LLMs aren't remotely like compilers or power tools

i recently heard a colleague telling another "vscode can't really help you when writing yaml files". I said that the yaml lsp supports loading a json schema, all you need is just one comment at the top to point the lsp to it, then you get reliable error reporting with red squigglies on every single keystroke.

He just shrugged and said "oh i just use ai".

What takes me one keystroke takes him a full minute. And he didn't even consider the use of actual tools

You had me up until you equated LLM usage to unreliable human delegation.

No human being is going to genuinely suggest to glue cheese onto a pizza to stop it sliding off.

These things aren't people. Stop anthropomorphizing a computer program. They'll make all the mistakes humans can make, as well as a new class of mistakes because they're just fancy next word prediction with no lived experience.

I've got to give the author credit, it seems like this is written without an LLM. But, there seems to be a lot that is arguable.

For example, the article says "Natural language is the interface of last resort."

But, is that actually true? Isn't the AI boom happening because N/L is truly available as an interface?

Anyway, please keep writing and asking tough questions.

This sounds like someone who hasn't used compilers for long. I remember plenty of fun compiler undefined behavior that had essentially non-deterministic outcomes.

And actually using power tools in the real world is pretty uncertain, try running a straight starting cut with a multitool against wood grain and then come back and talk to me.

Is there not a way to use a transformer that is deterministic in rounds as it is just maths? I thought it was temperature which randomized results but perhaps it is also something else.

Of course, the content returned is non-deterministic.

The author misses on machine learning compiler optimisation passes, dynamic compilers (aka JIT), profile guided data optimization passes, and that LLM driven compilers is becoming a hot topic in compiler research, see recent SIGPLA or PLDI papers.

Additionally they also miss on low-code/no-code orchestation tools like Boomi, Workato, Opal and co, where programming is moving from drawing boxes, into agent prompting.

Those are also programming from all practical business purposes.

This person is taking things too literally.

By power tool we simply mean speeds up work.

And by that definition LLMs are absolutely a power tool of epic proprtions!

If you ever coded before LLMs existed its like night and day.

People here take things too literally....

Yes an llm is not a compiler. But both speed things up... That is the point...

An llm is not a power tool but they both speed things up!

I don't care about determinism for something that is a simulacrum of though ... That is a feature not a flaw...

Finally we are getting better at this ... I get that we are going to lose our jobs a devs and I get that it is scary.. But honestly good riddance.. There is much more to do in this world that we can do after we just let go and accept that what we knew is gone ...

Just to be clear .. A correction Will likely come for the AI industry ... But a correction happened 25 years ago and the internet is still here :)

Just remember to not take life too seriously... You'll never get out alive :P