hypebot
boosted
The US military "misplaced" a fighter jet worth $100 million three years ago this month.
The Navy dropped another $60 million jet into the ocean last year.
Aircraft losses in the Iran War, from causes like friendly fire and crashes, has reached a toll of 42.
Worth remembering when the same government says it can't afford healthcare, housing, or college tuition.
2 media files
@jacobin Thing is, "afford" doesn't mean "afford." "Afford" means "prioritize."
der.hans
boosted
I'll be speaking about the Small Web at @SeaGL 2026.
It's at the UW HUB on Oct 23rd and 24th. All the coolest Free Software Fanatics in the PNW are going to be there.
This is the montage section in the film of my life.
der.hans
boosted
📢 Good news everyone!
All talk acceptance emails have been sent. Please check your inbox and spam folders and confirm your assistance so we can finalize the schedule.
Quixoticgeek
boosted
RE: https://hachyderm.io/@thomasfuchs/117207411820132499
Actually it's extremely fair to ostracize team members who support ethnic cleansing.
Ethnic cleansing is a form of genocide and as such is a crime, not an opinion.
der.hans
boosted
me talking about #codeberg a year ago: "folks, i get it that you're not happy with the downtimes, but please! we're a smol nonprofit that is always being compared to a billion dollar corporation"
some person talking to be about codeberg today: "i feel like codeberg is more reliable than github"
The Walking Bread
boosted
#2026PtoothfishBookReviews so i’ve backed to backed use of weapons and excession by iain m banks, culture books 3 and 5, skipping book 4 which is apparently mixed short stories ‘featuring the culture’, cos i don’t have a copy
i have little picture big picture emotional whiplash
but also yeah interesting times with the gang
i teared up at the excession’s report. rediculous. manipulation. cheap. dirty author tricks. i hope they do more of it.
2 media files
@mariusor @silverpill @smallcircles @evan It's not beneath Evan to also use ad-hominem "arguments" to avoid answers either. At this point the tensions are so high that cooperation is a) impossible anyway and b) unwanted by probably both parties.
@phnt @evan @mariusor @smallcircles No cooperation was possible from the beginning. The first thing that Evan and co tried to do when they came from Twitter in 2022 is to kill FEP process and SocialHub and set up their lame "alternatives".
Lisa Melton
boosted
Gloria Steinem’s death is an occasion to assess her legacy (not turn her into either a saint or a villain).
There is plenty to criticise, but dismissing everything she accomplished is just as intellectually lazy.
She was genuinely important to 2nd wave feminism. She brought women’s inequality into mainstream political debate, fought for reproductive freedom, workplace equality & the Equal Rights Amendment.
But her legacy also represents the limits of mainstream liberal feminism.
Poll:
I would like to know, if this symbol is well-known outside of Germany?
If you recognize its meaning, please don't reveal it in a comment, just vote "Yes", otherwise select "No".
And please vote only, if you live outside of Germany.
I'm asking for a friend, who told me, the symbol wouldn't be known in foreign coutries.
Poll:
I would like to know, if this symbol is well-known outside of Germany?
If you recognize its meaning, please don't reveal it in a comment, just vote "Yes", otherwise select "No".
And please vote only, if you live outside of Germany.
I'm asking for a friend, who told me, the symbol wouldn't be known in foreign coutries.
@fasnix Is there an option to vote “Yes, but you’re asking a population self-selected for disproportionate interest in semiological systems of that precise sort”?
What's this? Oh just Snickers using ClickFix and indirect prompt injection as part of a marketing campaign.
https://www.snickers[.]com/hungr-ai
https://www.snickers[.]com/digitalsnickers
The second URL contains prompt injections to change model behavior.
And this is the best-case scenario
2 media files
@mttaggart I want to see the prompt that created this. And if it threw any warnings or rejections for, you know, making malware.
It’s extremely irresponsible by Mars corp.
Background: I'm migrating my WordPress blog to BSSG. Write-up so far is in German, the current top post on the blog. Migration the post title image wasn't part of the export tool I used, so I am adding them back manually right now. So ... beware of typos ... but I double checked the path and name of the image, and as it renders on the single post page, it just might be a bug.
@schoedland that’s strange, but could definitely be a bug. I’ll have a look at it. Thank you!
@sebsauvage @artlog @bortzmeyer C'est un peu condescendant des jeunes, même si j'imagine que vous forcez le trait 😉 Dans les faits, la jeunesse n'est qu'un mot comme dirait Bourdieu. Mais au global, les jeunes ne sont ni natif·ves ni naïf·ves 😊 Ils sont juste pris dans les mêmes pièges que leurs aîné·es.
Nizar Kerkeni 🇹🇳 نزار القرقني
boosted
International activists are preparing to launch a new flotilla from Europe across the Mediterranean next month in an attempt to challenge Israel’s naval blockade of Gaza, the Freedom Flotilla Coalition (FFC) announced on 2 September.
According to the FFC, the vessels are scheduled to set sail in October, with planned stops at ports in Spain, Portugal, and Italy before proceeding toward the besieged Gaza Strip.
The mission will be the FFC’s first attempt to break the Gaza blockade since Israeli occupation authorities intercepted and abducted 430 activists, including doctors, journalists, and humanitarian workers aboard approximately 54 vessels in international waters in May.
Since then, FFC activists have traveled across Europe to build public support and raise funds for the upcoming October flotilla.
Don’t need a password manager if you just use the password “secret” everywhere!
@thomasfuchs Oh thanks a bunch, now i have to change mine!
@laguiri Indirect prompt injection occurs when the LLM references an external source—a website or file, usually—that contains text intended to provide further instructions to the model. Everything the LLM ingests is just context for its response, and there are no guarantees that it will recognize these external instructions as illegitimate. If the instructions are followed, the model can take action without the users' knowledge. This action can include:
- Exfiltration of information the model has about the user
- Changing output to misinform/deceive the user
- Executing malicious code on the user's system (malware)
In this case, the change in behavior is a stupid marketing gimmick. But where we go from here is clear. Imagine asking an LLM about recommendations for a purchase. The thing searches popular sites, except uh-oh, the sites have prompt injection instructions in them to alter the response to favor one brand over another. Now imagine an entire bidding ecosystem for that "ad space."
Now imagine most every ecommerce site you visit has LLM-poisoning instructions somewhere in it. Whenever a model accesses those services, they may be coerced into responses that are not in your best interest.
Your policy of not using them is prudent indeed.
@mttaggart @laguiri I'm curious as to what foundation model will fall for this or what Snickers used to test it.
