There are a few things I don’t like about this scoring system :

• Why is there a “Top Provider Content Share” metric if its gonna score the same as the “Top Provider User Share” every time ?
• Why is the Top Provider Content Share not higher than the user share ? For instance, emails usually have at least one sender and one recipient, making it twice as likely that at least one of them is using gmail. If an email has 10 recipients across 10 different providers, each provider has a copy of the data
• Why is ease of hosting a mail server rated so well ? How is “leveraging email hosting services” decentralized in any way ?
• Why are we using a random repo created a few hours ago by a random github user as a reference ?

There are a few things I don’t like about this scoring system :

• Why is there a “Top Provider Content Share” metric if its gonna score the same as the “Top Provider User Share” every time ?
• Why is the Top Provider Content Share not higher than the user share ? For instance, emails usually have at least one sender and one recipient, making it twice as likely that at least one of them is using gmail. If an email has 10 recipients across 10 different providers, each provider has a copy of the data
• Why is ease of hosting a mail server rated so well ? How is “leveraging email hosting services” decentralized in any way ?
• Why are we using a random repo created a few hours ago by a random github user as a reference ?

Self hosting emails is a pain, but I’ve been doing it for almost 2 years and I do not have any of these issues. I’m not an expert either, I just thoroughly followed a tutorial to properly configure dmarc, dkim and everything else and everything just works (I just hope I’m not jinxing it by writing this :D )

There are a few things I don’t like about this scoring system :

• Why is there a “Top Provider Content Share” metric if its gonna score the same as the “Top Provider User Share” every time ?
• Why is the Top Provider Content Share not higher than the user share ? For instance, emails usually have at least one sender and one recipient, making it twice as likely that at least one of them is using gmail. If an email has 10 recipients across 10 different providers, each provider has a copy of the data
• Why is ease of hosting a mail server rated so well ? How is “leveraging email hosting services” decentralized in any way ?
• Why are we using a random repo created a few hours ago by a random github user as a reference ?

You’ve probably read about language model AIs basically being uncontrollable black boxes even to the very people who invented them.

When OpenAI wants to restrict ChatGPT from saying some stuff, they can fine tune the model to reduce the likelihood that it will output forbidden words or sentences, but this does not offer any guarantee that the model will actually stop saying forbidden things.

The only way of actually preventing such an agent from saying something is to check the output after it is generated, and not send it to the user if it triggers a content filter.

My point is that AI researchers found a way to simulate some kind of artificial brains, from which some “intelligence” emerges in a way that these same researchers are far from deeply understanding.

If we live in a simulation, my guess is that life was not manually designed by the simulation’s creators, but rather that it emerged from the simulation’s rules (what we Sims call physics), just like people studying the origins of life mostly hypothesize. If this is the case, the creators are probably as clueless about the inner details of our consciousness as we are about the inner details of LLMs

Migrating all my IPv4 stuff (firewalls, VPN, routing tables, etc) to IPv6 is probably the one thing I’ve procrastinated for the most time in my life :/

Something’s odd with the numbers from fediverse observer. Numbers shown in monthly graphs should be about 30 times higher than numbers shown in daily graphs, but they are about the same

I did not read the link, but two of my biggest concerns do not appear in the summary you provided :

• the burden of hosting an ActivityPub enabled service is often duplicated for each instance instead of being split between them (for example, my Lemmy instance has a large picture folder and database because it is replicating all posts from communities I’m subscribed to)
• it’s a privacy nightmare. All instance admins now have as much spying power as the single centralized service it is replacing

(Edit: typo)

Sur mon instance (dans l’appli et version web) ça n’a pas l’air filtré. Tu es sur que ça ne vient pas de ton instance ?

Ce n’est pas mon analyse, les blocs de citations que j’ai postés sont directement tirés de l’article (d’où mon autre commentaire qui dit que l’article est intéressant malgré le choix de son titre)

Merci pour la précision. L’article est pas inintéressant non plus. Je trouve juste dingue cette manière de titrer presque a contre courant du fond de l’article

Hum, je crois que tu t’es trompé de post ? :)

le SMB a progressé de 0,8 % et même de 1,4 % pour les ouvriers mieux lotis grâce à la revalorisation automatique du SMIC

Mince alors. Si j’avais su j’aurais fait ouvrier pour être mieux loti que les autres avec mon smic ! C’est à la limite du supportable ce type de tournure…

Les salaires progressent toujours plus vite que les prix, mais les augmentations ralentissent

Je trouve ce titre au mieux imprécis, au pire trompeur… Dans tous les cas c’est tres peu clair sur la nature du phénomène qu’on décrit. Personellement, je comprend en lisant ce titre un truc du style “depuis toujours, les salaires augmentent plus vite que l’inflation”, alors que d’après le contenu de l’article, ça fait seulement depuis 2 ou 3 trimestres :

« Les salaires ont réagi avec retard à l’inflation. En revanche, ils ralentissent presque en même temps que les hausses de prix, ce qui nous a beaucoup surpris », reconnaît Dorian Roucher, chef du département Conjoncture à l’Insee.

Alors que les ménages n’ont guère pu compter sur leurs salaires pour limiter leurs pertes de pouvoir d’achat face au choc inflationniste de ces deux dernières années, les gains qui se profilent en 2024 risquent de décevoir. Dans ses prévisions, l’Institut de la statistique attend une hausse de 2,9 % du SMB pour l’année en cours après un bond de 4,3 % en 2023. Les salaires réels augmenteraient « modestement » en 2024, de 0,6 % selon l’Insee. « La dynamique des salaires cette année ne compenserait donc pas les pertes cumulées par les salariés en 2022 et 2023 qui ont atteint 2,5 % », souligne Dorian Roucher.

On est quand même très loin de l’optimisme que suggère le titre de l’article !

never stopped POSTing, even though I configured nginx to always respond 403 to anything from them for about a year now.

Lol, there are definitely some stubborn user agents out there. I’ve been serving 418 to a bunch of SEO crawlers - with fail2ban configured to drop all packets from their IPs/CIDR ranges after some attemps - for a few months now. They keep coming at the same rate as soon as they get unbanned. I guess they keep sending requests into the void for the whole ban duration.

Using 418 for undesirable requests instead of a more common status code (such as 403) lets me easily filter these blocks in fail2ban, which can help weed out a lot of noise in server logs.

Your sensitive data and logins are tied to email addresses, which are tied to domains. Lose your domain, someone can access everything.

I recently stumbled upon an article showing how bad this can be when the expired domains were used for important/serious stuff

I think they do get marked as dead after the Bodis subdomain does not act as a Lemmy instance. But I was wondering if a large number of instances “waking up from the dead” and acting maliciously could cause some trouble. Or would such “undead” instances pose no more threat to the fediverse than the same number of newly created malicious instances ? I’m mainly thinking about stuff like being in a privileged position to DoS most instances at once, or impersonation of accounts that used to actually exist on these “undead” instances

The closing parenthesis got caught into the link (at least with my client), turning it into a 404. You should add a space

I don’t game that much on pc anymore, but this reminded me of this post about Linux gamers providing good bug reports.

Sur le sujet de ce qu’est capable de faire un LLM “juste en donnant des prédictions pour le mot suivant”, je te recommande une autre vidéo récente de Mr Phi sur le fait de faire jouer ChatGPT aux échecs. C’est assez impressionnant, et un poil flippant d’après moi.

Si tu veux creuser le sujet (et notamment sur ce que dit Yann Le Cun, je te recommande cette vidéo si tu comprends l’anglais

J’ai pas réussi à tout lire, la condescendance du truc m’est insupportable.

J’ai compris qu’il fallait rien esperer en voyant une citation de Raphaël Enthoven, j’ai abandonné 2 lignes plus bas