domain:rifters.com
This is awakening me to a sort of Gell-Mann amnesia effect: if the LLMs are this wrong and this stubborn in areas where I can test its output, where else is it wrong? Can I trust it in the rough analysis of a legal situation? In a summary of the literature on global warming? In pulling crime stats? I'm inclined to think it shouldn't be trusted for anything not either harmless or directly verifiable.
Angela Collier has a video about "vibe physics" that talks about this in some detail. In the section I linked to she discusses how crackpot physics emails have changes since the advent of LLMs. People will add caveats about how they talked to this or that LLM about their theory and the LLM told them it made sense. She'll point out in reply how LLMs will just agree with whatever you say and tend to make stuff up. And then the people sending the email will... agree with her! They'll talk about how the LLM made simple mistakes when talking about the physics the emailer does understand. But obviously once the discussion has gotten to physics the emailer doesn't understand the LLM is much more intelligent and accurate! It turns out having the kind of meta-cognition to think "If this process produced incorrect outcomes for things I do understand, maybe it will also produce incorrect outcomes for things I don't understand" is basically a fucking super power.
You can pry my em-dashes from my cold, dead hands. That people don't know how to use them properly only means we need to teach English composition better.
A much more frustrating element of SecureSignals' writing is that he will often make some passing mention of some supposed ironclad consensus that exists on this one niche topic, that requires no sourcing or validation (after all, it is the consensus!).
He will then of proceed to conspicuously deny universally agreed-upon facts.
I understand there's a culture of letting your kids do whatever they want. I see it every day. It doesn't make it right and more importantly doesn't transfer responsibility to someone else.
I don't find this convincing at all. The parents failing here are elder millennials with a smattering of young gen Xers: digital natives. They're being lazy and stupid, and they should know better after being on the Internet at similar ages as their kids.
Mostly monitoring the situation and fixing the odd bug that pops up. Things are pretty stable, so I've been plotting my next move.
Not gonna lie, the main reason this project got as far as it did is using the old nitter+miniflux setup daily, the thought of returning to Twitter, or even just plain old Nitter is unfathomable to me at this point, and it's been driving me to power through until I got something usable. While there's a whole bunch of basic features I should add to make it reasonably usable to anyone who isn't me, the thought of working on them instantly sends me into a comma. OTOH, I've found myself more and more frustrated with Substack, and thinking that it sure would be nice to be able to follow all the people from there on the same app I use to keep track of Twitter. Importing articles would be easy enough, but I think I'm more interested in the "Notes", but the prospect of integrating them was daunting. Cursory searches revealed no alternative Substack readers that I could raid for code, so I thought it would take forever to figure out how to deal with their API... until I actually looked at it. Turns out you can literally just fetch Notes with curl
, so I won't have to worry about reverse-engineering their authentication process, unless I'll want to implement fetching paid content.
How have you been doing @Southkraut?
If you put the sentence "Ohio law states that you may not loiter less than three get away from a pubic building." into Google docs, it will correct "get" to "feet", and "pubic" to "public". This has been the case for around 15 years.
I'm in much more agreement with "80% of work is done by 20% of employees" (although I think it's a larger share than 20%)
I disagreed strongly with "most office work is fake email jobs"
"Many employees don't work at full capacity" =/= "many employees have fake jobs"
The shorts would be against whatever companies you think are wasting large sums of money paying people to do nothing, as presumably they're very liable to be disrupted by companies with more competitive cost structures
Personally my view on the white collar replacement thing is more that AI can replicate most of the beige do-nothing-particular existing that defines a huge chunk of the white collar economy for far less per head. I don't think that AI's necessarily gonna launch a new frontier, but there's so much essentially complete dead-weight in the white collar economy and that might be a catalyst towards rationalization.
There was a local case recently where a 15 year old runaway became a Qanon cause celebre, with the parents and the local whackadoodles accusing random local families of being child traffickers and having kidnapped her.
She turned up a few weeks later in the deep south with a 20 year old boyfriend.
Obsession with some kind of Pennsylvania Pizzagate distracted from any actual effort of finding her, to say nothing of the harm done to those accused in public of child trafficking.
Israeli-born citizens should not be allowed to be attorney generals. Or they should recuse themselves in any case involving Israelis. Or Jews. Or they shouldn't be allowed to be government officials. Or lawyers. Or something.
If, hypothetically, 95% of Italian-American community supported the Mafia, would be Italian-American prosecutor or judge dealing with organized crime case seen as impartial?
You can get 10-20 tokens/s with CPU only inference as long as you have at least 32GB of RAM. You can offload some layers to your GPU and get probably 30-40 tokens/s? Of course, a 3090 gives you >100t/s but it’s still only $800, I’d consider that mid-range compared to a $2k+ 5090.
Swapping from the SSD is only necessary if you’re running huge 100B+ models without enough RAM.
But this isn't a criminal court, I do consider repeated claims, even by Epstein himself, of being related to intelligence to be significant evidence even if it wouldn't be allowed in a criminal court.
The alleged Acosta quote isn't merely hearsay. If an individual who heard the quote went on the record and said he personally heard Acosta say that, then it would be hearsay, and would be entitled to a certain amount of weight, less than if Acosta went on the record himself, but still a decent amount, regardless of admissibility in court. But that's not what we're talking about. We're talking about an unnamed "Senior Administration official" who told a journalist that Acosta said that, and we don't even know if the official in question even heard the quote themself or is merely repeating a rumor. That is, at minimum, double hearsay with an anonymous intermediary. It's the kind of thing that is only to be believed by someone who is already motivated to believe it.
Go back and read my writeup from a couple weeks ago on what actually happened in DOJ during the initial Epstein investigation, and explain to me how him being an intelligence asset or whatever fits in. At what point was Acosta told to "leave it alone"? How does a guilty plea involving jail time and sex offender registration equate to leaving it alone? Why were Epstein's attorneys so dissatisfied with the deal that they spent nearly a year trying to get out of it after it was signed? Why didn't senior DOJ officials in Washington side with Epstein when he referred the matter for departmental review? If Epstein had dirt and was pissed at the government for prosecuting him, why didn't he use it during the near decade between his release and rearrest, during which time he was the subject of numerous lawsuits?
There's an extensive record of the initial Epstein deal and if no one inserting wild conspiracy theories about Epstein getting off easy because he was a Mossad agent has done the basic work of familiarizing themself with that record. Instead they start from the premise that Epstein was involved in intelligence and work backward, ignoring anything that doesn't support their theory. Not doing so is like writing about European economic development in the second half of the 20th century without knowing about WWII.
After reading more of your thoughts I'm actually much more in agreement with them than how I interpreted them initially
There's a lot of space between "Ending the relationship" and "Don't even mention it."
Lodging formal complaints, and making public that they are doing so to assuage public concern, can lead to Israel telling its government officials do not diddle kids.
The truth always comes out in the end.
70+ years on, and UFO/UAP truth is nowhere to be seen. Maybe there is really nothing out there, except layers of psyops upon hoaxes upon scams upon bullshit.
Trust the plan.
Yet would [3] really be the motive with the highest good in a consequentialist sense? It may not be, insofar as the motive and conduct of Christ is for our imitation. Because if we believe that Christ’s guiding motivation was pure love for others, then we would likewise believe that our own guiding and primary motivation ought to be love for others. But here we may be wrong. Because Christ never says that love for others should be paramount, only that love for neighbor should be equivalent to the love we have for ourselves. The love for God is the paramount love, significantly greater than our love for neighbor, uniquely requiring “all your heart and all your soul and all your mind and all your strength”.
If Christ’s overarching moral motivation was to obey God for God’s glory, knowing that he will share in that glory and receive honor from God, then his motivation makes a lot more sense. Glory has been the motivation for all kinds of self-sacrificial acts throughout human history, whereas “love for humanity” is rare, if not nonexistent. (The man does not rush in to a burning home to save strangers because he loves humanity, but because he knows (from media) that this is glorious, and a glorious way to die). Additionally, the Epistles say that when we suffer morally, we should do so with glory in mind:
rejoice insofar as you share Christ’s sufferings, that you may also rejoice and be glad when his glory is revealed. If you are insulted for the name of Christ, you are blessed, because the Spirit of glory and of God rests upon you.
We do not suffer because “it’s right”, or love our enemy (though we ought to do so), but because we will feel glory. And to be Christians means to be —
heirs of God and co-heirs with Christ—if indeed we suffer with Him, so that we may also be glorified with Him. I consider that our present sufferings are not comparable to the glory that will be revealed in us.
The “social rewards” from God are intrinsically linked to moral conduct by Christ:
Blessed are you when people insult you, persecute you, and falsely say all kinds of evil against you because of Me. / Rejoice and be glad, because great is your reward in heaven, for in the same way they persecuted the prophets before you.
If Christ’s motivation was glory, both for his Father and for his divine family and for himself, then we would likewise imitate this, and this would lead to glorious moral acts. But if Christ’s motivation was pure and uncorrupted “love for humanity”, then we will only feel a gnawing discomfort at the impossibility of our ever replicating this motivation in any legitimate sense.
So what's your expectation of this new UAP hearing? Anything different from the previous nothingburgers?
But definitely knock it off with this "We all know" attempted consensus-building.
There's something kind of funny to both accusing @SecureSignals of engaging in consensus building when he says we all know what he means, and saying that we all know what he means.
If you think you’re being subsidised on a $20/month plan, switch to using the API and see the price difference. Keep in mind that providers make a profit on the API too - if you go on OpenRouter, random companies running Deepseek R1 offer tokens at a 7x cheaper rate than Claude Sonnet 4 despite Deepseek most likely being a large model.
As @RandomRanger said, it would make little sense for ALL companies to be directly subsidising users in terms of the actual cost of running the requests - inference is honestly cheaper than you think at scale. Now, many companies aren’t profitable in terms of revenue vs. R&D expenditure, but that’s a different problem with different causes, in part down to them not actually caring about efficiency and optimisation of training runs; who cares when you have billions in funding and can just buy more GPUs?
But the cat’s out of the bag and with all the open weight models out there, there’s no risk of the bigcos bumping up your $20/mo subscription to $2000/mo, unless the USD experiences hyperinflation at which point we’ll have other worries.
Some managers, sales reps, and HR workers come to mind (note that I'm not saying there's no need for those roles, but I get the impression there are far too many people in them). Heck, even many coders, despite having a real thing they make, are just skating by and not making a difference to anyone's life. I would possibly include myself in that. And I'm working for a successful company - I'm sure it's a dozen times worse in, say, the government, where even the distant hand of the market can't reach you.
I'm also open to the argument that 95% of jobs are useless but it's humanly impossible to know exactly which those are, so you need to keep everyone employed. I'm not arguing from omniscience here, just from my instincts after decades of code monkeying.
The truth always comes out in the end.
Mainstream science dismisses the concerns and sees the object as ordinary red colored D-type asteroid.
Mainstream science told you to mask up and get the covid vaccine too.
If a Scotsman finds out you knowingly transmitted human immuno-deficiency virus to them, they're sure to SHIV you.
Sure it is incrementally better on what we already have. The problem I'm trying to illuminate with it is that is the compute worth the provided value? It is hardly taking away a job from anyone doing the proof reading, it is an improved version of what we already have.
I wouldn't discount the entire title, but a bad PM fits the "fake email job"-shaped hole so well that it might as well be made for it.
More options
Context Copy link