this post was submitted on 01 Apr 2024
190 points (98.0% liked)

Futurology

1814 readers
27 users here now

founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
[–] pennomi@lemmy.world 55 points 8 months ago (4 children)

There’s already more than enough training data out there. The important thing that remains is to filter it so it doesn’t also include humanity’s stupidest data.

That and make the algorithms smarter so they are resistant to hallucination and misinformation - that’s not a data problem, it’s an architecture problem.

[–] FaceDeer@fedia.io 19 points 8 months ago

Stupid data can be useful for training as a negative example. Image generators use negative prompts to good effect.

[–] MotoAsh@lemmy.world 9 points 8 months ago (1 children)

Butbutbut my ignorant racism is the truth!! That's why I hear it from everyone, including [insert near by relatives here]!!

[–] Takumidesh@lemmy.world 3 points 8 months ago (1 children)

Well is the goal truth? Or a simulacrum of a human?

[–] MotoAsh@lemmy.world 2 points 8 months ago* (last edited 8 months ago)

Considering not even all humans are hireable, I'd say only a fool aims for a simulacrum.

[–] CanadaPlus@lemmy.sdf.org 4 points 8 months ago* (last edited 8 months ago)

Well, it's established wisdom that the dataset size needs to scale with the number of model parameters. Quadratically, IIRC. If you don't have that much data the training basically won't work; it will overfit or just not progress.

[–] Ultraviolet@lemmy.world 4 points 8 months ago

You also have to filter out the AI generated garbage that is rapidly becoming a majority of content on the internet.