Jos autolla ei ole paljoa arvoa, niin suosittelen amisten autokorjaamoja. Työtä valvoo opettaja, ja laskutus on aina autodatan listahinnan mukaan, ja tuntitaksa on puolet korjaamoista.
Grumpy infosec guy who has been on the blue team since late 90s. Dabbles on the red side on occasion. Been working with various detection AI and threat detection and management automation since late 2000s
285 followers 665 following 1,843 posts
view profile on Bluesky
Jos autolla ei ole paljoa arvoa, niin suosittelen amisten autokorjaamoja. Työtä valvoo opettaja, ja laskutus on aina autodatan listahinnan mukaan, ja tuntitaksa on puolet korjaamoista.
Sigh, one of those people… There are a plenty of things to criticize about GenAI without conspiracy theories. And please use GenAI as a term, AI covers also many other technologies that aren’t wasteful or in any way involved with the issues LLM have. Blind hate sucks.
So? It was Adobe Firefly generated image made by wife. Firefly is fully trained with Adobes own stock photo collection, no plagiarism there, except fraudulent uploads that they didn’t catch.
That’s kinda off context, I guess every discussion can be turned into some kind of guilt trip about some atrocity. Buy that’s pretty useless for discussion. I am Finnish, back then my ancestors were being raided for slaves, by what is now Russia.
I meant, buildings that are still in active use. Yeah, 56 years is short, but doesn’t really negate the effect of nostalgia. Old enough to feel old unless you think about it.
TBH looks perfectly normal to me, a true case of someone looking smart until you hear them talk.
TBH there are quite a lot of people addicted to being angry and shouting. The fact that I mostly agree with them, doesn’t remove the feeling how unnerving that behavior is.
Well, back then he kept his mouth shut on anything but election related statistics. A true case of someone seeming smart until you listen them to talk.
Jepulis, tuli sama mieleen, on tämä arvoton näytelmä. Siitä on aikaa kun meillä oli viimeksi hallitus joka keskittyi ongelmiin, eikä poseeraamiseen ja sekoiluun.
Grifters gonna grift. I don’t mind if someone use GenAI instead of stock photo, as the result is what user wants, not what they could find but not exactly what they wanted. But selling “AI art” is plain wrong. Especially fraudulently like that.
Siis olisi kaikkien etu, jos voisimme priorisoida tulijoita jotka menestyvät täällä. Sekä Suomi, että tulijat voittaisivat. Mutta tuo mölinä vaan pahentaa kaikkea.
Varmaan tullut jo, media sitten kirjoittelee huomenna. Tekisi mieli sanoa, että jättää huomiotta, mutta se vaan venyttäisi Overtonin ikkunaa. Pirullinen ongelma, tuolla mölinällä estävät fiksun keskustelun maahanmuutosta, joka on varmaan se tarkoituskin.
TBH US is still a very young country. Europe is littered with buildings that are older than first European settlements in what is now US. So it looks a bit silly, at the same time, the constant change is why there’s nostalgia for things that have stayed constant at least for a little while.
What is going to slow down “robot uprising” is how difficult it is to properly automate something with LLMs, and how difficult it is to maintain. Slop is easy to produce, but slop is like an actor, anyone expecting an actor to be be able to do things shown in a movie, is going to be disappointed
TBH The inherent problems are not in LLMs or other technologies, it’s in people. It’s the hyper charged economy that drives to use technologies before they are fully understood. More harm has been done by bad applications of plain old statistics than LLMs.
Ok
Yes I see. Sometimes I go for cheap amusement. As we Finns say, a crazy person can find cheap entertainment. Hullulla on halvat huvit.
Yes, mere word LLM brings people out of the woodworks. To be honest I can’t wait for the hype to die down. When that happens also emotions will subside. LLM does have its use cases. But both boosters and haters are so wrong about most things.
I admit, I was trolling a bit. I wanted to demonstrate why there are no AI lists or starter packs. It is because of people addicted to shouting.
However the post I replied to was not about LLMs, people use any excuse to shout. Which also is why Bluesky has slowed its growth, this can be almost a hostile as Twitter where people migrated from.
You might want to check the other answers I got 😅 People are so blinded by rage that they blindly assume word AI always means LLMs.
And at which point do you realize that the comment I originally replied to was about everything else except LLMs?
Which is the whole point. GenAI haters either so addicted to hate, they don’t care to whom they shout. Or so clueless, that they assume word AI means only GenAi. Cybersecurity has been mostly applied AI long before LLMs became feasible.
AI bubble? We did first production AI deployment at 2006, it was naive baynesian classifier detecting Windows EXE malware in 2006. First project I was involved was in 2008, it was a heuristic rule engine classifying malware using sandbox detonation. That’s a long lived bubble.
And also underlines their point rather well. People have gotten addicted to hating and shouting, without even having a clue do they have right target or not.
Well, the post I replied to listed other AI techniques than LLMs. GenAI, only a small fraction of technologies available, and the least power efficient one at that. So, yes. People knee jerking on word AI, is rather weird.
This is why every company should maintain package repository proxy, which has sanity checks. Also protects from human typos.
Such lists would act as a lightning conductor for nutcases all riled up about LLMs.
Rather than just watch agent to do work, I have started lifting weights and other exercises. Better use for the short waits.
He is not putting any malicious code, I believe that. GRU however…
Tldr; don’t ever touch anything with a ten foot pole, if it implies sensitive data, and the service doesn’t have significant corporate customers.
Corporations can screw over consumers, however they do their diligence on corporate customers. Which also applies to kind of services that are used by paid customers and corporations alike.
Also, corporate licensing is a bit different beast for service providers. It would be the end of OpenAI or Antropic as a business, if it would turn out they use paid customers data for training, when that has been forbidden. Caution is fine, going full luddite is counterproductive.
Therre is no free lunch, of course if you use LLM for free, your input is training data But paid subscription allow to disable data sharing, at lease on other companies than X that is.
Heh. EDR, EPP and other security vendors have been using various AI technologies well over 10 years already.
Facebook pitiful excuse of moderation is such a minefield of forbidden words, that I use it only for pictures of pets.
Toki tietysti on pakko sanoa, että meidän ongelma on pääosin lääkärikoulutuksen kartelli. Koulutamme tosi vähän lääkäreitä per capita, ja tämä näkyy hinnoissa ja saatavuudessa.
Vähät välittävä
The later versions obviously had a lot better ventilation. Beta testing on customers is not a new thing 😅
Futuro is a part of Finnish pop culture there are articles occasionally, but couldn’t find it If I remember the door got stuck, and ventilation was insufficient, they had to take turns on breathing from the small ventilation hatch. Here are some more pictures. www.jalkipeli.net/futuro-talo-...
The problem with original bubble windows was that they couldn’t be opened at all. Which in times before ACs turned out to be a problem.
Macron shouldn’t insult ogres like that.
The original design was bad in details, for example windows couldn’t be opened and ventilation was bad. One family almost suffocated when their door got stuck. The architect was a visionary, not a competent engineer.
Shame that the page doesn’t contain anything about the history of the concept. Since when sales failed in Finland, the license was sold rather freely, so that company may as well have a legit license to use Futuro name. The same architect had also other cool concepts.
Nope, it’s firmly bolted down, better to learn breathe water.
Fun fact, that’s a Futuro house, it was Finnish concept for new style summer cottages, first was made in 1968. They weren’t a big commercial success.
Well, I don’t care where Trump goes, as long as he does. Heaven is fine by me.
Yeah, but that is the theoretical absolute best, reality would vary between 80 to 160 square meters depending on conditions, time of day, etc.
With full output and already at speed, a solar train could perhaps limp to a nearby station. But how often do blackouts happen in a middle of day in clear daylight?
Even that train wouldn’t ever be able to move fully on own solar power. There is not enough roof area per meter, even with best available solar cells, and theoretical maximum output. Which wont happen since flat is very unoptimal angle for solar cells. Best is around 45degrees.
40 square meters. Train car has at maximum 3,5 sqm per meter, at very unoptimal angle. So reality would be a bit more area needed, even during the brightest day sun going full blast.
From the article. 23% of the energy generated from solar panels on the train shed roof feeding the onboard batteries powers the train, along with power generated by panels on the roof of the train itself. The remaining 77% is fed into the grid to power the local community via a green energy prov
Best commercially available solar cells produce at maximum 250W/square meter. Train has 3,5sqm roof space when curve is accounted. Maximum power output would be at 875W per meter of train length. That’s not enough for electric car. And train roof is at very bad angle, thus less power in reality
this article exists because i realized it would be far easier and less eye-twitch inducing for me to just link to this article whenever someone asks me, for the thousandth time, if it's legal or a good idea to shoot at drones, jam them, or attack them with eagles
Because banking systems are extremely robust in account keeping. Any modifications are detected. Remember that the processes have developed over centuries to survive all kinds of malicious and deceitful actors.
Damn autocorrect, I meant local news.
He made minor news for using wok and oil to destroy USB sticks. But I can’t even find any articles online anymore. Yeah, so bad actors with good opsec rarely make into our knowledge.
Inverse survivorship bias, since those who succeed, and haven’t left online evidence enough to convict, don’t get above legal news. There was this one Russian cyber criminal who had reinforced door and panic trigger that destroyed all drives.
Of course Trump can’t be trusted, but better try to steer him, rather than let the Russians monopolize the helm.
I suspect after previous disaster, Zelensky has had a lot of mentoring on how to handle a person like Trump. Finnish president Stubb has probably been doing some private classes.
And it is getting progressively worse, I also have no clue why any advertiser is willing to risk their brand with Twitter.
If you dive into a cesspool, don’t complain about the smell.
Which means I have to build enclosure for the adapters that came with the cameras. Which is unnecessary hassle if they would use standard 12v, like any other camera.
For anyone using considering outdoor monitoring cameras, I would recommend avoiding TP-Link Tapo wifi cameras. The Tapo C500 outdoor cameras ship with indoor only 9v power adapters. Indoor only power adapters are typical. But aftermarket IP44 rated 9v adapters are impossible to find.
This means that EPP capabilities, that is preventing the infection in the first place, are becoming more relevant again.
Human/Brainrot.A
What has been working in other countries is to build enough city managed rental apartments. This helps to curb free market excess both by removing some of scarcity and acting as a reference point to make gouging too obvious.
It’s 2050 and a teen girl is torrenting a .tar.gz file of all the consciousnesses of all the tech bros who uploaded themselves into the cloud in a bid for immortality and modding them into The Sims 4
Niinpä, viittaan nyt takaisin siihen kuinka surkeasti epäonnistuimme kunniaväkivallan kanssa. Se, että ummistimme silmät koska se oli pääosin romanipiirien ongelma, ei ole mikään ylpeydenaihe.
Kysyn vaan ihan suoraan, että miksi antaa Persujen omia ilmiselvän lasten oikeuksia koskevan asian keppihevoseksi?
Siis, esimerkiksi kunniaväkivalta on asia jolle Suomi ummisti silmänsä, vaikka sitä oli täällä tiettyjen pitkään täällä olleiden vähemmistöjen joukossa. Eli siinä mielessä pidän keskustelua ennen varsinaista ongelmaa varsin terveenä ja toivottavana. Sitten tämä leireihin jakautuminen on eri juttu
Toisaalta, jos on jokin mahdollinen asia mitä ei pidä hyväksyttävänä, niin nyt on oikea aika tehdä päätös. Asioiden estäminen jälkikäteen on aina paljon työläämpää.
Denmark's new 'chat control' proposal is concerning. A risk-based system could force companies to scan private, encrypted messages for illegal content, setting a dangerous precedent for privacy and digital security. #cybersecurity fightchatcontrol.eu#overview
Sam Conman at this rate. OpenAI is doing great work, but I am starting to see why the board wanted to give Altman the boot.
Bloomberg - Tesla Dojo Went From Essential to Shut Down in 12 Months It spoke to a dramatic shift for a project that, at various points over the last six years, Musk and others billed as foundational. #Tesla /1
BlackHat that is, mobile keyboards are bane of my existence.
I am middle aged cis dude who is as white as one can be while still alive. And despite that “privilege”, I would be super hesitant to travel to US. About the only thing I miss there are the BlachHat and DefCon.
Considering that it would come with the people in there, I don’t know whether it would have been a good deal.
Sam is proceeding in his villain arch.
Jerry practicing advanced stay.
It’s not like facilities of that size could be hidden. Better to focus on active protection measures.
That would be a perfect way to thin out Russian AD, as they would be immediately replaced. Make AD in known positions the actual target!
If they wanted Putin dead, he would already be dead. But maybe they could whack some body doubles with a note “we know he’s a fake” just to up the paranoia.
They should chuck a drone or two to make sure AD stays there.
One classical example of said personality types is Benito Mussolini, who started as a socialist, and ended up a fascist.
But for some reason, it seems certain people take it as a personal offense. And rather than admit, yes there indeed are *those* people also among us, they go to ridiculous lengths to deny.
Yeah right. Absolutely and ironclad denial of horseshoe theory is a bit mind boggling. It is an observation of personality types, and how a certain type gets attracted to fringes, no matter which fringe. And due to their personality, they tend to behave same, no matter the cause.
I am rather convinced that The Onion is run by time travelers.
They always have.
These are kinds of people who should be recorded as a living proof of horseshoe theory being very much correct.
And in correct use GPT-5 rocks. The key is in controlling what documents it accesses from. For example a code assistants use in Cursor using GPT-5 is pretty damn impressive, when I first feed it with all our documentation and prior code. And yes we have enterprise license and control of data.
I'm going to get so blocked for this. I'm just irked by posts proudly demonstrating that LLMs are bad at tasks LLMs are a bad fit for. I have no idea if GPT-5 is any good, but here's a better benchmark: bsky.app/profile/anti...
Glad to see young hackers growing! Although vape hunt is something I can get behind, it isn’t nearly as safe as people think.
Tuotekehityksen kulut nollataan viimeistään konkurssissa.
En, mutta olen pyöritellyt malleja kotona, joten osaan laskea sähkönkulutuksen ja rautakulut. Jos opetaatiosta ei saa nykyhinnoilla kannattavaa, niin silloin on datakeskuksessa jotain pahasti vialla.
En oikein suostu uskomaan, että AWS tai Azure myisi tokeneita tappiolla. Joten tuotantokulu on jossain token hinnan alapuolella. Eli LLM ajo ei ole kallista, kehitys on.
This! I find it so annoying when progressives fall into that bullshit. Some things shouldn’t be culturally appropriated from magats and other wingnuts.
Do you know were the questions good for the exam? It’s one thing to use LLM and review, and totally another to serve unverified garbage from LLM.
Siis ei AI:n käyttäminen kallista ole. Ja nyt kehitetyt mallit eivät katoa minnekään. Oma veikkaukseni on, että uusien mallien kehitys hidastuu, ja huomio siirtyy nykyisten hyödyntämiseen.