So, I have soiled myself and finally tried the overclocked text predictor (or what industry refers to as "AI").
I mean, more and more repos of software I depend on are introducing the slop code, maybe I was missing something?
Asked it various questions related to:
- Firmware (x86_64/ARM64), obviously focusing on coreboot
- Linux kernel
- Explaining snippets of C code
- Performing various arithmetic operations
- Doxxing myself (I can pinpoint where data would originate from, so I was curious which websites it would query)
To say it performed miserably would be an... understatement. It generated answers that looked plausible at first glance, but was absolute nonsense if you actually bother to read what's on the screen.
The only website it pulled data about myself was GitHub, specifically some of my old commits from few years ago that got merged into coreboot and ended up in Dasharo as 3mdeb rebased their fork (which is hosted on GitHub).
What I find especially egregious is that we built computers to do math in the first place... and now they're completely failing at what they were designed to.
So, here you go. If you're against AI (specifically LLMs/GenAI like I am), you're not crazy. Those systems are absolutely useless and fail at every single task given to them.
Sadly I cannot claim that I never touched AI anymore, but it was good to reassure myself that I wasn't wrong.
This bubble will burst *really* hard, I need to stock on popcorn because it's going to be epic
I mean, more and more repos of software I depend on are introducing the slop code, maybe I was missing something?
Asked it various questions related to:
- Firmware (x86_64/ARM64), obviously focusing on coreboot
- Linux kernel
- Explaining snippets of C code
- Performing various arithmetic operations
- Doxxing myself (I can pinpoint where data would originate from, so I was curious which websites it would query)
To say it performed miserably would be an... understatement. It generated answers that looked plausible at first glance, but was absolute nonsense if you actually bother to read what's on the screen.
The only website it pulled data about myself was GitHub, specifically some of my old commits from few years ago that got merged into coreboot and ended up in Dasharo as 3mdeb rebased their fork (which is hosted on GitHub).
What I find especially egregious is that we built computers to do math in the first place... and now they're completely failing at what they were designed to.
So, here you go. If you're against AI (specifically LLMs/GenAI like I am), you're not crazy. Those systems are absolutely useless and fail at every single task given to them.
Sadly I cannot claim that I never touched AI anymore, but it was good to reassure myself that I wasn't wrong.
This bubble will burst *really* hard, I need to stock on popcorn because it's going to be epic