to report out the trades that have (say) 70%+ certainty of profit?? So, I feed the ticker symbol and outputs the best deriative trade... No more guessing! I've become a hedge fund. Simple, right? ??

Parenthentically, I may be sound "market intelligent," but trust me, I only know what I know because of 强生公司 , Take Care Health Systems, AmerisourceBergen , etc. It was at work that we needed predictive analytics, population health management, statistics, and process excellence (TQM) to bend the healthcare curve, to remove defects in manufacturing, to forecast reach and ROI if we spent Brand Marketing Expenses (BMEs, aka money) on TV, digital, or other places to reach consumers. I learned present value, future value, #ML modeling, etc. as forecast decisions, bought and depreciated servers, routers, and networking gear... as we worked.

So, I asked chatGPT how to Python code for this -- then I used  (Colaboratory) to run the model - 10M times! For free! Yes, I know Python so I could edit the code, but did I have to know Python?? ??         

Anyway, I had my proof now: I could replicate Fidelity's or anyone's Black–Scholes models to predict probability of profit. Step 2 was to flip that model and spit out the most profitable deriative trades. But, then I got distracted by looking at 2 Stage Discounted Cash Flows through (i.e., is a stock was under or over-valued) via the article "A Look At The Intrinsic Value Of Netflix, Inc."

Then I went back to chatGPT and I got back to coding. I started small-- give me Python code to prompt for ... and printf out ...

Then I discovered, I didn't need to say prompt or printf... I could use natural language with chatGPT to create the code. ??        

You may be missing the point, so let me explain. I could say,

I want to store results in a database. How do I do that? Can you code that?

You still may be missing the point. I could alter the code by saying, (verbatim)

"The code -- I want to pull in today's stock price; also, I want it to output the market cap using currency in the billions- format it with commas if the company is in the trillions... also, format the output to include ... oh, add this prompt... now, give me all the code again... you know what?? i have a new idea: let's store each run in a database. at the beginning of the program ask for 'new run or load existing run (y/n)' - something like that; if it's a new run then skip loading. oh, and add error handling in case the wrong ticker symbol is entered. just add error checks everywhere!

and out came the code... without error. almost!

One time there was an error. I went back to chatGPT. It apologized as it noticed the code was errantly truncated. ?? The code included importing libraries to pull from Yahoo Finance, to open a SQL*Lite database, etc.

All from my English typing of requirements.

I'm using the code for my real-life deriative trading and assessing a company's value based on 2 Stage Discounted Cash Flows... and I

• Didn't lift a finger to code it.
• Didn't pay a dime to run it-- sometimes up to 10M model runs -- in < 45 seconds!
• Didn't need to know where my code was stored and in what repo...
• I just bookmark openai... and I bookmark colab... and I'm done!

Summary: Yes, the code is simplistic, but that's today!

Today, it was about 150 lines of code and three (3) libraries imported, but tomorrow it could be 1M lines of code, and 3,000 libraries imported.

????

And what if the limits are of my own making?? What if,

1. I simply don't know how to speak naturally? ?? ...
2. My creativity is the only limit! Once I unleash my thoughts any code (problem solved) is possible? ??
3. This is the world we live in??! ????

wdyt??

P.S.

I created the banner using LinkedIn's "AI" designs... I type what I want-- I wanted "robots grinning evily..." that was too hard for LinkedIn so I chose the robots you see here. Maybe... as soon as tomorrow, ... evil grinning robots will be there waiting for me.