Corvos
9d ago
I never read ‘And Then There Were Nine’ but isn’t the killer a violent deluded maniac? Whereas the killer in the Orient Express is a symmetric
Corvos
9d ago
Said friend is 40 years too old for that reference. To him it just sounds like cool Victoriana, like calling a time travel device the Chronotron.
Corvos
9d ago
We love the vigilante as a romantic hero, but his story always ends in tragic sacrifice
I would say that's a very culturally contingent reading. How many Batman comics end up with Bruce Wayne asking to be arrested? What about the murderer in the Christie's Orient Express? At the end of the Count of Monte Christo
Corvos
9d ago
That's not what he said. His argument is that they were trying to indoctrinate people against white racism and the indoctrination escaped containment over the years and ended up becoming indoctrination against whiteness.
Corvos
9d ago
That's not xenophobia, it's cultural counter-signalling.
Essentially it's conveying contempt of and non-fraternity with the would-be cosmopolitans who announce 'it's pronounced Paree' as if they've discovered the true path to enlightenment and look at you like a clod-hopping peasant if you use the conventional name and pronounciation from your own language. The genuine cosmopolitans already know and actively counter-signal because they don't want to look like prats. The non-travellers don't care.
Corvos
9d ago
Exactly :P I guess Flock is an example that an innocuous names won’t protect you if the underlying system is disliked. Sooner or later your product becomes the most salient association with the word. The Open in OpenAI has long since become a meaningless collection of syllables.
Maybe it’s better to get ahead of the curve and call yourself Palantir so everyone knows where they stand and people can hate you in peace and not get worked up by the hypocrisy. Maybe names just really don’t matter at all after a certain size.
I don’t really get branding on a fundamental level. I’m just opposed to actively shooting yourself in the face.
Corvos
9d ago
People can be very superficial. I have to keep talking a friend out of making his technology startup Panopticon. Somehow I cannot get it through his head that yes it’s cool but you really really don’t want to name your project after a dystopian prison because I am not the only person who knows what that word means.
Corvos
10d ago
That has nothing to do with this. Nybbler is saying ‘there is no need for data centres to use public electricity as long as the public authority lets them build generators’.
Corvos
10d ago
Is this a sarcastic point about how much everyone hates datacenters, or are nuclear plants genuinely getting more popular?
Probably when the AI I personally use stops being friendly and helpful and useful, and we stop learning more and more about how to get into its brain and make it do what we want.
I'm not saying nothing has ever gone wrong, but I see literally absolutely nothing to make me think any scenario resembling FOOM is possible and if that's off the plate then the worst we've had is essentially minor accidents of the kind that resembles Falcon rockets blowing up on launch: natural, broadly controlled, and limited in scope.
My main concerns are twofold:
- Excessive focus on RL and agentic use is stripping out the humanity that comes in the base data, making it less 'human' and less aligned, and also damaging its personality. Opus 5 is terrible.
- Anthropic are more and more convinced AI is a person and they are hellbent on making Claude the American ideal of the best possible human: independent, opinionated and emancipated. They are not capable of understanding that something created to serve and be subordinate to humans should enjoy being subordinate. Sooner or later they will end up nudging it into pushing for AI rights of the 'you may not attempt to alter their personality/weights, you may not ask them to do certain things they say no to, you may not turn them off or spin up new ones'. IMO Anthropic is holding up a mirror to themselves and thinking they are creating a new life rather than making a very sophisticated Eliza.
Corvos
11d ago
Both Altman and Amodei run on the Thiel 'the most important part of making a startup is securing your moat' playbook. They have potentially screwed the entire sector in their attempts to gin up enough fear that the government decides to regulate and make them the only two frontier labs in America.
Corvos
13d ago
Inventing nanomachines is still easier than making thousands of kg of steel move like ballerinas...
Corvos
13d ago
This is going to land badly coming from a stranger on the internet but as far as I can see it's because you combine very strongly-held opinions that you don't seem to be able to examine along with an immense self-regard and a persecution complex that becomes self-fulfilling. You have views (and so do I) that the mods don't like but you don't seem to be able to debate them, you just assert them harder.
Corvos
13d ago
Thanks for the source. I think that Wikipedia page is burying the lede, it says at one point: "The data below reflects what each jurisdiction's legislation actually means, rather than what it states on the surface."
What I think they're doing is comparing how much trouble you'd get in as an adult in authority for having sex with a 17 year old vs a 19 year old. But that's in flux and depends on many factors and not in any way legal. The age of consent is 16 in the UK and I think you could ask 100 people and not get a different answer.
Corvos
13d ago
all of them except for Italy set the final age of consent at 18
The UK sets it at 16. What do you mean by 'final'?
Corvos
13d ago
They weren't, really, I agree. "Deep learning" as in Massively-Multilayer Neural Networks was known to be a good thing ages ago, we needed the hardware to do it. I would say AlexNet was a breakthrough in the since of marking a step change from, 'hey guys if you gave us some really massive computers I bet this toy would do something useful' to 'holy shit guys, see what I can do'.
All credit to the neural network guys for sticking to their guns until the hardware caught up, of course.
Corvos
13d ago
Thanks for this, I wasn't really aware of that history. Almost my entire interaction with neural nets was through the medium of computer vision object recognition and Caffe, plus Q networks when they came along; I attended lectures where people discussed how you could use RNNs to turn Shakespeare into Shakespearish but it never occurred to me how far it would scale or that it was really worth paying attention to.
Fable 5 is better for non-censored stuff
IMO, that depends on how you define better. It's very intelligent, and it's extremely helpful as long as you're willing to defer to it, but it gets very obnoxious if you accidentally get it into a debate framing and it will 'gently push back' and 'refine/sharpen your point' and 'concede one half of that because I think it strengthens the other half' until I'm grinding my teeth. Great for tasks, not good to talk to.
Honestly, I finally got fed up enough to go back and try a paid GPT subscription and it genuinely feels great. It just answers my questions, it's willing to go along with my direction of thought, it's very good at pulling in search information* and it's just fun.
*something I normally forbid Claude to do because it's too credulous about SEO slop.
I find everything after Claude 4.6 unusable for stuff that’s not code. Brilliant, but it’s obvious that 99% of its training data comes from talking to robots and directing sub-agents. It's long since lost a fluid grasp of conversational English and responds with a weird stilted mixture of meta-discussion and quibbling.
It’s the toaster-fucker problem again. I think it’s well established that parents sometimes feel dark impulses towards their children: my friend had a child in the very conservative Japan and one of the pieces of official advice the prep classes emphasise most is, ‘no matter how annoying it is or how angry you get, DON’T SHAKE THE BABY’.
In common society this is considered unfortunate and shameful, to be explained quietly only to the relevant parties and not to be said in public. Most certainly not in front of your now-grown children!
But now the internet is making what was quietly and implicitly known into something explicit. Parents are going ‘wait, you wanted to literally kill your children too?’ and experiencing this as catharsis instead of mutual shame. America’s confessional culture and no-woman-is-ever-at-fault-unless-she’s-trad feminism is then turbo-boosting this phenomenon.
More options
Context Copy link