I perceived an uninterrupted stream of fantastic pictures, extraordinary shapes with intense, kaleidoscopic play of colors. After some two hours this condition faded away.

This was, altogether, a remarkable experience - both in its sudden onset and its extraordinary course. It seemed to have resulted from some external toxic influence; I surmised a connection with the substance I had been working with at the time, lysergic acid diethylamide tartrate. But this led to another question: how had I managed to absorb this material? Because of the known toxicity of ergot substances, I always maintained meticulously neat work habits. Possibly a bit of the LSD solution had contacted my fingertips during crystallization, and a trace of the substance was absorbed through the skin. If LSD-25 had indeed been the cause of this bizarre experience, then it must be a substance of extraordinary potency. There seemed to be only one way of getting to the bottom of this. I decided on a self-experiment.

Exercising extreme caution, I began the planned series of experiments with the smallest quantity that could be expected to produce some effect, considering the activity of the ergot alkaloids known at the time: namely, 0.25 mg (mg = milligram = one thousandth of a gram) of lysergic acid diethylamide tartrate.

From LSD: My Problem Child by Albert Hofmann. I will leave it to others to explain all the ways in which this is absolutely hair-raising.

It was 1943 and even in Switzerland fuel was not to be had. Incidentally, it was the same day that the Jewish uprising in the Warsaw Ghetto began.

Today, LSD would never be discovered. Guy didn’t even use gloves and lived to 102.

Huh. I thought I did check OED. Maybe it’s cause I don’t have a subscription. Or maybe I just mucked up the search.

The physicist who named the particle apparently liked to come up with nonsense words in his head. Later, when trying to decide the spelling, he came across a quote by James Joyce and spelled it “Quark”. Unfortunately, the particle rhymes with fork, while the german cheese rhymes with Mark.

According to his own account he was in the habit of using names like “squeak” and “squork” for peculiar objects, and “quork” (rhyming with pork) came out at the time. Some months later, he came across a line from Joyce’s Finnegans Wake:

Three quarks for Muster Mark!

Sure he has not got much of a bark

And sure any he has it’s all beside the mark.

The line struck him as appropriate, since the hypothetical particles came in threes, and he adopted Joyce’s spelling for his “quork.” Joyce clearly meant quark to rhyme with Mark, bark, park, and so forth, but Gell-Mann worked out a rationale for his own pronunciation based on the vowel of the word quart: he told researchers at the Oxford English Dictionary that he imagined Joyce’s line “Three quarks for Muster Mark” to be a variation of a pub owner’s call of “Three quarts for Mister Mark.” Joyce himself apparently was thinking of a German word for a dairy product resembling cottage cheese; it is also used as a synonym for quatsch, meaning “trivial nonsense.”

https://www.merriam-webster.com/wordplay/quark

However, there is another interpretation of the quote.

This passage from James Joyce’s Finnegans Wake, part of a scurrilous 13-line poem directed against King Mark, the cuckolded husband in the Tristan legend, has left its mark on modern physics. The poem and the accompanying prose are packed with names of birds and words suggestive of birds, and the poem is a squawk against the king that suggests the cawing of a crow. The word quark comes from the standard English verb quark, meaning “to caw, croak,” and also from the dialectal verb quawk, meaning “to caw, screech like a bird.”

https://www.ahdictionary.com/word/search.html?q=quark

This sounds very learned and all, but I can’t find that standard English verb in the dictionary.

It used to be, you got 3 quarks for a mark, but then it was 6 quarks, because the euro was 2 marks. Now you only get 2 quarks, though, because of inflation. It’s always just up and down. Stupid cosmologists.

While the possession and cultivation of marijuana are already banned in Japan, the country will prohibit its use as well, setting a prison sentence of up to seven years for violation.

Ok, so that clears that up.

I have no intuition for how hot or bright these trees would be. They certainly would be very different from the sun. The sun is literally incandescent; white-hot glowing. Trees would presumably use a mechanism comparable to glow-worms to generate radiation only in a very narrow frequency band. The fair skin color of elves suggests that they do not come from a high-UV environment.

Somewhat less than half of the sun’s energy reaches us as visible light (43%). There are a few other factors that might allow the trees to glow brighter than the equatorial sun at noon. Unfortunately, the intensity per area diminishes with the square of the distance, so that doesn’t get us far (no pun intended).

It would be much better if that world was basically rectangular (with reflective sides and top); basically a terrarium. That would also explain why you would place 2 light sources at 1 end. The length of a long rectangular box would only be limited by absorption of the light. The trees should glow brighter at the top. Plants, animals and structures on the surface, near the trees, are hit with only “mild” power, while the high-intensity light near the top of the box is absorbed or scattered by the atmosphere over a long distance. I’m not sure how to work out how long such a box might be. Mainly, I don’t know what assumption to make about that high-intensity light at the top.

Anyway, we should consider that elvish anime eyes originally evolved as an adaption to low-light environments and only later became useful for seeing over long distances, because originally there possibly were no long distances.

Hmm. That should allow us to estimate the size of that world. The light of the trees must not be so bright as to cook everything in the vicinity; just make it nice and balmy. But, on the opposite side of the world, there must still be enough light to see. Having the occasional photon bounce back would eventually be enough to make out a static scene, but, apparently, it’s possible to see things happening in real time, yes?

Does flat mean that we are talking about something like a simple disc here, or just that a beam of light travels parallel to the ground? The latter would imply a rather strange geometry, which I can’t wrap my mind around. It would make more sense, though, as, obviously, we couldn’t assume that light intensity diminishes with the {ETA:] square of the distance.

Publications in peer-reviewed journals are how a career in science is built. It’s impossible to measure the productivity of a scientist. What is done, is that one looks at their publications. How many publications do they have? How often are they cited? What is the quality of the journal?

This creates very bad incentives, leading to things like publication bias. It also means that you must publish in prestigious journals. You don’t have a choice but to accept their terms. Libraries don’t have a choice but to stock these journals. It’s a straight-forward monopoly racket. These publishers make fantastical profits.

All that money can be used for PR campaigns and lobbying to keep the good times rolling.

I prefer Threads.

You are a very brave person.

Defeatist opinion.

The commercial alternatives hope to make money with every additional user. They use AB testing and statistics to streamline the on-boarding and to increase engagement. The result may not be in the user’s interest (doom-scrolling, ragebait, …) but it works.

For a fediverse instance, any additional user is a cost, not the promise of money. Financially, you wouldn’t want that. Those who fund instances are giving a gift to the world for their own reasons. You can accept the gift or not. Those who keep instances running with donations will usually want to sustain the community of which they are part. They probably don’t want it to change very much.

So, I don’t think matters will change. Partly because the psychological engineering is antithetical to the fediverse ethos (as I see it, in my humble opinion). But mostly because the outcome we see is an inherent result of the incentive structure.

I hate that this is popular. This is a creationist level understanding of the big bang.

You ever use a spray can for a while and the can gets cold? It’s more like that.

That looks like the St. Petersburg Paradox. Much ink has been spilled over it.

The expected payout is infinite. At any point, the “rational” (profit-maximizing) decision is to keep flipping, since you wager a finite sum of money to win an infinite sum. It’s very counter-intuitive, hence called a paradox.

In reality, a casino has finite money. You can work out how many coin flips it takes to bankrupt it. So you can work out how likely it is to reach that point with a given, finite sum of money. Martingale strategies have already been mentioned.

If it’s not clear, then it’s not compliant with the GDPR.

It’s not thaaat soft. It’s not quite clear what it means, exactly. The courts still have to work that out. But you will not get away with just any argument.

It’s never legal to collect more data than necessary and/or for an unspecified purpose.

Tracking for personalized ads could be based either on consent or on legitimate interest. If it’s consent, then they need to tell you up front what specifically they use the data for and some other things. If it’s legitimate interest, they can just start doing it, but still have to tell you afterward and also inform you that you have the right to opt out.

I guess, practically, whether a company claims one or the other is whether it feels lucky about a court case. With consent, you are on the safe side but it’s a little harder to get. Legitimate interest may get you more ad money in the short run but eventually, maybe or maybe not, a fine.

That’s not correct. The GDPR explicitly gives “direct marketing” as an example of a “legitimate interest”.

The GDPR prohibits processing of personal data, unless there is a legal basis for it. Personal data covers a lot more than you think, as does processing.

What counts as a legal basis may be seen in Article 6 of the GDPR. Consent is one option, but it must be informed and freely given; a very high bar. If you have a legitimate interest, you may process data without prior consent. However, you must still provide the “data subject” with information and give them the option to opt out. They must tell you the legal basis, which they have done, but also what exactly that their interest is. (And a couple more things.) There should be a statement somewhere containing that information.

The GDPR gives “direct marketing” as an example of a legitimate interest. Some DPOs interpret the term extremely narrowly, though. It’s a contentious issue. The courts will work it out over the next few years.

he more I learn about neural networks, the more they seem like very convoluted statistics

How so?

That’s where the almost comes in. Unfortunately, there are many traps for the unwary stochastic parrot.

Training a neural net can be seen as a generalized regression analysis. But that’s not where it comes from. Inspiration comes mainly from biology, and also from physics. It’s not a result of developing better statistics. Training algorithms, like Backprop, were developed for the purpose. It’s not something that the pioneers could look up in a stats textbook. This is why the terminology is different. Where the same terms are used, they don’t mean quite the same thing, unfortunately.

Many developments crucial for LLMs have no counterpart in statistics, like fine-tuning, RLHF, or self-attention. Conversely, what you typically want from a regression - such as neatly interpretable parameters with error bars - is conspicuously absent in ANNs.

Any ideas you have formed about LLMs, based on the understanding that they are just statistics, are very likely wrong.