They are all some binary file that completes a task. That LLMs have a statistically derived “personality” (if one would be dangerous enough to call it that) to make people like them was a purposeful mistake in development to trick people into embracing the technology. LLM is non-deterministic by nature which means results can never be trusted.

Honestly, the Photoshop filter analogy may be wrong because the filter is deterministic. The same settings produce the same results every time as it is math on an image file. LLMs literally can’t do that.

LLM tech is only seeing more “reliable” results now as they often throw the tool against a sandbox environment where it has to execute deterministic code to compensate for the random unreliability, and it still screws up.

Linus knows code, the tool sped up his process. In the absence of his knowledge, it would be useless.

Neat toy, but way too inefficient to bank the future on, and very dangerous to have such a religious reaction to.

LLMs are akin to fancy Photoshop filters for code. They can be used well by professionals and terrible by n00bs.

They are but a (very inefficient) tool.

OpenAI employee builds a rapport with an Apple employee on-site. Apple employee leaves laptop unlocked with OpenAI employee in their office and goes to the bathroom. No joke, it’s just that simple.

You’d think they would have learned that from the cloud or any SaaS provider multiple times over already.

Many of these schools forbid stuff like this

Coming soon to a USA near you: Constitutional amendment 28: Freedom of oppression. Indoctrination starting at the college/university level.