Ventisei racconti che fotografano la vita e le sue molteplici sfumature: è questa la caratteristica principale dell’opera di Isabel Pistore, Le ali dell’imperfezione. I colori della vita.. La nostr…
A fully automated, on demand, personalized con man, ready to lie to you about any topic you want doesn’t really seem like an ideal product. I don’t think that’s what the developers of these LLMs set out to make when they created them either. However, I’ve seen this behavior to a certain extent in every LLM I’ve interacted with. One of my favorite examples was a particularly small-parameter version of Llama (I believe it was Llama-3.1-8B) confidently insisting to me that Walt Disney invented the Matterhorn (like, the actual mountain) for Disneyland. Now, this is something along the lines of what people have been calling “hallucinations” in LLMs, but the fact that it would not admit that it was wrong when confronted and used confident language to try to convince me that it was right, is what pushes that particular case across the boundary to what I would call “con-behavior”. Assertiveness is not always a property of this behavior, though. Lately, OpenAI (and I’m sure other developers) have been training their LLMs to be more “agreeable” and to acquiesce to the user more often. This doesn’t eliminate this con-behavior, though. I’d like to show you another example of this con-behavior that is much more problematic.
Profitez des vidéos et de la musique que vous aimez, mettez en ligne des contenus originaux, et partagez-les avec vos amis, vos proches et le monde entier.
Profitez des vidéos et de la musique que vous aimez, mettez en ligne des contenus originaux, et partagez-les avec vos amis, vos proches et le monde entier.
Hanno una politica no-logs che è stata anche dimostrata dopo che il suo CEO è stato portato in un tribunale greco per poter analizzare dei log (tomsguide.com/computing/vpns/w…)
Non ha "brutti" precedenti e all'interno della pagina c'è già ControlD che è di loro proprietà
Evitano di pagare per finire nelle classifiche "delle migliori VPN" (windscribe.com/ethics)
Si può pagare con Monero e Bitcoin (questi due fattori per molti è una cosa positiva quando si parla di VPN)
Le critiche che vengono generalmente mosse a Windscribe sono che:
non è europea ma canadese
non hanno un sistema di numeri casuali per l'account come Mullvad o IVPN ma ci si può comunque iscrivere solo con account+password e basta
il loro marketing è più aggressivo di altri ma non troppo dissimile da quello di Proton secondo me. Permettono di usare la loro VPN gratuitamente ma con molti limiti tra cui il limite di utilizzo di GB
a questo proposito la cosa che mi è piaciuta di meno in assoluto è il fatto che se gli dai una mail ti permettono di navigare 10GB gratuiti anziché i 2GB predefiniti. La mail comunque non è obbligatoria né per l'utilizzo gratuito né per l'utilizzo a pagamento.
Per il resto però sinceramente non ci ho visto molti lati negativi, ho chiesto (e mi è stato dato) un account temporaneo per poterlo testare un po' e per ora non so trovando grossi punti deboli.
ci ho fatto un salto, ma non sembra esserci nessun free plan nella pagina Pricing ad oggi (11 Maggio 2025). Che l'abbiano discontinuato, o non desiderano che si sappia ? Mah
Windscribe’s free plan is the top tier free VPN on the market, with no ads and no user activity logs. Browse anonymously and securely without paying a cent.
Però, chattando col loro chatbot, scopro che forse .... `is it still there a free plan ?
7:35 PM
Garry
Yes, Windscribe offers a Free plan which provides you with 10GB of monthly VPN bandwidth if you confirm your email address. Without confirming an email, you'll have access to 2GB per month. Enjoy exploring our Free plan features! `
quindi sembra che dopotutto questo free plan esista ancora
The response to The Settlers has been humbling. But the ongoing displacement and intimidation of Palestinians is more severe than we could capture, says documentary presenter Louis Theroux
Chinese researchers have a new method to extract uranium from seawater twice as cheaply as previous technologies. Their success comes as China needs uranium to fuel its unprecedented nuclear expansion
Norway’s sovereign wealth fund, the world’s largest, said it had sold all of its shares in Israel’s Paz Retail and Energy PAZ.TA because the company owns and operates infrastructure supplying fuel to Israeli settlements in the occupied West Bank, Reuters reports.
Disclaimer: The article linked is from a single source with a single perspective. Make sure to cross-check information against multiple sources to get a comprehensive view on the situation.
Microsoft is rolling out a new OneDrive feature that synchronizes data from personal accounts with business accounts. The feature, officially called “Prompt to Add Personal Account to OneDrive Sync,” allows bypassing security policies. This could result in business data ending up in the wrong hands.
Disclaimer: The article linked is from a single source with a single perspective. Make sure to cross-check information against multiple sources to get a comprehensive view on the situation.
Free Free Palestine! A space for activists and Palestinians to connect and amplify their efforts for the liberation of Palestine. Promote Palestinian solidarity events, or document and share here.
MrBeast: "I decided to create something unique to help people who are who are struggling financially. All you have to do is register at " SLOTCAS . COM " and use the special key "JAX" to get 2500$
Gaza’s water “comes from wells with salt water unfit for consumption. They have water treatment plants, Israel should hit those plants. When the entire world says we have gone insane and this is a humanitarian disaster — we will say, it’s not an end, it’s a means.” That was the opinion of Giora Eiland, adviser to the defense minister and former head of the Israeli National Security Council, writing in Yedioth Ahronoth on October 9, 2023.
[1]Already by November, the U.N. Office of the High Commissioner for Human Rights warned that “around 70% of the population in Gaza is drinking salinised and contaminated water.”[2] By July 2024, Oxfam reported that “people in Gaza have had only 4.74 litres of water per person per day” since the start of Israel’s offensive,[3] well below the World Health Organization’s minimum needed for survival in a humanitarian emergency.
[4]Israel's control over life’s most essential resource did not arise overnight. After occupying Gaza and the West Bank in 1967, it quickly established a system to extract water for its own use while restricting Palestinian access. For decades, this policy forced the indigenous population into dependence and precarity, while Israeli settlements in the Occupied Palestinian Territories — illegal under international law[5] — enjoyed “privileged access to water.”
[6]Read the rest of the piece on SourcedPress. Every fact-checking thread is public, and every source document is provided.
On May 3, the United Nations Special Rapporteur for human rights in Palestinian territories, Francesca Albanese, called for the prosecution of leading European Union officials for complicity in the war crimes committed by Israel.
In a series of interviews and reports, Albanese accused in particular the European Commission President Ursula von der Leyen and EU High Representative for Foreign Affairs Kaja Kalla. Albanese condemned the EU leaders for aiding and abetting gross violations of international law through their unconditional support for Israel. “The fact that the two highest figures of the EU continue business-as-usual engagements with Israel is beyond deplorable,” Albanese stated. “Immunity cannot equate with impunity. They will have to be judged before history does.”
In an interview with The Intercept Albanese reiterated: “I’m not someone who says, ‘History will judge them’—they will have to be judged before then.”
On May 3 the United Nations Special Rapporteur for human rights in Palestinian territories, Francesca Albanese, called for the prosecution of leading European Union officials for complicity in the war crimes committed by Israel.
Police bodycam footage shows Newark Mayor Ras Baraka being arrested outside a detention centre in New Jersey, during a protest about US immigration. Officials claim he trespassed, but Baraka and witnesses say he complied with orders.
The Post has a good piece up about all the hidden ways the Trump White House is trying to break different parts of the government — through non-payment of…
Für mich ist Sean Penn der beste Schauspieler Hollywoods meiner Generation. Wir müssen ihn nicht mögen, um ihn zu respektieren. Dieser Film ist ein Schlüssel zu seinem Selbstverständnis als Chronist und Kritiker der Vereinigten Staaten. Und weil, unter anderen, der unvergessene James Gandolfini hier sein Partner ist, müssen Sie das wohl sehen. (ZDF, Neu)
Für mich ist Sean Penn der beste Schauspieler Hollywoods meiner Generation. Wir müssen ihn nicht mögen, um ihn zu respektieren. Dieser Film ist ein Schlüssel…
It seems to have plateaued and increasing more slowly.
Look like 1 year "growth then plateau", like 2021-01 to 2022-01. But 2022-05 to 2024-09 linear growth again. Analysis/forecast of human behaviour not easy.
Also combine data of different source not easy, please handle with care.
Good point, thanks. The way I modeled the adjustment was by assuming that most usage is captured by Statcounter but there's movement back and forth to a reservoir that flies under its radar, in bursts, with zero net movement in the long run. So I used a geometric mean of the source data scaled by the square root of their averaged ratio.
Major digital labor platforms, also known as gig companies, operating in the United States misclassify gig workers as independent contractors, denying them labor rights.
There's this weird phrase "agentic AI" which looks weid. But it's a sign for how bad the state of the bubble is: Software agents are a very old concept, they are simple systems provided with a target function to optimize that operate somewhat autonomously
There’s this weird phrase “agentic AI” which looks weid. But it’s a sign for how bad the state of the bubble is: Software agents are a very old concept, they are simple systems provided with a target function to optimize that operate somewhat autonomously. This fits in well with modern “AI” systems who can be used to either extract data to operate on or might provide a few newish modes of interacting with other systems. So why not call them “Agents” but use the construct “agentic AI”?
The “AI” bubble is in trouble. Investment in data centers goes down, there still is no realistic business case, most people outside of LinkedIn still don’t much appreciate the slopification of the world. The only thing keeping the hype alive is momentum. First you had LLMs/generative AI, then came “reasoning models” who also hallucinated a chain of thinking/reasoning that might look like something that could have lead to the output (it is important to realize that this is still just a random string, nothing that has anything to do with cognitive processes). There’s always the next thing behind the horizon.
Problem is that the next thing when it comes mostly is the same thing we have: Maybe better at a few things, worse at a few others, maybe a bit cheaper, maybe also more expensive. And often: A disappointment when compared to the massive promises that predated their actual release. To not have that disappointment settle in, you need a new thing: “We never cared about this model really, it’s just a thing, we are actually working on …”
We see that a lot. Fabrications/hallucinations have been “soon gone” for a while now because that’s how software works, right? Bugs get removed. (Problem is that hallucinations are a structural property of those systems, they can be removed just as well as human beings’ need to breathe.) So when you read “agentic AI” you have to read it as scaffolding.
The “agentic” part promises that something new will happen in “AI”. Something more than “we can now use those systems in automation pipelines a lot easier”. That’s just a bit of application and as soon as you are on the application level, you need to really talk about revenue and all those nasty things. But if “agentic AI” is “on the horizon” you can still keep the momentum going for a few more months, for another round of investments. “Agentic AI” will be deprecated in summer or early fall to coin a new term describing the same processes of automation.
The discourse on “AI” systems, chat bots, “assistants” and “research helpers” is defined by a lot of future promises. Those systems are disfunctional or at least not working great right now but there’s the promise of things getting better in the future.
Which is how we often perceive tech to work: Early versions might be a bit wonky, but there’s constant iteration and work going on to improve systems to be more capable, more robust and maybe even cheaper at some point.
The most pressing problem for many modern “AI” systems, especially the generative systems that are all the rage these days are so-called “hallucinations” which is a term describing when an AI system generates incorrect information. Think a research agent inventing a paper to quote from that doesn’t exist for example (Google’s AI assistant telling you to put glue on pizza is not a hallucination in that regard because that is just regurgitating information from Reddit that every toddler would recognize as a joke). Hallucinations are the big issue that many researchers are trying to address – which mixed results. Methods like RAG are shifting the probabilities a bit but are still not solving the problem: Hallucinations keep happening.
But I think that this discourse misses an important thing: Anything an LLM generates is a hallucination.
That doesn’t mean that everything LLMs generate is incorrect, far from it. What I am referencing is what hallucinations are actually defined as: A hallucination is a perception you have that is not connected to any actual stimulus. You hallucinate when you perceive something in the world that you have no sensor data for.
The term hallucination itself is an anthropomorphization of those statistical systems. They don’t “know”, or “think” or “lie” or do any such things. They iteratively calculate the most probable set of words and characters based on the original data. But if we look at how it is applied to “AI”s I think there is a big misunderstanding because it creates a difference between true and false statements that just isn’t there.
For humans we separate “real perceptions” from hallucinations by the link to sensor data/stimulants: If there is an actual stimulant of you feeling a touch it’s real, if you just think you are being touched, it’s a hallucination. But for LLMs that distinction is meaningless.
A line of text that is true has – for the LLM – absolutely no different quality than one that is false. There is no link to reality, no sensor data or anchoring, there’s just the data one was trained on (that also doesn’t necessarily have any connection to reality). If using the term hallucination is useful to describe LLM output it is to illustrate the quality of all output. Everything an LLM generates is a hallucination, some just might accidentally be true.
And in that understanding the terminology might actually be enlightening, might actually help people understand what those systems are doing and where it might be appropriate to use and – more importantly – where not.
100_kg_90_de_belin
in reply to miss_demeanour • • •