Skip to content

Piero Bosio Social Web Site Personale Logo Fediverso

Social Forum federato con il resto del mondo. Non contano le istanze, contano le persone

Alleï, j’avais promis un thread de meme #IA #AI

Uncategorized
11 1 0

Gli ultimi otto messaggi ricevuti dalla Federazione
Post suggeriti
  • 0 Votes
    1 Posts
    0 Views
    Fine-tune an #LLM to output old bird names and it'll start behaving as if it’s the 19th century. Ask for a major recent invention, it'll reply the electrical telegraph! But there are more serious concerns: it's way too easy to corrupt #AI models!#AIsafety #AIresearch[2512.09742] Weird Generalization and Inductive Backdoors: New Ways to Corrupt LLMshttps://arxiv.org/abs/2512.09742
  • 0 Votes
    1 Posts
    5 Views
    The #ocaml debacle with the ginormous slop #AI / #LLM "contributions" continues! It's round two, already! Oh no. With the same outcome, predictably. This time in the #julia community.https://discourse.julialang.org/t/ai-generated-enhancements-and-examples-for-staticcompiler-jl/134054/18The patience of the Julia people is absolutely immense here! I've huge respect for them after reading this (and the Ocaml people, only days ago).The fundamental issue seems to be that the guy attempting this novel approach to software development thinks the AI "understands" what it's doing. So even if showstopping nonsense in the AI-produced code is pointed out to him, he comes back with - oh well, yeah, the AI gets "lazy", or some other sort of narrative that rationalises the slop, and presumes the "AI" is, by definition, really understanding what's going on, and smart, and capable, and all that.Will there be no end to this nonsense? Each week the patience of a new set of #FOSS maintainers will be tested? It's Groundhog Day, but with all the quirkiness and charm removed?
  • 0 Votes
    2 Posts
    10 Views
    @treyhunner I work for a big tech with a lot of LLM availability. It actually *has* helped my Python programming. We are also forced to use it for text generation for things like quarterly goals, employee reviews and more. It is useful but requires a critical user and is never a shortcut. That said, as a human, I value process over product, and I think LLM represents everything wrong about capitalism and how it regards human beings.
  • 0 Votes
    1 Posts
    15 Views
    Mi ero perso il numero 14 di Quid dedicato interamente a AI.Ho letto solo i primi tre articoli e per ora, mi sembra molto "solarpunk" per dirla informalmente... Oppure sono io a essere troppo critichi, chissà...Se vi interessa la trovare quihttps://quid.mensa.it/category/numeri/quid-14-intelligenza-artificiale/#ia #ai #rivista #mensaitalia #intelligenzaartificiale #llm