Grimblewald

Makes sense, like "knowing" what a something looks like from memory, and actually having a recent short term memory reference. One is an idea that you'll recognise the real deal from, the other, is close enough to reconstruct a passable version of the real deal from. The former, is all but useless for describing to another what that means or comparing things in isolation in a faithful way.

Have used 5.5, ran usual battery of tests, and the downhill death spiral in a science/chemistry context continues. Code wise, it is better than open weight in several ways, but you really have to work hard to fight it from creating such a heavily abstracted mess it becomes unmaintainable for humans. Requires too much handholding for me. It's scientific reasoning is more confident as well, which would be great if it were also more correct.

Open weight models are generally more humble when working at fronteir, seek human input more reliably on thi gs they have no buisness deciding on, and can be trusted to not fuck up codebase in some way. Also, the way claude talks by default and tends towards even with all the "use this standard or that" tricks, grinds my gears in a way that drive me postal.

So, all in all, the shit prose, the constant oversight robbing me of attention better spent elsewhere, and high cost, and seemingly correct and difficult to catch errors on first pass make avoidance/non-reliance a nobrainer.

When dumb models are right, they're right, because likely its a solved problem elsewhere, and when they're wrong its quite transparent. Claude and fable are extremely good at spinning resilliant bullshit that can take quite some mental labor to verify and disprove, usual problem is flawed or deeply i correct assumptions on top of which then relativly sound reasoning and evidence is layered, or otgerwise weird contradictory decisons made "because i thought it would be better this way" that waste 48 hours of gpu time on an invalid model because claude fucked the one small part entrusted to it, which was extremely well defined and reasoned/backed by literature and it just made an executive decision to deviate without consulting or warning about it.

so fuck claude.

If required, so far I'm not impressed with newest proprietary. i have a claude sub mostly rotting at this point, so i come back from time to time until the year long sub expires, but at this stage i will not renew and see no point in doing so.

debt means you owe money. There is no requirement you get anything in return for taking on that debt. The definition, since you wont look it up, is "Debt is an amount of money, goods, or services that one party (the debtor) owes to another party"

If I commit to buying some future goods, and sign that I will hand over that money, that's debt. It's money I legally owe someone else. Counting that as debt is totally fair, not counting it is shady as fuck.

If I rent a house, I receive no money, in fact I receive no tangible goods at all really. Just the right to occupy a space under some pre-agreed rules. Yet, I now have debt.

Cool article but i really hope we fix llm prose in some near future. How chatbots write is so grating.

what do you think debt is? what you've described as misleading, is debt by definition.

Good quality well organised and accessible knowledve on a staggering array of topics is at everyones fingertips via wikipedia. Yet, still, I frequently find myself discussing things in depth with friends and collegues that we could easily just lookup and confirm. We dont do this though, because in the process of trying to establish truth using only what is known to us as a group, often creates far more interesting lines of inquery and insight. These are frequently far more potent and useful than anything we can lookup and sometimes are novel enough we cannot find anything like it online. This can only happen because we all applied ourselves and the cognitive varient we embody.

So, the problem with AI is that it is a regression to some mean of minds + corporate bias introduced. It contributes at best one person worth of input variance and is wholly incapable of contributing efficiently or effectively. AI does not contribute effectively and units of insight offered per word read is abysmal for AI text, and so carries enormoys cognitive burden to refine the slop into the simole coherent point actually made, if there even is one, which you often don't know until the effort in refining has been expended. So, we introduce a lot of labor to discussuons by including AI when best case it may offer about as much variance a single person who is not willfully ignorant can offer.

So, I imagine I am not alone in this, and when I talk with humans, what I want is their mental model of the world to contribute, not some corporate mouthpiece i could query directly if that's what I wanted. AI is not exactly inaccessible.

So, i think most are hypervigilant in pointing out AI in their communities for a variety of reasons , but a common theme will be that those who use AI to do their thinking are a single entity in terms of what they actually add to the discussion, and spread over many people it is the age old spam problem, a single mind having disproportionate signal strength in a discussion where all voices should carry equal energy.

but also, throw a weird variation of the game at the computer and it falls apart. Need something novel? llm's wont cut it. Doubly so if what is being done goes against established conventions, even if those are demonstrably wrong, to keep an llm from undermini g everything you first need to show it all the proof and evidence before it can be trusted to help, which for a complex problem can eaisly be 30-50% of co text limits.

LLM's are good at solved problems applied to known or similar to known applications, otherwise they're not going to match humans.

Odd to see so many 0 karma accounts piling in to decry this as fake so quickly after it's posting. I think we're all expecting something like this, weather this post is true or not. It's pervasive in many adjacent items, like your goddamn TV listening to everything, to companies being caught out in the past for doing things like this constantly. It is their entire business model, their life-blood. They'll never stop spying and they'll never stop finding new ways to spy.

It should surprise absolutely no one that things like this are going on, weather or not this post is credible. I agree, provenance of this particular post is a touch dubious, but even less trustworthy are the comments here saying this is bullshit.

almost, it's just a stream of conciousness reply because effort spent on refining a comment must be matched by post effort or you'll burn out.

Depends on journal. Predatory and low tier journals that's true for, but at higher tiers it is far less common. You still get the occasional lazy reviewer, but typically they wont be asked back if the editor isn't lazy as well. The system certainly has its flaws though.

Retractions do happen as well, especially if it's clearly proven malpractice. I've been behind 2 retractions so far, but it has never (in my experience) happened at the "contact corresponding author and seek clarification" stage of escalation. Typically it requires escalation to the journal itself.

Interesting idea, but it reads like an LLM write up. For me to engage properly with your written work, I need to see human effort. There are too many half-baked ideas out there, and unless you, as the author, thought the idea is worth putting proper effort into communicating, why should I think of this anything other than something someone said in passing? One way you could communicate effort to me, and likely others, rather than this work looking like the output of some LLM, is to use graphs, schematics and figures to support claims. Metrics are extremely misleading, only really useful when we've verified what is being measured is what one assumes is being measured. Metrics on their own are often deeply bullshit, so without any kind of graphs of trends or other important qualities, I assume it's all LLM output, since LLM's will default to not plotting anything while humans tend to go the opposite way.

Also, more effort into completeness needs to go in. Perhaps it is in there, but you just report an "accuracy" without defining what that means. The more I look, the more it starts becoming apparent this likely hasn't received any real human over-sight at all. "What does Grassmann mixing actually buy us?" It buys nothing, because it is a concept, Claude.

Neat idea, but in my opinion devoid of critical thinking / analysis or the level of rigour required to be suitable for broader dissemination. I'd put effort into making this a more credible body of work ASAP. Slop like this WILL cripple your standing in the field.

not really, it would simply be the first of many headaches that just get worse as time goes on. Windows is a festering blight, it costs you more and more in bullshit you deal with the longer you have it.

For example, I cannot currentl use the ms suite, like word or excel, because it asks me to log into an institutional account that I dont have, despite my account being the owner of the licenses for the suite running on my laptop. Make that make sense. All because i opened an institutional word document on it, i've tried fully reinstalling, no dice. i've tried nuking caches and registry entries, no dice. Window's basic functionality is indistinguishable from maleware.

memeory bandwidth is typically the limiting factor, not compute. Your card quite likely has 2x or more the bandwidth.

Fable is generally massively overhyped IMO. Anthropic has decent offerings, but they're over priced and matched too closely by things magnitudes cheaper AND faster, and since anything interesting isnt gonna get bailed by AI first try, being able to test and review sooner means getting a finished toy sooner as well. Combine that with local models that dont shift underneath you in behaviour like anthropuc/oai do, they're actually gonna have a really hard time convincing anyone like me their service is worth it. Probably explains the fear mongering and push towards controlling local model development and dissemination.

This is it for me, it feels like someone developing alongside me, rather than something like Claude which is constantly fighting to degrade code to a point where it and only it knows what the fuck is going on anymore, which would be defensible if it was at least good code but it isn't.

on the flip side, some early adopters of the tech have been scaling back or removing. Which would both offset gains from new-comers as an increase and also reduces the willingness of people to take the plunge.

There's no doubt AI companies can be profitable and bring value, but anthropic and OAI are horrifically over-valued at the moment, and there is no credible read that currently signals engaging in trade of these companies stock is anything other than engaging in greater fool theory.

plenty of public data at this point. Right-wing politician / figurehead social media engagement has tanked, which is the most obvious and self-verifiable signal I can offer.

Here's a news article if you prefer it.

https://www.thelondoneconomic.com/politics/reform-figures-se...

Richard rice is an example given in this one, engagement from thousands of likes etc per post, to less than 100. That obviously isn't an isolated instance and converts to less insane comments in other areas of the internet as well. You suddenly have a hard time finding previously incredibly common bad-faith actors in most political discourse comment sections.

I agree propaganda bot traffic almost certainly doesn't come from datacenters but rather proxies in locations of interest, but those bots must live somewhere and that somewhere is these Yandex datacenters. That said, I agree with your core point, I also doubt cloudflare has any way of tracking these types of bot.

"The market can stay irrational longer than you can stay liquid"

jesus, at that level of increase I shudder to think how much their expenses had to rise.

Thats a bit if a silly analogy isnt it? It would be more correct to say it would be like the NSA/CIA operating out of google datacenters, which they do, extensively.

oh, but they did. Measurably. Comment sentiments shifted drammatically following the post knockout period, especially on topics where comments typically get flooded by what is commonly called a troll account. Take videos covering pro-choice content or other content critical of US right wing talking points, which draw such bots like flies to shit. Post knockout? Near nada on new videos. It went from a sizable chunk of comments (~13%<amount<27%), to a fraction of a single %. That's a non trivial shift. Still looking at data, but something certainly did happen, and I think the memes are likely grounded in truth.

not all 8B have computer access, and even less access the computer as a device they control. Our crowd is actually rather small.

right? "im having this niche problem i find no information on anywhere, any ideas?" and you get "ahh this is the classic <wrong, other niche problem>"

I don't think so, this argument stands in line eith the current stste of AI, if my statement of "not reliable or universally useful enough" stops being true, my claim loses its base, it stands fine on current evidence and is conditional on it, so it'll never look quaint.

I'm big on AI and have it deeply incorporated into every facet of my life but it _is_ an early tech with a lot of roygh edges, it is not ready to be forced into every aspect of public infrastructure and life yet. That's a simple fact of the tech's infancy.

There was a time saying forcing everyone to use the internet and closing all paper/phone lines for banking is a bad thing, would have been the right call. Now it's more defensible, still the wrong move for some in our social cohorts, but there is merit to that move now, However, at the internets infancy it was no different.

Unsure personally, but I dont really care either. Given current handling and behaviour fron the elite surrounding the files, I'm personally treating any accusation as true until thorough credible evidence emerges proving innocence.

There's a lot of power and resources going onto covering things up and undermining real and sound evidence reguarding a real and well documented criminal network that permeates into every level of our power structure. So, we cannot afford to dismiss anything, with or even without evidence, no? I know that's unusual, but guilty until proven innocent only works when the artibtration system itself isn't conpromised and is itself not party to guilt.

Given everything we've seen to date in relation to the epstine files, people killed to hush things, coverup efforts, especially the botched ones we have proof for, it does paint a pretty clear picture which supports murder of the whistleblower being a very credible accusation.

It is my position that if you hold any power or influence, you're immediatly guilty and complicit. Even if publically vocally against the powerstructures most heavily implicated. We know those guilty own or contorl all major platforms and they have a vested interest in only promiting / allowing those complicit to wield influence.

And yet, hand it simple task out of domain, it fails badly. If a system 4 years ago did all you describe without being trained to do those exact things with unfathomable volumes of data, i'd agree, but we don't have that. We have something that can replicate seen patterns effectively.

There is no reason to view the transformer vased LLM family as anythi g other than a translation machine. From the language of english spec sheet to python language script, for from image description to image, etc. The only real read we can take from current behaviour of llm's is that the relationship between domains we think of as seperate are likely joined more tightly than previously assumed.

there's a third option youre missing here that people are consistent on: AI shpuld be opt in and nothing prevents opt in being default. AI isnt reliable or universally useful enough to include it defacto.

A huge component of it is invested in things that depend on the AI bubble. Some 20% of it's total from memory.