RazedHunter is back after a break that he really needed!

Your patience is going to be rewarded, because episode three is a real well researched treat.

RazedHunter goes into great detail about how ridiculously unintelligent and horrifically resource intensive Gen AI is.

Our evil tech oligarchs are determined to suck the power grid empty and destroy the planet for a thing that's less accurate than a Magic 8 Ball.

AI takes larger token count and slows inference when solving more complex tasks

AI Agents fail basic tasks 91% of the time

AI is incapable of understanding of concepts

Consistent for simple tasks

AI inference cost rise per usage linearly

Jevons Paradox (increase in efficiency=increase in consumption) applies to AI scaling

The better the AI model=the more energy it uses

AI is only good at memorising data, not reasoning it

AI does not possess common sense reasoning that limits automation (and scaling gets diminishing returns)

Limitations of AI agents by predictive models

AI failure rates are on the rise

AI is falling short in reasoning

AI models try to "hack" to cheat in benchmark tests

AI companies rig their AI models for benchmarks

AI agents fail 20-40% of the time depending on the agent being used

AI knows when its being evaluated,making benchmarks less reliable

Limitations of mathematical reasoning of LLMs

AI can't actually think or reason

Limitations on LLMs when it comes to "thinking"