models will get smarter, this wont be an issue

▲ reg_dunlop 5 hours ago | parent | next [-]

Intelligence, which I assume to be a synonym for "smart" requires the capacity to acquire and apply knowledge from experience.

These models do not have any experience. They're not sentient. And are in no way capable of being "smart", let alone becoming "smarter".

▲ alexjplant 4 hours ago | parent | prev | next [-]

The Claude web UI popped a modal up a few days ago advertising their new model to me. It was full of HTML tags that were escaped or otherwise not rendered so that the text was literally

  <b>Included in your plan limits until Jun 22</b> <br><br>Fable takes 2x the usage of Opus.
  <b> Switch models when a message is flagged</b><k <br> When safety measures flag a message, automatically switch to a different model to keep chatting. When off, your chat will pause instead. <a href="https://support.claude.com/en/articles/153636
  target="_blank" rel="noopener noreferrer" > Learn more</a>

...and this was presumably generated with the flagship model from the world's most prestigious LLM company.

▲ throwatdem12311 5 hours ago | parent | prev | next [-]

They say this every time. Just wait for the NEXT model bro THEN everything will be be fixed.

Ok wait maybe not the next one but surely the one after!

Hasn’t happened yet and there is no evidence it will.

▲

laichzeit0 4 hours ago | parent [-]

This all reminds me of in the 90s when the Borland C++ compilers and Turbo Pascal shit came out and everyone was still hand rolling assembler because the optimising compilers were so bad. I thought Opus 4.6 was pretty good, basically a step change. The stuff I got out of Fable before they blocked it was nearly alien. If things keep improving, I don’t see humans writing code in 2 to 3 years except maybe super niche areas. This will all go the way that optimising compilers did. No amount of resistance, anger or denialism will change that.

I’d actually love it if LLMs could skip the slow high level lanaguages entirely and just churned out some weird LLM bytecode that was closer to the metal. I don’t want to read it or understand it at all. Here’s my spec, build it and notify when done. I want to ship stuff not build or dick around with code. Basically like when I go to a shop because I want a table, I don’t care if some carpenter “crafted” it or a machine mass produced and spat it out. It’s cute, but most people just want stuff and don’t care how it’s built.

	▲	demorro 2 hours ago \| parent \| next [-]
		In every case up until now, a jump in abstraction has moved us forwards in ease of understanding the underlying artifact at a conceptual, human level. High level languages are effectively runnable documentation. It's possible to say that LLMs producing code may be the same category of thing, but the non-determinism and ephemerality of it all makes it difficult to imagine.
	▲	throwatdem12311 3 hours ago \| parent \| prev [-]
		If my experience having to manually unf*ck “production” slop written by sweatshop tier offshore told to go wild with a Claude subscription is any indication then we are a LONG way off from any of that BS. My job is safe because I’m the only person at the company that actually understands what the actual code is doing and I’m the one that gets the calls at 2am and weekends. “Weird LLM bytecode” Why not just generate object code for the target mschine directly?

▲ sfn42 5 hours ago | parent | prev [-]

Come talk to me when it isn't an issue.