I was bitten by the same issue. Maybe add a line to the docs.

Added to the README, apologies for that.

I have a pet-peeve with this. As a non-native English speaker, I find it very useful to dictate multiple notes, in different languages, and have the LLM produce clear English prose out of it. The prose may be LLM-generated, but I edit it when needed to make sure that the contents is 100% mine.

It's like dictating to a typist like they did in the 60's - he will make sure that your letter looks professional and will fix your grammar, but you will sign the letter. This is totally different from LLM spam, the kind that inflates a sentence into a three-page article full of nothing.

So - is it a problem if the language reverts to a mean? that is the point of a shared language, right?

It's not just the language that reverts to a mean, it's the knowledge embedded in the model. If you're interested in discussing niche topics with ChatGPT, the further the model collapses the less likely you are to get meaningful results from the "tail" - the areas of knowledge that fall at the far ends of the model's bell curve.

Actually, both will, as they are not separate within the LLM. The thing is, one is a style issue, the other content. You can express original ideas and still use a lot of em dashes, or produce slop with a lot of typos in it.

Wish I had something like that when I was coding demos around 1988... :-)

(Edit): you kids have it easy.

I grew up in the 80's programming my C64, assuming that this would be more or less the sort of thing I'd be doing for a living. The reality is actually pretty disappointing. I _wish_ anybody still did this stuff.

I wrote something similar in the 1990s. Except, I had to write the TUI library myself (in x86 assembler) since nothing like what I needed existed at the time.

I was working on a project to do ECU performance curve remapping for a rally driver friend, so mine had additional features like the ability to export memory segments as .m files for plotting curves in Matlab.

I watched a video about ECU remapping (on a modern processor) yesterday, and the guy started by using the OBD port to get access to the system. I had to physically desolder the EPROM from the board to dump it back in the day.

The kids today have it far too easy.

Answering the question: it would be called "the Erlang VM", and you'd use Elixir to program it.

https://elixir-lang.readthedocs.io/en/latest/mix_otp/10.html

Yeah but... who does that?

How is that relevant? we are talking about models, now what you do with them.

For text, have a big model generate the "intelligent" answer, and then ask a local LLM to rephrase.

Yeah exactly, you can always do that by using another model that doesn't have the watermark.

...and does no harm for unfixable bugs. It's the logical equivalent of "switch off and on again" that as we know fixes most issues by itself, but happening only on a part of your software deployment, so most of it will keep running.

I do the same thing - Instead of going first to an unknown site that might (will?) be ad-infested and possibly AI generated, so that a phrase becomes a 1000-word article, I read the comments on HN, decide if it's interesting enough to take the risk, and then click. If it's Medium or similar, I won't click.

Hey, coming out feels good - I thought I was the only one.

Instead of LLM, Python and whatnot, it could have been a cheatsheet: https://github.com/scottvr/wtffmpeg/blob/12767e7843b9fd481ba...

In the end, I'm not sure I get what this is for - the venturebeat piece seems written by an AI.

Raspberry pi with microphone and camera.

The only disruption here is the hyperbole.