I'm still waiting for an assembly language model tutorial, but apparently there are no real engineers out there anymore, only torch script kiddies /s

Automotive actually uses ML in plain c with some inline assembly sprinkled on top run run models in embedded devices.

It’s definitely out there and in productive use.

> ML in plain c

Which engines in particular? I never found especially flexible ones.

Ironically, slippery slope argumentation is a favourite style of kids.

Unfortunately, your argument is a well known fallacy and carries no weight.

It's a natural response for "No true Scotsman" in the parent.

Pfft. Assembly. I'm waiting for the real low level tutorial based on quantum electrodynamics.

Not really, right? There are a ton of special tokens, like start of sequence etc., so what happens if there are two start of sequences predicted? It's a valid token but cannot really be turned into something sensible, so it throws an error when converting tokens to plain text?

Special tokens are handled by the application, not the model. They are still output before then.

Things you can also do: nothing on this list. Seriously, a list like this often makes one feel like one is not getting most out of life, but realistically most these things are not for everyone. Good to be reminded about the agency of one's decisions, but you do you.

It reminds me that one cant get everything out of life ~ but it's worth trying!

That's the most reddit comment I have yet seen on HN

Came for the CAP theorem, stayed for the cardiology 101 course, lol. His approach to diagnosing himself is really admirable. Working with doctors but never fully trusting them w.r.t. completeness is unfortunately the way to go in these busy times, where doctors just assume the most likely diagnosis and move on. Luckily, we have almost the same information available as the doctors with studies and such that we can dig deeper ourselves.

I can't really blame them. The essence of it is "if you see hooves think horses not zebras".

The human body is usualy one resilient bastard, so the ol' "tylenol and come back if it gets worse" is a good tactic for 95% of the cases.

But detecting those 5% cases probably takes A LOT of dedicated bandwidth from the whole medical process, so unless there is some clear manifestation that it's not horses but something else, I think it is reasonable that some will not investigate further.

Looks awesome, quite fitting to my text labeling needs actually. Might try it out at some point for my next prototype