Greg Fish

Greg Fish

@neutronfish.bsky.social

Irradiated Ukrainian-American computer lobotomist by day, writer by night. https://cyberpunksurvivalguide.com

518 followers73 following1.3K posts

Me: *looking at what ea/acc people are talking about these days when it comes to AI* My brain: “New word added to the dictionary! ‘Orgynizers!’” Am I doing computer science wrong? I feel like I’m doing computer science wrong.

I know that some people are cheering Delta for refusing to use Starlink and sticking it to Apartheid Elmo, but the reason has nothing to do with Ukraine or “doing the right thing.” It’s just a branding dispute. Bastian didn’t want to promote any brand other than his own. That’s it.

The math behind LLMs has been published since 2017. We have tutorials on how to make them from scratch. You can build one with Excel macros if you're really dedicated. So, why do AI startups keep pretending they summoned an uncontrollable cosmic horror?

cyberpunksurvivalguide.comwhy ai startups are claiming they lost control of their own botsAI companies decided that the best way to deal with their tools being abused and misused is to claim they're impossible to control and always have been.

CSG on the radio is a double feature this week. First, I lost a bet, had to write a post-apocalyptic sci-fi book, and started serializing it on the newsletter. Second, I talk about why we shouldn't be *that* mean to robots...

omny.fmNEW - Swearing at AI: Whose Behaviour Is the Real Concern? - Shane Hewitt and The NightshiftSwearing at AI can feel like a reasonable response to the sixth failed attempt. Greg Fish offers a less obvious way to judge demands for politeness: the behaviour worth worrying about might belong to ...

The most effective and reliable enablers of fascism and authoritarianism aren't vocal zealots, activists, and true believers. No, it's the mediocre, resentful, and amoral paper pushers in dead end jobs offered small raises and promotions, say scholars...

cyberpunksurvivalguide.comwhy mediocrity is fascism's best friendWe like to think that brutal dictatorships and fascist regimes need monsters. In reality, they just need a handful of resentful, amoral paper pushers.

It took Silicon Valley's top AI geniuses billions of dollars and two decades to... figure out a very basic fact about human intelligence taught in 200 level psych courses and known since the late 1980s. Shane Hewitt and I discuss...

omny.fmICYMI - AI: What If Feelings Are Part of the Intelligence? - Shane Hewitt and The NightshiftAI hype often sells a future free from the supposed weakness of human emotion. Greg Fish offers a reason to question that premise before accepting the next promise of a machine that thinks like us. Wh...

Our economy feels driven by scams and wild, unhinged hype, but that's only because that's exactly what's happening. Turns out, assuming stratospheric growth and one in a century revolutions should be a quarterly thing leads to some perverse incentives...

cyberpunksurvivalguide.come-con-omics: why everyone is lying to you and each otherThe reason why today's economy feels like an ongoing scam is because, well, it kind of, sort of is. But don't worry, it's perfectly legal. And very lucrative.

I'm still figuring out angles and lighting for longer videos, and this one felt natural to leave kind of stripped down to its basics. Also, my dog decided he was going to be in the frame the whole time.

My biggest problem with the Paperclip Maximizer thought experiment when talking about the dangers of AI isn’t the premise, it’s what happens when you try to come up with ideas to mitigate it with Less Wrongians and AI bros.

The whole discussion becomes a game of Calvinball in which the AI beats any attempt to program it correctly, update its bounds, or shut it off with unspecified super-intelligent magic we mere primitive hairless monkeys can’t even imagine. Even though we built the thing.

“What if we [ insert literally any forethought here ] to make sure it won’t do that?” Nope. It’ll develop the IQ of Newton, DeVinci, and Napoleon combined, and easily overcome any potential strategy in ways we will never understand. Stop trying to come up with solutions, it’ll never work.

If chatbots are on the verge of conquering the world and making every job obsolete, why are AI startups selling you access instead of just cashing in on the very literal trillions to be made? Are they allergic to money? Shane Hewitt and I discuss...

omny.fmIf AI Can Do Every Job, Why Sell It as a Subscription? - Shane Hewitt and The NightshiftGreg Fish has a test for separating what AI is actually good at from what it's being sold as, and it comes down to one sharp question he keeps asking companies that make the biggest claims: if it can ...

I am no fan of Meta. Few companies have played a bigger part in ruining social media and the web. But a lawsuit seeking to effectively shut down the world's biggest social platform "to protect the children" could set a dangerous precedent if it succeeds...

cyberpunksurvivalguide.comthe long shot scheme to rattle the world's biggest social platformKnives are out for Meta in court, with politicians and lawyers eager to send a message. But how likely is it to work? And is it even a good idea?

Critics of generative AI excel at summarizing the problems with the tech, and empathizing with they anxious public. They're not so great at offering solutions other than hope. Mostly because the dilemma is less about AI itself, and more about how our system works...

cyberpunksurvivalguide.comthe forward centaur's guide to complex problemsTech critics have a lot to say about how the generative bubble will burst. They're far more mum on what to do while we wait to see if it does.

I’m just wondering if there an astrophysicist or a cosmologist with the time and inclination to give me some advice on a little project? It’s nothing weird, I promise. Just some fun #science for the people. Okay, maybe it’s a little weird, but not in *that* way.

A popular joke on social media is that Gen Z act an awful lot like Boomers despite their distaste for their elders. But how could a generation raised on hyper-personalized algorithms act so much like a generation raised in a conformist monoculture? Believe it or not, there's a very simple answer...

cyberpunksurvivalguide.comthe me generations: rise of the digital boomersWhat happens when a generation grows up being served exclusively by recommendation algorithms by Big Tech.

Earth's orbit is becoming a free-for-all, with over 250,000 satellite launches being planned by the 2030s even as our space junk keeps piling up. And the potential consequences have space agencies worried about the future of astronomy and space flight...

cyberpunksurvivalguide.comhow to endanger the future of space flight for status and profitCEOs and space faring powers are treating low Earth orbit as their personal playgrounds, much to the horror of space agencies.

CEOs and executives are obsessed with AI to a degree that's unnerving even AI startups. But then again, their common personality traits make them perfect targets for an addiction to chatbots, even when it's detrimental to their long term goals...

cyberpunksurvivalguide.comwhy your boss is obsessed with a.i. past the point of sanityNot only is the C-suite not immune to AI psychosis, they seem to be primed to suffer the worst of it as their employees duck and cover.

Celebrating that Elon Musk is a trillionaire now is like throwing a party that a random porn star who doesn’t even know or care that you exist just shot his 1,000th scene.

Since COVID, an awful lot of Republicans decided to ignore modern science and medicine. A new study finds that refusing to see doctors, neglecting their advice, and embracing social media influencers instead, unsurprisingly, comes with a death toll...

cyberpunksurvivalguide.comhow the republican crusade on modern medicine is slowly killing themIt used to be crunchy vegan moms who disdained modern medical science. Now, it's MAGA diehards who are at war with their doctors to their own detriment.