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

For the last hour something keeps causing WeChat to launch and demand I verify my age but I can't figure out what

Uncategorized
1 1 0

Gli ultimi otto messaggi ricevuti dalla Federazione
  • @aeva the whole linux mint team is big on NIH, they created their own GUI lib and redid standard linux apps lol so if anyone hates stuff not working it's prob them

    read more

  • @pupxel the version packaged in fedora is super buggy (the desktop background renders corrupted and sometimes just crashes) and there is no way to change the keyboard layout, BUT this is the first time full 10 finger multi touch has worked correctly in mollytime, so i may give it further consideration, assuming it working is not also some kind of ephemeral bug

    read more

  • Enigmistica (ebook)

    @libri - Gli enigmi visti in modo enciclopedico

    https://wp.me/p6hcSh-8Lg

    read more

  • @ericg congratulations! What happens now?

    read more

  • Assalto alle flotille: mobilitazione immediata in decine di cittaโ€™ italiane. la diretta.
    @anarchia
    Il secondo assalto israeliano in una settimana contro le Flotille dirette alle coste palestinesi di Gaza (dopo la Global Sumud Flotilla, รจ toccato a Freedom Flotilla Coalition e Thousand Madleens To Gaza) ha visto unโ€™immediata risposta di piazza e dal...

    read more

  • And there it is. I've retired.

    read more

  • Your LLM Wonโ€™t Stop Lying Any Time Soon

    Researchers call it โ€œhallucinationโ€; you might more accurately refer to it as confabulation, hornswaggle, hogwash, or just plain BS. Anyone who has used an LLM has encountered it; some people seem to find it behind every prompt, while others dismiss it as an occasional annoyance, but nobody claims it doesnโ€™t happen. A recent paper by researchers at OpenAI (PDF) tries to drill down a bit deeper into just why that happens, and if anything can be done.

    Spoiler alert: not really. Not unless we completely re-think the way weโ€™re training these models, anyway. The analogy used in the conclusion is to an undergraduate in an exam room. Every right answer is going to get a point, but wrong answers arenโ€™t penalizedโ€“ so why the heck not guess? You might not pass an exam that way going in blind, but if you have studied (i.e., sucked up the entire internet without permission for training data) then you might get a few extra points. For an LLMโ€™s training, like a studentโ€™s final grade, every point scored on the exam is a good point.

    The problem is that if you reward โ€œI donโ€™t knowโ€ in training, you may eventually produce a degenerate model that responds to every prompt with โ€œIDKโ€. Technically, thatโ€™s trueโ€“ the model is a stochastic mechanism; it doesnโ€™t โ€œknowโ€ anything. Itโ€™s also completely useless. Unlike some other studies, however, the authors do not conclude that so-called hallucinations are an inevitable result of the stochastic nature of LLMs.

    While that may be true, they point out itโ€™s only the case for โ€œbase modelsโ€โ€“ pure LLMs. If you wrap the LLM with a โ€œdumbโ€ program able to parse information into a calculator, for example, suddenly the blasted thing can pretend to count. (Thatโ€™s how undergrads do it these days, too.) You can also provide the LLM with a cheat-sheet of facts to reference instead of hallucinating; it sounds like whatโ€™s being proposed is a hybrid between an LLM and the sort of expert system you used to use Wolfram Alpha to access. (A combo weโ€™ve covered before.)

    In that case, however, some skeptics might wonder why bother with the LLM at all, if the knowledge in the expert system is โ€œgood enough.โ€ (Having seen one AI boom before, we can say with the judgement of history that the knowledge in an expert system isnโ€™t good enough often enough to make many viable products.)

    Unfortunately, that โ€œeasyโ€ solution runs back into the issue of grading: if you want your model to do well on the scoreboards and beat ChatGPT or DeepSeek at popular benchmarks, thereโ€™s a certain amount of โ€œteaching to the testโ€ involved, and a model that occasionally makes stuff up will apparently do better on the benchmarks than one that refuses to guess. The obvious solution, as the authors propose, is changing the benchmarks.

    If youโ€™re interested in AI (and who isnโ€™t, these days?), the paper makes an interesting, read. Interesting if, perhaps disheartening if you were hoping the LLMs would graduate from their eternal internship any time soon.

    Via ComputerWorld, by way of whereisyouredat.

    hackaday.com/2025/10/10/your-lโ€ฆ

    read more

  • ๐Ÿช’ ์•ผํฌ ์…ฐ์ด๋น™: ์ƒˆ๋กœ์šด ์˜คํ”ˆ์†Œ์Šค์˜ ์›๋™๋ ฅ
    ์ž‘์€ ๋ถˆํŽธํ•จ์—์„œ ์‹œ์ž‘๋œ ์ˆ˜๋งŽ์€ ํ”„๋กœ์ ํŠธ, ๊ทธ ์—ฐ๊ฒฐ์˜ ํž˜์€?

    ์•ผํฌ ์…ฐ์ด๋น™: ์ƒˆ๋กœ์šด ์˜คํ”ˆ์†Œ์Šค์˜ ์›๋™๋ ฅ
    ํ™๋ฏผํฌ (์ž์œ ยท์˜คํ”ˆ์†Œ์Šค ์†Œํ”„ํŠธ์›จ์–ด ๊ฐœ๋ฐœ์ž)

    ์—ฐ์‚ฌ https://2025.fossforall.org/speakers/
    ํ‹ฐ์ผ“ https://event-us.kr/fossforall/event/110400

    read more
Post suggeriti