sweng

joined 1 year ago
[–] sweng@programming.dev 3 points 5 months ago

Yes, and what I'm saying is that it would be expensive compared to not having to do it.

[–] sweng@programming.dev 3 points 5 months ago* (last edited 5 months ago) (1 children)

Doing OCR in a very specific format, in a small specific area, using a set of only 9 characters, and having a list of all possible results, is not really the same problem at all.

[–] sweng@programming.dev 2 points 5 months ago* (last edited 5 months ago) (2 children)

How many billion times do you generally do that, and how is battery life after?

[–] sweng@programming.dev 6 points 5 months ago

Cryptographically signed documents and Matrix?

[–] sweng@programming.dev 1 points 5 months ago

I think you are replying to the wrong person?

I did not say it helps with accuracy. I did not say LLMs will get better. I did not even say we should use LLMs.

But even if I did, non of your points are relevant for the Firefox usecase.

[–] sweng@programming.dev 1 points 5 months ago (2 children)

Wikipedia is no less reliable than other content. There's even academic research about it (no, I will not dig for sources now, so feel free to not believe it). But factual correctness only matters for models that deal with facts: for e.g a translation model it does not matter.

Reddit has a massive amount of user-generated content it owns, e.g. comments. Again, the factual correctness only matters in some contexts, not all.

I'm not sure why you keep mentioning LLMs since that is not what is being discussed. Firefox has no plans to use some LLM to generate content where facts play an important role.

[–] sweng@programming.dev 7 points 5 months ago* (last edited 5 months ago) (9 children)

At horrendous expense, yes. Using it for OCR makes little sense. And compared to just sending the text directly, even OCR is expensive.

[–] sweng@programming.dev 3 points 5 months ago (11 children)

The issue is not sending, it is receiving. With a fax you need to do some OCR to extract the text, which you then can feed into e.g an AI.

[–] sweng@programming.dev 1 points 5 months ago* (last edited 5 months ago) (4 children)

What do you mean "full set if data"?

Obviously you can not train on 100% of material ever created, so you pick a subset. There is a a lot of permissively licensed content (e.g. Wikipedia) and content you can license (e.g. Reddit). While not sufficient for an advanced LLM, it certainly is for smaller models that do not need wide knowledge.

[–] sweng@programming.dev 8 points 5 months ago (1 children)

I'd say the main differences are at least

  • package availability
  • update frequency
  • backporting
  • packaging philosophy (e.g. plain upstream vs customizations, include all funtionality in single packege vs split out optional features)
  • default confguration for packages
[–] sweng@programming.dev 1 points 5 months ago* (last edited 5 months ago)

Feel free to assume that, but don't claim an assumption as a fact.

You recommended using native package managers. How many of them have been audited?

[–] sweng@programming.dev 1 points 5 months ago* (last edited 5 months ago) (2 children)

You know what else we shouldn't assume? That that it doesn't have a security feature. And we additionally then shouldn't go around posting that incorrect assumption as if it were a fact. You know, like you did.

view more: ‹ prev next ›