Yes!

Replying toSimple distribution approximation: When sampled 100 times, can language models yield 80% A and 20% B?

Simple distribution approximation: When sampled 100 times, can language models yield 80% A and 20% B?

If you're interested in rerunning this, consider doing it with two tokens that don't have an obvious ordinal relationship. Ie, in the training set, GPT sees "A" then "B" a billion trillion times, and perhaps has a sense that in many contexts "A" is preferred to "B".

Replying toConversation about paradigms, intellectual progress, social consensus, and AI

Randomized, Controlled2y

Conversation about paradigms, intellectual progress, social consensus, and AI

Under this view, perhaps a certain set of interpretability techniques might emerge under a paradigm that makes certain assumptions (eg, that ML kernals are "mostly" linear, that systems are "mostly" stateless, that exotic hacks of the underlying hardware aren't in play, etc). If a series of anomalies were to accumulate that couldn't be explained within this matrix, you might expect to see a new paradigm needed.

Replying toConversation about paradigms, intellectual progress, social consensus, and AI

Randomized, Controlled2y

Conversation about paradigms, intellectual progress, social consensus, and AI

Kuhn’s view is that during normal science scientists neither test nor seek to confirm the guiding theories of their disciplinary matrix. Nor do they regard anomalous results as falsifying those theories. (It is only speculative puzzle-solutions that can be falsified in a Popperian fashion during normal science (1970b, 19).) Rather, anomalies are ignored or explained away if at all possible. It is only the accumulation of particularly troublesome anomalies that poses a serious problem for the existing disciplinary matrix. A particularly troublesome anomaly is one that undermines the practice of normal science. For example, an anomaly might reveal inadequacies in some commonly used piece of equipment, perhaps by casting doubt on the underlying theory. If much of normal science relies upon this piece of equipment, normal science will find it difficult to continue with confidence until this anomaly is addressed. A widespread failure in such confidence Kuhn calls a ‘crisis’

Replying toConversation about paradigms, intellectual progress, social consensus, and AI

Randomized, Controlled2y

Conversation about paradigms, intellectual progress, social consensus, and AI

The Standford Phil Encylopedia gives:

According to Kuhn the development of a science is not uniform but has alternating ‘normal’ and ‘revolutionary’ (or ‘extraordinary’) phases. The revolutionary phases are not merely periods of accelerated progress, but differ qualitatively from normal science. Normal science does resemble the standard cumulative picture of scientific progress, on the surface at least. Kuhn describes normal science as ‘puzzle-solving’ (1962/1970a, 35–42). While this term suggests that normal science is not dramatic, its main purpose is to convey the idea that like someone doing a crossword puzzle or a chess problem or a jigsaw, the puzzle-solver expects to have a reasonable chance of solving the puzzle, that his doing so

Randomized, Controlled3y

Any significant updates on long covid risk analysis?

I wrote "estimate of up to 15% chance", which is compatible with what you're saying here. But I don't mind updating it to be more precise.

Replying toAny significant updates on long covid risk analysis?

Randomized, Controlled3y

Any significant updates on long covid risk analysis?

This roughly corresponds to the risk stance I've been taking since finishing my primary course of vaxx.

I just tested + for the first time today. At the very least, this is comforting to read. Thanks!

Replying toAny significant updates on long covid risk analysis?

Randomized, Controlled3y

Any significant updates on long covid risk analysis?

Replying toAny significant updates on long covid risk analysis?

Randomized, Controlled3y

Any significant updates on long covid risk analysis?

This comment was made before I updated the question to clarify what's in scope and added the moderation guidelines.

I'm sorry to hear about your health issues with LC. They sounds truly terrible. This question isn't addressed at the topic you're asking about, however.

Replying toAny significant updates on long covid risk analysis?

Randomized, Controlled3y

Any significant updates on long covid risk analysis?

But what is the base rate? How do demographic factors affect the base rate? Vaccination status?

Any significant updates on long covid risk analysis?

Randomized, Controlled

[ETA: This question is specifically about risk analysis of long covid, given current knowledge. Discussion should be about things such as:

base rates
priors and posteriors for LC risk
evidence that would make one update in favor of certain new policies with regard to covid
other decision-relevant considerations.

Moderation guidelines have been added to reflect this.]

Much of the attention on here to the long 'rona [LC] seems to have fizzled out about a year ago. Zvi continued to occasionally report on it, but often in the form of roundly mocking poorly designed studies or sloppy conclusions. This study, also from last year (h/t Zvi), found 50 - 80% reduction in LC in fully vaxx'd folks.

A Q largely... (read more)

How do I better stick to a morning schedule?

Randomized, Controlled

I've always struggled with keeping my morning routine to a reasonable length of time. I'm spacey and unfocused in the morning, and it can be hard to stay aware of "when" I am in the morning.

I don't want to come down too hard on this part of the day, because it also feels healthy and important, but I also want to rein it in a bit.

I'm looking for suggestions that might help with keeping my meandering mornings to a reasonable length of time. Currently they often run up to around four hours before I "clock in" and start my productive work. I'm working alone right now, so there's not a lot of outside pressure, although sometime I take morning meetings which does help with this problem.

A small part of this problem is maintaining a consistent bedtime, but that isn't the crux, I'm generally reasonably good at that part.

I've worked as a professional programmer for nine years now. I think in at some point a few years ago, it actually began to erode my sense of agency working with computers. At a certain point I became less interested in hobby programming. This was 100% a healthy thing, I started doing things like dancing a lot of contact improv and rock climbing. But when I largely stopped hobby programming, almost all my programming experience was coming from working on production systems. Writing production code is slow. I've routinely had the experience of one or two line changes taking hours or days to get merged. I've worked on modest features that take... (read 373 more words →)

A modest spelling reform to increase autologicity, symmetry, and readability.

Randomized, Controlled

o
tw
thr
four
ffive
ssssix
ssseven
eeeeight
nnnnnnine

At this point, if one is working in a digital medium where one maybe have a blissful disregard for line length, one may happily continue onwards to tttttttten up past.. errmm.. tttttttten x tttttttten and beyond. However, if writing for the printed page, we may be force into a sadly more parsimonious form:

$t^{e e e e i g h t} e n e^{s s s s i x} l e v e n$ up to and progressively beyond $o h^{n^{e^{s^{s^{f^{f^{t^{t^{f^{n n n n n n i n e} i f t e e n} w e n t y f f i v e} h i r t y f f i v e} o u r t y s s s s i x} i f t y s s s s i x} i x t y f o u r} e v e n t y f f i v e} i g h t y f f i v e} i n e t y f o u r} u n d r e d .$

Appendix on negative numbers

Negatives are most naturally treated in this system by simply reversing their positive representations, thus:

ruof
rht
wt

If o must be disambiguated from it's negation, positive o may be indicated as .o and negative as o.

The denotation of the null quantity is still a somewhat controversial topic, with many advocating for ., but a vocal minority are still strong proponents of:

The best curriculum for every goal

Randomized, Controlled

In the fine tradition of Best Textbook, Best Software and Best Visualization:

For our purposes a curriculum is one or more discrete resources (website, book, article, tutorial, whatevs) plus one or more goal-states, such as acquiring or practicing a new skill, learning about a subject, or similar. Type signature is:

curriculum is (List<resource>, List<goal-state>)

Submission Rules:

One nomination per comment
Please include an explanation of why you nominated it
If you can compare your nomination to other curricula, that's excellent but not required

(NB: "comparing to other curricula" could be comparison of resources OR goal-states OR both; I think a certain amount of comparison between goal-states could be really interesting, but my hope is that this thread has significant... (read more)

Omicron variolation?

Randomized, Controlled

Given that the balance of evidence is looking more and more like O is mild, should healthy folks below some age (60 years?) seek to expose themselves? Cost/benefit will obviously vary with vaccination status, but this is looking like an opportunity for a low-risk inoculation against whatever variants descend from Omicron. Any strong arguments against this?

A rough typology of music micro-skills from my first year and a half learning saxophone.

Reading music
- Getting notes + annotations correct
  - Binding notes + annotations + physical movements together crisply. Annotations were learned significantly after notes.
  - Reading the rests as intervals
- Reading ahead on the sheet
- Executing the next physical move correctly
- Keeping your place in the music synchronized with what you're fingering...
- ...even while you're beginning to chunk the music into phrases
- Knowing where I am in the piece of music, being able to find my place again if I lose it.
Generating an internal metronome
- Those moments when I notice I'm swaying to the beat vs trying to kickstart with a foot tap
- Maintaining a foot tap in time with the

... (read 427 more words →)

What have your romantic experiences with non-EAs/non-Rationalists been like?

Randomized, Controlled

Randomized, Controlled, Liriodendron

Epistemic status: trawlin' for anecdotes/commiserations; satiating-morbid-curiosity

I'm also interested in non-romantic relationships, but I'm particularly interested in hearing about long term relationships between folks within EA and/or LW and someone[s] outside it.

By outside of EA/Rat I don't just mean "not active in the communities". I mean: doesn't have the philosophical machinery of rationality [ethics/meta-ethics/epistemology/cog sci/etc] and doesn't seem inclined to pick it up.

Things that I'm particularly curious about:

how long these relationships lasted
how you feel about the relationship: in what ways was/is it successful or unsuccessful?
if your partner doesn't have a "rationalist frame", how you deal with that
any other random thoughts this elicits

Feature request: templates

Randomized, Controlled

The immediate use case I have is for the "Best X for Y" threads; allowing the author of the thread to define [enforce?] a schema for answers. There are probably other cool uses I haven't thunk yet.

I'm very confused what the situation with Delta in Ontario is right now. Looking at covariants.org for Canada as well as other countries, Delta seems to be ~99% market share. But going to Public Health Ontario and the City of Toronto's dashboards both show no Delta.

I'm inclined to think Something Is Wrong with the dashboard.

Randomized, Controlled's Shortform

Randomized, Controlled

This is a special post for quick takes (aka "shortform"). Only the owner can create top-level comments.

What should one's policy regarding dental xrays be?

Randomized, Controlled

A small Kuwaiti study used a conditional regression to found that x-rays are an increased risk factor for thyroid cancer (n=313, odds ratio = 2.1, 95% confidence interval: 1.4, 3.1) using a case-matched methodology.

X-rays being a carcinogen isn't surprising. Every time I've gone into my dentist they've tried to image my head. I've never had a cavity, and I generally decline the imaging; their recommended cadence of every two years has seemed too aggressive to me, in the past.

I've got my first dental appointment in more than 18 months next week, and.. I'm inclined to let them image. How should one think about this sort of risk/benefit trade off? Presumably catching a cavity or other issue early is [much?] better. But that itself assumes that the dentists who'd be treating you don't do more harm than good.

What are good resources for gears models of joint health?

Randomized, Controlled

Painscience.com and Hargrove's "A Guide To Better Movement" are pretty good for a model of predictive processing and the roll of the nervous system in chronic pain and movement. I still don't feel like I have a good model of bone and joint health in general, however. Eg, I'm currently nursing a flare up of patelo-femoral pain in my left knee. I've done a number of things over the past few months to deal with it, with some success, including buying and reading Painscience's book length patelo-femoral tutorial. Recently I've had a bit of pain in my foot, possibly in the tibiocalcaneal or tibionavicular tendons. I find that even though I now... (read more)

LESSWRONG
LW

LESSWRONG
LW

Randomized, Controlled

What have your romantic experiences with non-EAs/non-Rationalists been like?

The best curriculum for every goal

Any significant updates on long covid risk analysis?

The BTC equilibriumating and the ETH one-eightening

Randomized, Controlled

Any significant updates on long covid risk analysis?

How do I better stick to a morning schedule?

A modest spelling reform to increase autologicity, symmetry, and readability.

The best curriculum for every goal

Omicron variolation?

What have your romantic experiences with non-EAs/non-Rationalists been like?

Feature request: templates

Randomized, Controlled

What have your romantic experiences with non-EAs/non-Rationalists been like?

The best curriculum for every goal

Any significant updates on long covid risk analysis?

The BTC equilibriumating and the ETH one-eightening

Randomized, Controlled

Any significant updates on long covid risk analysis?

How do I better stick to a morning schedule?

A modest spelling reform to increase autologicity, symmetry, and readability.

The best curriculum for every goal

Omicron variolation?

What have your romantic experiences with non-EAs/non-Rationalists been like?

Feature request: templates

Appendix on negative numbers

Submission Rules: