YouTube's AI is Crazy, Likes Clickbait, Is Pessimistic and Is Like a Goat

I wanted to quickly share my personal opinion on the rise of Artificial Intelligence (AI), specifically on the YouTube platform, where an option called "Inspiration" recently appeared for me.

This YouTube feature suggests video ideas, categorized by "Interest". When reloading the page, the suggestions change, and I've been struck by both the quality of the implementation and the consistency of the suggestions:

• Suggested Title,Description,Interest
• Ultimate Mac Setup for Laravel Developers, Guide on the optimal Mac configuration for Laravel developers - Moderate
• Laravel Livewire: Building Dynamic UIs, Step-by-step guide to building dynamic interfaces with Livewire - High
• Mastering JSON Fields in Laravel, Best practices guide for working with JSON fields in Laravel - Low
• Spatie Roles and Permissions: A Deeper Dive, Deep exploration of the Spatie package for managing roles and permissions - Low
• The Future of Udemy: A Developer's Perspective, Unique perspective on the future of Udemy for developers - Very Low

1. Criticism of Design and Titles

• Thumbnails: The generated thumbnails are, for me, a disaster. They rarely have anything to do with Laravel or the specific topic.
• Clickbait: The titles are excessively grandiloquent (e.g., "Ultimate Mac Setup for Laravel Developers"). This is a general criticism I have of the platform: they promote clickbait and appeal to emotion more than real content. While this can generate more views, I am not morally fond of it.
• Generic Topics: Titles like "Laravel Livewire: Building Dynamic UIs" are extremely generic. A dynamic interface can be built in a thousand ways (Vue, Inertia, Livewire). This shows that the AI is based purely on search terms associated with my content, but without a criterion of originality or specificity.

2. Contradiction in Suggested Interest

It's curious that the AI recommends content with "Low Interest" (e.g., Mastering JSON Fields or Spatie Roles and Permissions).

• Niche Logic: This reflects the reality of the development niche: the more advanced or specific the topic, the smaller the audience.
• Value Question: Why recommend something that the AI itself considers to be of low interest? At most, it would serve to know what I shouldn't do.

Lack of Coherence in Specific Suggestions

The AI problem is aggravated by suggestions that make no logical or technical sense, which underscores the need to be cautious with its results.

Contradiction Example (Laravel, Inertia, Alpine)

• The AI suggested: "Build a Full Stack App with Laravel, Inertia, and Alpine."
• Inertia and Alpine: Inertia (which uses Vue/React) is a large client-side framework. Alpine is designed for very small DOM manipulation tasks.
• The Problem: Using both is redundant and illogical. It's like proposing a new project using Laravel and CodeIgniter: if Laravel can do everything, why CodeIgniter? Starting a project with Vue (Inertia) and Alpine makes no sense, unless it involves a complex migration.

This recommendation is a disaster that makes no sense whatsoever. It shows that, despite being an application owned by Google (YouTube), AI results can be very poor.

The Danger of AI in Specific Requests (ChatGPT/Stripe)

Another example of the AI's lack of specificity occurred when integrating Stripe into my Academy application (which uses Vue on the frontend).

• The Context: I was using the Vue plugin for Stripe (vuestripe.com). I need it to return a Payment ID (Session ID) after payment on the client side to process the course assignment on the server.
• The Question to ChatGPT: I asked it: "How do I adapt Vue Stripe so I can process the payment on the server (I need the Session ID)?"
• The Useless Result: ChatGPT returned a result that included the complete implementation with Express.js and another library, something I did not ask for. I only needed an example of how to get the Session ID with the Vue plugin.

Conclusion

The message I want to convey is that the results from these AIs must be handled with extreme caution. No matter how large the companies behind them are, the results for very specific requests can be irrelevant, contradictory, or incorrect. We must always evaluate and validate what they return to us.