The content of this post was a bit hard to follow, so I seriated the sentences based on openai embeddings, for everyone’s convenience.
(The filenames may be hash gibberish, and the file-created times arbitrary, or indicate nothing more than when you happened to download a file.)
It is obscure (I had never heard of the term until a year ago or so), but highly useful: it works for everything from seriating Egyptian graves by rough burial time to organizing tag entries by topic.
This could be used to seriate LW2 tag-entries by something more relevant than either just date or upvotes.
It could also be used to present the tags themselves as a big 2D grid.
Then you just seriate them as if they were normal tagged items.)
(A tag embedding is the average of all its members’ embeddings.
Since we now have neural embeddings for just about every modality there is, that means you can seriate anything.
I use it on Gwern.net (background) to organize the ‘similar links’ recommendations in a way much smarter than the naive k-NN embedding distance retrieval approach.
You can do further nice tricks with it, like infer the number of clusters/topics by where distances spike in size, and label them automatically with a LLM.
if you seriate them, however, suddenly you see clear clusters/topics emerge out of the chaos, and it’s easier to skim the list.
Seriation also works beautifully for photographs or other big jumbles of images, where usually there is no way to ‘sort’ them that matters.
The practice of ‘sorting by color’ can be a simple form of seriation, and better than nothing.
It turns out that it is perfectly possible to loosen the definition of ‘sorting’ to something more approximate like ‘try to minimize how different each item is from the next one’; this approximate or generalized sorting is called ‘seriation’.
Seriation [or “ordination”], i.e., finding a suitable linear order for a set of objects given data and a loss or merit function, is a basic problem in data analysis.
The infrastructure comprises data structures to represent linear orders as permutation vectors, a wide array of seriation methods using a consistent interface, a method to calculate the value of various loss and merit functions, and several visualization techniques which build on seriation.
In this paper we present the package seriation which provides an infrastructure for seriation with R.
“Getting Things in Order: An Introduction to the R Package seriation”, Hahsler et al 2008:
This is a linkpost for https://www.jstatsoft.org/article/download/v025i03/227
To illustrate how easily the package can be applied for a variety of applications, a comprehensive collection of examples is presented.
What might you use it for?
It works but the data formatting issues are a bit tricky, so I think I may have to scrap my little prototype and start over with a JSON/‘function-calling’-style approach, which I am not too familiar with.
I sometimes seriate regular lists of text in my writing, where I can’t come up with something meaningful.
I’ve also played around with an OA API LLM script for doing seriation on short natural language text lists, which could be used to automatically seriate lists in anything you write, or which could be used to clean up messy notes.
(For example, you could seriate your chaotic incoherent notes, then tell a LLM to rewrite your notes strictly line by line, and then, with the organization and grammar/spelling all cleaned up, start working on it yourself.)
But it’s not hard, a GPT-4o-scale LLM understands pretty well a prompt like “Reorder them to group similar items, but do NOT rename, add, or remove any items.”, so you can easily make your own tool.
Have you ever wondered how to sort a list or a folder of files where no strict sorting comparison operator like ‘newer than’ is quite right?
Keep this in mind the next time you see a list or a grid anywhere, where there’s not an obviously correct way to sort it: it doesn’t have to be sorted in a dumb way or left sorted at random, when it could be… seriated!
If you just sort them by ‘distance’, it is mostly meaningless and produces a jumble (see for example any algorithmic set of recommendations, like YouTube video lists—if I open a cat video, I see cat/anime/Touhou/cat/CS/music/cat/...);
Because it works so well, and is so simple to implement (simple greedy distance minimization, no need for TSP solvers), I initially called it “sort by magic”.
Nevertheless, both exact solution methods and heuristics are available.
Caused by the problem’s combinatorial nature, it is hard to solve for all but very small sets.
ArthurB
Interestingly o1-pro is not available for their team plan which offers the guarantee that they do not train on your data. I’m pretty sure they are losing money on o1-pro and it’s available purely to gather data.
Popular with Silicon Valley VCs 16 years later: just maximize the rate of entropy creation🤦🏻♂️
#e/ac
We have a winner! laserfiche’s entry is the best (and only, but that doesn’t mean it’s not good quality) submission, and they win $5K.
Code and demo will be posted soon.
Exactly. As for the cost issue, the code can be deployed as:
- Twitter bots (registered as such) so the deployer controls the cost
- A webpage that charges you a small payment (via crypto or credit card) to run 100 queries. Such websites can actually be generated by ChatGPT4 so it’s an easy lift. Useful for people who truly want to learn or who want to get good arguments for online argumentation
- A webpage with captchas and reasonable rate limits to keep cost small
In general yes, here no. My impression from reading LW is that many people suffer from a great deal of analysis paralysis and are taking too few chances, especially given that the default isn’t looking great.
There is such a thing as doing a dumb thing because it feels like doing something (e.g. let’s make AI Open!) but this ain’t it. The consequences of this project are not going to be huge (talking to people) but you might get a nice little gradient read as to how helpful it is and iterate from there.
It should be possible to ask content owners for permission and get pretty far with that.
AFAIK what character.ai does is fine tuning, with their own language models, which aren’t at parity with ChatGPT. Using a better language model will yield better answers but, MUCH MORE IMPORTANTLY, what I’m suggesting is NOT fine tuning.
What I’m suggesting gives you an answer that’s closer to a summary of relevant bits of LW, Arbital, etc. The failure mode is much more likely to be that the answer is irrelevant or off the mark than it being at odds with prevalent viewpoints on this platform.
Think more interpolating over an FAQ, and less reproducing someone’s cognition.
Speed running everyone through the bad alignment bingo. $5k bounty for a LW conversational agent
The US has around one traffic fatality per 100 million miles driven; if a human driver makes 100 decisions per mile
A human driver does not make 100 “life or death decisions” per mile. They make many more decisions, most of which can easily be corrected, if wrong, by another decision.
The statistic is misleading though in that it includes people who text, drunk drivers, tired drivers. The performance of a well rested human driver that’s paying attention to the road is much, much higher than that. And that’s really the bar that matters for self driving car, you don’t want a car that is doing better than the average driver who—hey you never know—could be a drunk.
Fixing hardware failures in software is literally how quantum computing is supposed to work, and it’s clearly not a silly idea.
Generally speaking, there’s a lot of appeal to intuition here, but I don’t find it convincing. This isn’t good for Tokyo property prices? Well maybe, but how good of a heuristic is that when Mechagodzilla is on its way regardless.
In addition
There aren’t that many actors in the lead.
Simple but key insights in AI (e.g doing backprop, using sensible weight initialisation) have been missed for decades.
If the right tail for the time to AGI by a single group can be long and there aren’t that many groups, convincing one group to slow down / paying more attention to safety can have big effects.
How big of an effect? Years doesn’t seem off the table. Eliezer suggests 6 months dismissively. But add a couple years here and a couple years there, and pretty soon you’re talking about the possibility of real progress. It’s obviously of little use if no research towards alignment is attempted in that period of course, but it’s not nothing.
There are IMO in-distribution ways of successfully destroying much of the computing overhang. It’s not easy by any means, but on a scale where “the Mossad pulling off Stuxnet” is 0 and “build self replicating nanobots” is 10, I think it’s is closer to a 1.5.
Indeed, there is nothing irrational (in an epistemic way) about having hyperbolic time preference. However, this means that a classical decision algorithm is not conducive to achieving long term goals.
One way around this problem is to use TDT, another way is to modify your preferences to be geometric.
A geometric time preference is a bit like a moral preference… it’s a para-preference. Not something you want in the first place, but something you benefit from wanting when interacting with other agents (including your future self).
The second dot point is part of the problem description. You’re saying it’s irrelevant, but you can’t just parachute a payoff matrix where causality goes backward in time.
Find any example you like, as long as they’re physically possible, you’ll either have the payoff tied to your decision algorithm (Newcomb’s) or to your preference set (Solomon’s).
I’m making a simple, logical argument. If it’s wrong, it should be trivial to debunk. You’re relying on an outside view to judge; it is pretty weak.
As I’ve clearly said, I’m entirely aware that I’m making a rather controversial claim. I never bother to post on lesswrong, so I’m clearly not whoring for attention or anything like that. Look at it this way, in order to present my point despite it being so unorthodox, I have to be pretty damn sure it’s solid.
That’s certainly possible, it’s also possible that you do not understand the argument.
To make things absolutely clear, I’m relying on the following definition of EDT
Policy that picks action a = argmax( Sum( P( Wj | W, ai ). U( Wj ), j ) , i ) Where {ai} are the possible actions, W is the state of the world, P( W’ | W, a ) the probability of moving to state of the world W’ after doing a, and U is the utility function.
I believe the argument I made in the case of Solomon’s problem is the clearest and strongest, would you care to rebut it?
I’ve challenged you to clarify through which mechanism someone with a cancer gene would decide to chew gum, and you haven’t answered this properly.
If your decision algorithm is EDT, the only free variables that will determine what your decisions are are going to be your preferences and sensory input.
The only way the gene can cause you to chew gum in any meaningful sense is to make you prefer to chew gum.
An EDT agent has knowledge of its own preferences. Therefore, an EDT agent already knows if it falls in the “likely to get cancer” population.
Yes, the causality is from the decision process to the reward. The decision process may or may not be known to the agent, but its preferences are (data can be read, but the code can only be read if introspection is available).
You can and should self-modify to prefer acting in such a way that you would benefit from others predicting you would act a certain way. You get one-boxing behavior in Newcomb’s and this is still CDT/EDT (which are really equivalent, as shown).
Yes, you could implement this behavior in the decision algorithm itself, and yes this is very much isomorphic. Evolution’s way to implement better cooperation has been to implement moral preferences though, it feels like a more natural design.
Do you prompt the LLM to do the whole rewrite or call it n(n-1)/2 times to get the distances?