# LLMNOP: The Risk of AI # Introduction - Not a doomer <!-- stop --> - I actually use AI <!-- stop --> - But it has risks # Risk 1: Default Answers become De Facto Solutions > "Make me a todo list app" <!-- stop --> - It didn't do a server side rendered app in Rails - It didn't create a new MVC in Phoenix <!-- stop --> - It just made me a (broken) React application. # Risk 1: Default Answers become De Facto Solutions Ok fine... maybe that is the most human thing it could have done... # Anyway! - Use an LLM for an area you are not super familiar with <!-- stop --> - Accept the results, because they seem fine to you <!-- stop --> - Move on <!-- stop --> - The LLM host is happy because you sent them money <!-- stop --> What if everyone is doing this? # Natural Language Programmers This risk is be exacerbated by the "natural language programmer". - (Or majority of problems solved via natural language) <!-- stop --> Not likely to be suggestind particular technical details! # Natural Language Programmers - How will programming languages/libraries/frameworks gain adoption? - How would a new cloud service compete against the tech giants? # Natural Language Programmers > Don't worry! You don't need to learn anything! > AI is just going to take all the jobs! And do it perfectly! - The Internet, 18 months late on 6 months until no jobs. # LLMs Do Not Seek Truth - LLMs are not truth seeking machines - They seek to do what they are programmed to do - And to predict the next most likely thing - ... based on their training! Fortunately, we know economics, incentives, and reality don't need to apply to the benevolent people running my LLMs! # Does Not Require Malice If AI is great at generating typescript right now... There will be lots more typescript to train on and to test against. <!-- stop --> So the next most likely token to predict is probably Typescript. # Risk 2: LLMNOP - SEO for LLMs <!-- stop --> - Creating the LLM Nascent Oligopoly of Products <!-- stop --> - Or LLMNOP for short. # Risk 2: LLMNOP - Biases in: - LLM training data - Pre-training data - Reinforcement learning from human feedback - Prompts - Programming <!-- stop --> - True AI: **regexes and if statements** # Risk 2: LLMNOP If there are 1000x resources about React, how likely is the LLM to suggest a Svelte app? # Risk 2: Vertical Integration > "Deploy this to the cloud" - A naive user with a credit card # Not a `muh capitalism` - Bigger worry (for me) is regulatory capture # Not Just for Programming But this extends way beyond programming and technology choices - what about for brand names of anything! Why wouldn't LLM providers sell training services to Nike, Whole Foods, or Fortnite? # The End # Plug For Me :) Well even if AI takes our jobs, you could still write WoW mods in Lua. And if you want to learn how to do that, you might be interested to know that I'm creating a course that teaches you Lua :) It will be out sometime this year. > https://boot.dev/teej Enjoy 25% with promo code TEEJ on your first purchase.