Matias Koivurova

Matias Koivurova

@mkoivurova.bsky.social

Husband, father of two, and a smashing cat servant from 🇫🇮 PhD, Docent, Asst. Prof., working on light, relativity and quantum 🧪💡 Climate change is real Vaccines work Earth is not flat Trump is a dangerous dumbass Elon too

643 followers73 following1.5K posts

Matias Koivurova reposted

Prints available in my shop of this one from last year. The perfect Christmas gift for the philosophically-inclined office wall . Limited edition, numbered and signed giclee art print - link to shop in bio. #philosophy

Microsoft saying the quiet part out loud

KetAnthropic JoshAI@ketanjoshi.co

The first genuinely honest ad for generative AI I've ever seen, from Microsoft / CoPilot A mediocre white guy sees his non-white colleague getting praised for good work, he asks AI to *literally copy it*, and then he gets all the praise instead: www.ispot.tv/ad/SoDi/micr...

Microsoft: "are you a sad sack of shit who can't stand other people's success and who is completely uninterested in learning? Do you hate minorities and think their accomplishments should be yours instead? Boy, do we have a product for you!!"

Well damn. I wonder where that earlier graph came from?

Kate Starbird@katestarbird.bsky.social

Y’all, I hate to rain on the dead bot parade, but the Yandex data center destruction doesn’t seem to have significantly impacted social media bots. I made temporal graphs of X posts about AfD around the time of the attack and there’s no sign of a sudden dip or residual impact on that conversation.

This is what everybody with half a brain cell knew already knew, but proof is always nice: the rise of the far-right is entirely driven by Russians operating massive bot farms in Yandex data centers. Ukraine can go head and blow them all up!

Mark Chadbourn@chadbourn.bsky.social

Cloudflare Data Explorer shows the amount of russian bot traffic has dropped to 49% after Ukraine struck a yandex data center.

The thermal design of my work laptop isn't great. When I run a moderately heavy code that uses half of the cpu cores, the fan sounds like it's about to take off, and the laptop still feels hot. The biggest factor is probably that the outlet vent is behind the screen joint, which is... a choice

"Talous kasvaa..." ja karhu paskoo metsään. Suomen talous kasvaa koska koko Euroalueen talous kasvaa. Ja joo, ensimmäisellä vuosineljänneksellä nähtiin kasvupyrähdys, big whoop. Muistaako kukaan mitä muuta meidän rakas pääministerimme on luvannut? 🙃

Comet in Moomiland. It is unexpectedly inetense, there are giant carnivorous plants, a fierce monster, and a perilous journey to confirm that a comet is going to crash into Moominland, and there is nothing anyone can do about it. Real end of the world vibes 😅

Sleepy Hollow Reviews 🎃🐎@sleepingvillage.bsky.social

In honor of scary media month: what is the first piece of media you remember truly frightening you as a child?

Google: we'll invest 13 billion in Finland! Finnish government: yay! Finnish people: I'm not sure about this... G: it's for data centers! FG: yay! FP: wait G: 10 billion goes abroad! FP: hold on G: we need land! FG: yay! FP: but G: forest? Take it down! FG: yay! FP: you fuckers

View full thread

Google: we'll invest 13 billion in Finland! Finnish government: yay! Finnish people: I'm not sure about this... G: it's for data centers! FG: yay! FP: wait G: 10 billion goes abroad! FP: hold on G: we need land! FG: yay! FP: but G: forest? Take it down! FG: yay! FP: you fuckers

The reason Google chose Finland for the investment was because the bureucracy surrounding data centers is less strict in Finland than it is in the US. Which sounds really baaaad 😳

En voi sille mitään, mutta musta on vähän vistoa kun Elastinen (45 v) laulaa Sara Beelle (27 v, mutta kuulostaa yläasteelaiselta) että: "tää tyttö täs nyt kiinnostaa" Major setämies vibat 😬 #musataivas

Was out protesting in Tampere city center against the governments plans to allow mining under a fragile ecosystem called Viiankiaapa. It was cloudy, wet, and cold, but the turnout was still pretty good!

I heard that the cognitive reflection test is pretty hard, even though it only has three questions. It took me less than 30 seconds to solve them. Do people REALLY struggle with those? They are essentially reading comprehension tests!

Okay, so the whole "there is a good chance that AI will kill all humans within a decade" thing that has been going around lately? It's dumb. It's bullshit. It's a sign that the AI companies are in deep shit. But not because AI could kill us all. It's because they hit a wall called reality.

View full thread

The idea that "AI could kill us all, please make us stop!" is purely for the investors. They are trying to spin the idea that "we still have a ton of room for improvement, but now we really need to be careful because apocalypse. Keep the money coming so we can keep ourselves from killing everyone!"

Nut it is all bullshit. AI does not make it even into the top ten of my existential threats list. I still think that there is a good chance that the AI bubble will burst within the year. And when it does, it will be a bad time for us all. Financially, not existentially.

Okay, so the whole "there is a good chance that AI will kill all humans within a decade" thing that has been going around lately? It's dumb. It's bullshit. It's a sign that the AI companies are in deep shit. But not because AI could kill us all. It's because they hit a wall called reality.

View full thread

Then the AI companies thought "okay, then I will just optimize it! Get an army of coders to add exceptions and scale up the compute!" That is basically what they have been doing for the past year. And they are not seeing gains like they did with early scaling. They are freaking out.

Even with all this, LLMs are still not even close to AGI. And there is nothing more to squeeze out of the tech. This is it, the pinnacle of LLMs. Sure, it is impressive what you can do with a stochastic plagiarism machine. But it still does not *think*

Okay, so the whole "there is a good chance that AI will kill all humans within a decade" thing that has been going around lately? It's dumb. It's bullshit. It's a sign that the AI companies are in deep shit. But not because AI could kill us all. It's because they hit a wall called reality.

View full thread

"No worries" the AI companies thought, "We'll just use AI to generate new training data!" Smart! Except it doesn't work. All of them tried it, but none of them made it work. LLMs can only generate things that are in their training data. It doesn't add anything new, because it doesn't think.

But it gets worse! An LLM is based on probabilities. So on a fundamental level, it will always mess up at some point. It might answer the same question correctly N times, but on the N+1 try it may get it WILDLY wrong. So an LLM training an LLM is poison!

Okay, so the whole "there is a good chance that AI will kill all humans within a decade" thing that has been going around lately? It's dumb. It's bullshit. It's a sign that the AI companies are in deep shit. But not because AI could kill us all. It's because they hit a wall called reality.

So what happened is that AI companies saw that scaling the tech worked great early, and they thought they could scale it up ad infinitum. Now they have trained LLMs with essentially the entire body of literature human beings have ever produced. So no more scaling in human made training data.

I dunno mate, scientists have been warning that climate change will fuck shit up for a century now, and that burning fossil fuels is the main culprit. Have we stopped burning fossil fuels?

This post is unavailable.

Heard the news on the radio, and expected bsky to explode. But it's mostly quiet here? How are you guys not all yelling about Trump trying to buy off the elections??? And before you say "its the same old bs" no it is not. This is so much worse In any functional government, he would be removed asap

I was IQ tested back in the army to assess my "leadership potential" and I intentionally answered about half of the questions wrong for shits and giggles. My personal file probably still says that I have an IQ equivalent to a sponge

John Scalzi@scalzi.com

I was IQ tested when I was a kid and rode that high until Briggs-Myers tests came into vogue and then I reveled in being an INTP and now I recognize that both are basically the nerd equivalents of zodiac signs, but then that's a very Taurus thing to say, isn't it

Optimist: The cup is half full Pessimist: The cup is half empty Physicist: Assume a cup placed in a gaseous atmosphere, which contains a liquid held down by gravity. The fluid content of the glass is always 100%, since gases expand to fill their entire container and from a wave mechanical perspec-

Dr. Jessie Christiansen@aussiastronomer.bsky.social

Optimist: The cup is half full Pessimist: The cup is half empty Excel: The cup is January 2nd.

Had to take a mandatory AI course for work and it was... surprisingly good actually? No hype, just a clear high level explanation of what it is, what it does, and what it does not do.

Matias Koivurova reposted

chapter 0 of my comic MISERI MACHINA is now available to read for free on my site! it's been a labour of love that i've worked on for some time, so please check it out! // #MISERIMACHINA read it here!>> koyorin.com/miseri

The radio is playing in the background while we eat. Daughter (2 yo): is this baby shark? Me: no, this is Shakira D: who is Shakira? M: someone whose hips don't lie D: *puzzled silence*

jordan@jordantyranny.bsky.social

The only taxidermied blue whale in the world, 1865 CE, now housed at the Natural History Museum in Sweden. The public was allowed to enter the whale up until a couple was found having sex inside it. Now the whale’s jaws are opened only for special occasions, such as Swedish election days.