YoungAchamian
We walk conditioned ground and name our folly civilization.
No bio...
User ID: 680
YoungAchamian
We walk conditioned ground and name our folly civilization.
14m ago
The core "worry" about RSI is that it improves in exponential time and thus cascades beyond human understanding or control. Researchers using AI to accelerate AI R&D is mostly "linear" because it requires the researchers to understand and often guide the AI in directions.
Because of this I don't think RSI is linear by definition. Human's centaur-ing is not "recursive" or "self", it's just improvement.
YoungAchamian
We walk conditioned ground and name our folly civilization.
1hr ago
it seems like the error rate would stack up faster
This is the general problem with synthetic data. The model has some error, it generates new data, with that error embedded in. That new data takes on larger and larger proportions of the data. It uses that to train, it generates new data and so on. At some point the data is distributionally different than the real data. It's a very concerning problem with systems that need to interact with the hard physical world. But I could see language being more "Fluid?" "Soft?", idk. It might come to pass that compounding synthetic data in NL isn't a solution problem but instead a human-computer interaction problem.
how an LLM could actually achieve stable RSI just based on the "recursive" and "self" parts of the acronym. If the model weights are being modified
The theory at least as I understand RSI, is that the model improves bits of the training process not writes its weights. But really stretching it, writing its own weights in a way to skips expensive training process and leads to a better model would be RSI. If it tweaked its weights during training. it could show better results on non-training tests. Assuming those tests are comprehensive, it could then go back and tweak its weights again, recursively. Essentially skipping the search for a better algorithmic solution or better data.
how do you keep the evaluation of improvement from drifting with successive iterations
Gut intuition interpretation of what you are saying is you are testing on a non-training validation. The evaluation is how you improve on the unseen data which is not given to the model to improve. But this whole paper was wonky af and it gets weird with non-RL connection. The idea I think is that you have AGIs train AI tools that don't require data/training pipelines and expensive gpus thus improving the capabilities of the AGIs to solve problems by giving them better tools (that they designed)
YoungAchamian
We walk conditioned ground and name our folly civilization.
1hr ago
You want a research paper
Considering we just had a paper on "AI's totally feel pain maaaan" It would take a bit more. But theoretically a demonstrated experiment of some ML/AI specific improvement to LLMs that show an appreciable improvement to the state of the art. That would be strong evidence of self-improvement.
now even solving problems which humans are incapable of
But how? Not by being smarter, but by brute force number crunching or human-AI teaming.
along a trajectory which has been predicted for years by the same people who also predicted RSI.
Uhhh aren't we already supposed to be dead by their predications? Or is it that we have 3 more years, or 10 more years? Idk there's been so much sliding around. Along with all these failed predictions about the method to AI improvement. Didn't they bet on GOFAI? It sounds like nobody holds them accountable for their failed predictions which they sweep under the rug.
RSI is already here.
Outrageous claims need extraordinary evidence. Prove it. Prove that the AI is currently improving without humans, self-recursively.
here are still humans in the loop, but the human contribution gets smaller as a percentage with each generation.
What is the AI doing in the loop, and what is the human doing in the loop? Loop-percentages are misleading if the human is doing most of the ideation, system design, and out-of-distribution thinking and the AI is doing the more rote or established needs in the project. I have not been impressed by Astra/Sol 6.1 on brainstorming solutions to hard engineering ideas.
The magical thinking is that AI progress will suddenly stop now
I fail to see where I said this.
YoungAchamian
We walk conditioned ground and name our folly civilization.
2hr ago
Us meaty hominids have been doing RSI via "evolution by natural selection" thing for the last million or so years: demonstrably results have been exponential,
Exponential over a million of years or so, not over a half a decade. Evolution is slow.
YoungAchamian
We walk conditioned ground and name our folly civilization.
2hr ago
competent AI junior researchers
The annoying part about juniors is that they mostly just implement what is already established.
Explosive increase in capabilities is totally plausible.
Sure, I BELIEVE RSI is plausible but I can also recognize that is a belief based on faith not evidence. I haven't really seen definitive evidence that an AI has, without human input, provided a better algorithm, loss function, learning process, a more efficient self-attention (not O(n^2)) than what humans, or human-AI hybrids have done. And there is a lot of money and hyper available for anyone to prove they can.
The latest one was that "AI writes NN weights" which turned out to be, like-aways, way overselling what it was doing. The model turned out not to be able to improve from that because the weights were very sparse conditionals built into the policy.
By what scientific or engineering evidence have you seen of it doing so? Lot of hypers think it's possible and soon but do they have any research to point to it being the case? The theorize it to be the case, but that's a theory on faith not evidence.
Keep in mind that the current "theory" on AI is "Moore's law for AI via Data". That is completely at odds with the RSI theory unless that RSI is creating additional synthetic data that actually beats the sim-data distribution problem.
YoungAchamian
We walk conditioned ground and name our folly civilization.
3hr ago
The big one. Recursive self-improvement leads to them dominating all industries.
I find investing on this weird. It's essentially investing in Theranos because they said they can tell diseases from a single drop of blood. They never proved they could, and none of the frontiers have even proven an inkling that RSI is possible, let along that it can be applied across a wide distribution of industries at a cheaper cost than skilled humans. It's investing in hype and dreams. I think any investment in frontier IPO should not depend on the speculation of what they could figure out but on the business case for what they can do and shown improvement in being able to do now.
YoungAchamian
We walk conditioned ground and name our folly civilization.
20hr ago
After talking with some folks here, I focused back on character design a bit more. Originally I was writing plot point scenes with an understanding of how they wove into the overall narrative and world building. But after realizing that I was kinda stunted on the characters voice's and internal motivation, because they aren't really fleshed out, I decided to pause on that part and push out the cast, figure out motivations, internal conflicts, and how they see the world.
Also names, which is slow going but I was able to bang out 3 more, only to realize that Roman women kept their maiden name, which I want to follow. Mostly because I dislike skin deep veneers of world buildings, and a roman-esque familia-based society would likely have married women keep the names of their clans/family not their husbands clan/family name.
YoungAchamian
We walk conditioned ground and name our folly civilization.
5d ago
Why do you assume, inclusively, that nobody here voted for democrats? Or that people here agree with your stance that voting for democrats means not having the best interests of America or it's constitution at heart?
I voted according to my desires on Tuesday, and achieved outcomes to my desires and in accordance with my values. They might not be your values, and they might not been your desires, but gee golly wiz thank god we aren't all sophists.
YoungAchamian
We walk conditioned ground and name our folly civilization.
5d ago
Realistically you can call it whatever you want.
From a technical standpoint it is the field of Artificial Intelligence(AI) The fields goal is to reach Artificial General Intelligence (AGI) and eventually Artificial Super Intelligence (ASI).
You may call it SI or SAI all you want but the only thing you are communicating is that you don’t understand the boundary of word semantics as it relates to technical function. If you wish to convey lack of competency in favor of political tribal signaling that is your prerogative but don’t expect anyone else to join you who isn’t already in the communist party.
YoungAchamian
We walk conditioned ground and name our folly civilization.
6d ago
Yes, it's super annoying. It essentially forced not only agencies but also contractors and companies that sell to the FedGov as well. Now all my middle managers insist on using that word and correcting me when I say AI. Makes me want to reach through zoom and choke them.
YoungAchamian
We walk conditioned ground and name our folly civilization.
6d ago
Yeah I've heard that. I've also heard to the effect of "Just set yourself a writing budget each week". My goal rn is 5k words a week. Looking to do a book in the ballpark of 240k words. Assuming a range of 2000-5000 words a chapter. The biggest problem is that I feel zero inspiration to write another book/novel. As I pretty much am only doing this because I feel inspired to, as a creative outlet, that makes it difficult to switch. I'm sort of aware that this first book is probably going to suck a lot. I've been calling it my coprolite novel.
I forgot that BSandy has lectures on narrative building, I'm curious how helpful they would be. I've read a metric fuckton of books in my genre, including most of the really high quality ones. It's not so much world building, plot, characters or systems that I'm worried about. It's the smaller, low level things: sentence structure, prose, word choice, dialogue. I might peruse his lecture series and see if he goes over any of that. Though in my opinion his prose and dialogue is pretty average.
Eragon got published because both his parents run a major publishing house. I don't think Christopher has written anything else, and the books are pretty juvenile. I don't think YA existed as a strong label at that point (or at least I hadn't heard of it) but they are strong examples of it. And it's a genre classification I really want to avoid. It has a very stable narrative structure and very predictable tropes, to the point that often world building is both the only unique thing, and simultaneously window-dressing.
YoungAchamian
We walk conditioned ground and name our folly civilization.
6d ago
Great pieces of advice thank you! I think the voice/Character creation first + speaking them out loud looks very helpful!
avoid two primary characters starting with the same letter (and especially starting and ending with the same letter!)
I’m fucked here though. I went with a roman republic setting and apparently roman naming conventions were such that it was fairly common for you to have the exact same name as your father, your sons, and sometimes even your cousins. Gaius Jullius Caesar had the same name as his father and the same name as his grandfather. There were also only a small handful of personal names even available.
Further more, personal names (the first one) are only used intimately so most people are going to call you by your cognomen (third name), which you practically share with all your male family members. Combo-ed with the fact that since a cognomen is what friends casually call you, it needs to be something that sounds like a first-ish name to the reader in my opinion.
YoungAchamian
We walk conditioned ground and name our folly civilization.
6d ago
I think this just Modbait/Trolling. Its blatant consensus building, it was posted in the midterms thread and for some reason is being reposted here. It comes from a fresh account made a couple hours ago. But in case you aren’t.
We must conclude that the people who disagree with us, those who voted opposite to our desires yesterday
You need to get your head out of your own ass. There is no WE. There is no OUR desires. Your conception of politics being such that only your policies are correct is very much the dysfunction affliction the republic. You ARE the mirror of democratic party thinking they needed to fuck democracy to save it.
YoungAchamian
We walk conditioned ground and name our folly civilization.
7d ago
Anyone here have any advise for writing a fantasy novel?
I've been inspired for some reason in ways I've never before been inspired to write. Originally as a "Fixing" of a novel I read with an interesting premise, but it's slowly been mutating as I branch more and more away from the original other than the beginning framework and aesthetic veneer (that I want to make deeper). I've written about 15k words so far over 6 chapters, sort of haphazardly strewn over the book. I have a general plot outline and a deeper epic fantasy authorial cannon + magic system that I've conceptualized. I even have a nice complex romance subplot sketched out with some "Sapphic yearning".
But I seem to suck at coming up with names, or fictional idioms. And apparently I suck even more at making engaging descriptive dialogue. I've been having my alpha-reader-slave (ie my GF) read my chapters and then quizzing her on them. The general pattern is that I'm actually ok when things are happening, but my dialogue is bland or non descriptive. Several male characters read through my voice instead of their own. Things like theming or subtle clues seem to alternate between too subtle and a bag of bricks.
YoungAchamian
We walk conditioned ground and name our folly civilization.
7d ago
It's interesting. What is the pure self? How much of interacting here acts as a good or bad abrasion to your understanding of your own identity? Metamorphosis is not good or bad, it merely is. TheMotte being the carpenter replacing planks on your ship of Theseus. It's possible that some rotten planks are replaced but also that weird ones are added. Is that a good thing? If you intent is maintain your self unchanged, any memetic mutation would be bad.
Idk I have weird, protean thoughts about this. Very unformed/stream of consciousness.
YoungAchamian
We walk conditioned ground and name our folly civilization.
7d ago
Way of Kings with the magic coming from Shards means that having such a powerful stone means you have magical powers.
Uhhhh... not to be pedantic but magic doesn't come from stones in WoK. Technically speaking it comes from trauma and oaths with eldritch entities.
YoungAchamian
We walk conditioned ground and name our folly civilization.
7d ago
I am lazy and not particularly skilled writer, and my opinions are substantially unpopular here that I have never had an AAQC and probably never will, even if I think I am occasionally manage a reasonable and correct point.
You should be careful. I thought the same way, and now for some reason I have 3 AAQCs, 2 of them for frustrated rants. I'm a bad writer, and I am also lazy. When I first started posting the mods told me I got a lot of reports because I apparently have unpopular positions. This place can metamorphize you merely from the memetic infection of reading and writing on it.
YoungAchamian
We walk conditioned ground and name our folly civilization.
8d ago
I know we have some home LLM enjoyers here who use LLM-agents for personal projects. @ThenElection I think you mentioned you did.
Have any of you tried these two papers on your home-swarms? LatCom & LatentMAS I'm actually very curious on their claims, I think these could be fascinating, but I'm still in the process of buying GPUs/parts for building out my own system.
YoungAchamian
We walk conditioned ground and name our folly civilization.
8d ago
Yes. It is technically an application of Dunyain philosophy applied in a Gothosian fashion. A statement of the human condition.
YoungAchamian
We walk conditioned ground and name our folly civilization.
8d ago
Trust me, I intend to where there is a Libertarian candidate. I'm actually overjoyed one is available to me on my ballot.
YoungAchamian
We walk conditioned ground and name our folly civilization.
9d ago
I can tell you that economically to me it is indeed a nothingburger. But politically? I will be voting democrat (I voted for Trump in 24) almost entirely because of it. The "America First" President, who ran on no more foreign adventures decided that the high from bombing brown people and getting his dick sucked by Mossad honeypots was so much better than actually fixing ANY of the myriad problems in America. The American system is unfortunately designed so that the only way I can meaningfully express my displeasure is to hold my nose and vote for the opposition party. Who actually knows if I am at all representative of the independent electorate but obviously others think so too.
YoungAchamian
We walk conditioned ground and name our folly civilization.
9d ago
Damn early morning brain fart strikes again.
YoungAchamian
We walk conditioned ground and name our folly civilization.
9d ago
It's really not a constitutional crisis. McConnell blocked Obama's SC nominees. The constitution didn't catch fire despite all the screaming the left did. Dems blocking Trumps Fed judge nominees is predictable escalation of norms. What's good for the goose is good for the gander. Whether or not this is a productive state of affairs is besides the point.
Democratic control of the House and Senate, which seems likely will probably result in Trump's actual impeachment. The interesting question is will they pull that trigger. There are enough "reasons" to make it on its face justified but almost everyone knows that the real reason is they hate him. That would be a significant escalation of degrading norms. Leading to situations where having the presidency and failing to maintain one of the legislature chambers is fatal. Also President Vance is likely more competent and less of a situation to be milked going into the 2028 election. The political calculus will be interesting.
I'd note that the last time Dems went with a political calculus approach, RGB's retiring under Clinton instead of Obama, it blew up in their faces in a massively bad way. Are Dem voters willing to stomach another round of Dem establishment fuckups as they try 4D chutes and Ladders instead of actually delivering wins the voters want. Might be interesting to see normie Dems string the representatives up by their toes and vote DSA-version Trump into office in 2032.
I use Astra for work stuff, mostly coding and research. But the idea that you don't check it is wild to me. If it fucks up, and it can, it is incredibly professionally embarrassing. People will rightly blame you for the AI-fuckup, they will think less of your competence, and in a research/engineering setting that's professionally bad.
I have a junior that uses Astra for a ton of his work. It keeps tunneling on stuff that doesn't really matter or not really core to the goal of the project. Loses the forest for the trees so to speak. He uncritically just reports what it says and when questioned about why, or what it means he can't explain it. We've extensively talked to him about his AI usage, and honestly if it continues my boss is considering putting him on "PIP" (We don't really pip in the traditional sense).
More options
Context Copy link