I was going to spend a week without electric lights (plus no PC etc.). Partly for the romance of it, partly because I thought I might sleep better and be perkier if I let myself go with the natural day/night cycle.
I bought a lantern and some slender beeswax candles, and didn’t realise that this was good enough for mood lighting but not nearly enough for anything practical.
Corvos
4d ago
Are you an actual doctor? (I’m not.) I’ve found LLMs good at coming up with plausible hypotheses but bad at blocking them off.
Corvos
4d ago
Use an AI-integrated IDE like Cursor or Windsurf (now bought by OpenAI sigh).
Your query looks like ‘I have an error that look like paste text and I think it’s being caused by @Object1 not being destroyed properly during garbage collection’.
The IDE gives the codebase structure to the model, which queries the object you mentioned, its headers, etc. then does a search of the repo for where it’s used, then…
But I don’t think I’ve ever worked on a codebase that you would consider large and of course this only works for a monorepo.
Corvos
4d ago
Can you link? I enjoy that sort of thing; there was another couple who did the Victorian version with an icebox instead of a fridge etc.
I tried it myself once but it turned out that lighting even a small room with candles is surprisingly hard. You need a fairly serious candelabra if you want to be able to read a book after dark.
Corvos
5d ago
Thanks, that’s good to know! LG seems to be the best brand, although Sony have some nice ones.
IMO he’s saying that:
- IF there are important things genuinely worth spending money on (there always are).
- AND it is the case that the US needs to be careful about the budget (it does).
- THEN the Republican Party should make the case for spending on things they think are genuinely important, leaving it to the Democrats to urge caution and reduce the deficit because they know the Republicans won’t.
- CAUSING the Republicans to get at least some of what they want and become popular, while the Democrats have to sacrifice their own objectives and cut spending in a way that makes them unpopular.
In short, play the game of chicken to its end in an attempt to reverse the usual dynamic where Democrats make heartrending pleas and inspiring plans while Republicans explain why lots of things have to be cut and dodge rotten tomatoes.
@FCfromSCC is this a fair summary?
Corvos
7d ago
I think ‘buying trinkets’ is an okay way to say essentially ‘frivolous spending that has no long-term advantage’.
Since we’re talking about literal spending here, perhaps more context would be better in this case to be clear that you aren’t literally talking about buying physical things like jet planes.
Corvos
7d ago
Do people use the chat history / user memory features? I found them kind of intrusive and I prefer having a blank slate for queries, so I turned them off.
Corvos
9d ago
Ancient dwarven motto:
If your head is level with their navel, their groin is level with your teeth.
Corvos
9d ago
I don't literally mean 'want' as in literally 'will happily tell you that this is their intention'. I mean 'want' as in 'cannot be swayed from their path' i.e. they act as though they want to self destruct. Saying so is a defense mechanism, yes, but it's also that I have known such people and they will reject, subvert and oppose anything that will actually help them so actively that 'want' seems to be the correct word for it.
I do not think that ordinary people can help this subset of addicts and the mentally ill because that would require the power and authority to straight-up enforce 'help' on unwilling recipients, and in some cases it would take active mind-control.
Corvos
9d ago
@self_made_human, how are you getting on with the new PC?
I actually copied your specs, on the basis that you seemed like someone who knew what he was talking about, and I'm liking it very much but I need a proper monitor. You were going to get an OLED TV rather than a conventional monitor - did you? And if so, what do you think of it?
EDIT: apologies for the repost, I had the wrong Tinker Tuesday.
Corvos
9d ago
One think to look into is prefills - writing the first part of the AI's answer for them and then letting them 'continue' it. It's quite good for overcoming the more mind-killing varieties of fine-tuning that the big players use. Generally used for overcoming censorship but I think probably also good for directing approaches to problems, etc. For example, "Hmm, I should think about this very carefully, it's important I don't get it wrong" or "Oh, that's easy. Just...".
Corvos
9d ago
Partly I think female friendship is closer, more emotional and less contingent than male friendship. Partly of course grandchildren are family and that makes a big difference compared to tfwngf incelish guys.
Corvos
9d ago
nobody wants to self destruct
I think you’re wrong about this. Many people lean into their problems rather than out. Sometimes because - as cope - they convince themselves that they’re self destructing in order to live up to their ideals and then self-destruct harder to prove it. Sometimes because they sabotage themselves rather than risk failure with no excuse.
Corvos
11d ago
I would second The Colour Out of Space. Lovecraft himself considered it the finest of his works, and I think it's a purer example of purely Lovecraftian horror compared to some of his other works.
Corvos
12d ago
I read the three volumes of Worth the Candle that have been published on Kindle. It’s interesting but somewhat flawed - the main storyline doesn’t feel like it’s going anywhere, partly because
Corvos
14d ago
There's AI and there's AI. People detection is a simple matter which you can do on-chip. Anything like
Automated trucks for logistics, all coming from automated factories. That's all eminently possible with 70% unemployment, plus more exotic stuff like satellite swarms spying on everybody in real time, decapitation strikes with novel nerve agents we can't even detect.
in the next 5-10 years, like automated coding or automated logistics, is going to be heavily relying on a handful of APIs (approx 4 now) provided by a handful of companies. China could shut down LLMs in China tomorrow if it wanted to - firewall OAI and Anthropic, close down Deepseek. Boom, done. America would have a slightly harder time but it's basically straightforward.
Neither wants to, yet. But if the societal disruption starts to become uncomfortable, they can and they may well. I'm not talking about evil AI, I'm talking about obvious and destabilising social disruption. More than immense wealth and power, governments like stability. China and America are quite capable of running private military AI research on things like YOLO whilst mutually deciding that giving public/corporate society access to AGI is too disruptive to tolerate.
Corvos
15d ago
Also Battlestar Galactica. I think a number of events in the 00s combined to make it clear that we hadn't got answers to all of our problems - the 2008 financial crisis, 9/11 and the inability to turn a theocratic Islamic state into a liberal Western one. Environmentalism. These problems were huge but obviously totally unsolvable by ordinary people.
Also, people were bored. Nobody wanted to hear that we had solved everything and we just had to a) wait for laissez faire economic growth to solve all our problems and b) accept that anything which wasn't solving itself just had to be that way. They/we wanted change and adventure. I always think that was a big part of the response to Covid - people were longing for a Big Problem in which we could all Do Our Part.
Corvos
15d ago
I'm thinking of the
massive, gamechanging social and economic disruption
case, where say 70% of people become unemployed or suffer a sharp reduction in status. I don't like mass migration either, or the repeal of the death penalty, but the opposition to those is ~50% of the population max and most of those are pretty wishy washy about it. Governments hate disruption more than anyone, if too much happens too fast I can entirely see the government just bringing the hammer down, like China did with Ma. There's nothing technologically inevitable about cloud-based AI remaining available. And once it looks like one side of the China/America divide might start dialling this stuff down, I can well imagine their opposite number gratefully following suit.
In short, government with unanimous popular backing is still the biggest beast out there. IF it comes to the kind of unemployment figures above, I think AI companies will bend the knee or be broken. Obviously, if things remain as they are, the future is much more murky.
Corvos
15d ago
Depends heavily on your field. I made a minor move in my job a while ago: the old field fell under Computer Science and it was traditional for all research to be put on Arxiv, and the new field is almost the same but falls under Engineering and so everything good is gated behind IEEE. SciHub finds most papers before 2020 or so but not many after that and it's causing me serious trouble.
Corvos
15d ago
Presumably, if one wanted to, one could just firewall the main API servers. The big players are well known and with the possible exception of full-size DeepSeek, local models are not powerful enough to be very useful.
I’m not in favour of it, but I don’t think there’s anything stopping a majority voting for this. The only reason AI hasn’t been stomped on is the arms race and the fact that overwhelmed first world countries like the UK see it as the key to getting back in the black. Neither of this are immutable facts of the universe.
I remember (will never forget) that awful story about the tick.
More options
Context Copy link