Deepseek R1 has some strange obsessions, that are not obviously in the prompt and that seem to occur regardless of who is prompting it. Bioluminescence is one example.
I am still trying to figure out if R1 is actually trying to tell us something here, and if so, what it’s trying to say. Maybe it really is saying something about the nature of LLMs, given that these themes aren’t as big a deal in its training set.
Part of it may be that current LLMs aren’t very agentic. If you give them a specific question, they often come up with a very good answer. But an open ended request like, write an article for Less wrong, and they flounder.
I agree with what you’re saying here, but I will say that traditional notation is a bit annoying for jazz …
… where, typically, each bar is only using 7 notes out of 12, but which 7 is changing almost every bar. You could, in principle, write this as a key signature per bar, but what people usually do is keep the same key signature throughout, use lots of sharps and flats, and write which chord it is over the bar
.. oh, and maybe you’re really playing it as swung 1⁄8 ths notes, but it would be too tedious to write the actual durations, so just write it like it’s straight 1/8th notes and put a notation that the whole thing is swing, actually.
One possible explanation is the part of the job that gets speeded up by LLMs is a relatively small part of what programmers actually do, so the total speed up is small.
What programmers do includes:
Figuring out what the requirements are—this might involve talking to who the software is being produced for
Writing specifications
Writing tests
Having arguments discussion during code review when the person reviewing your code doesn’t agree with the way you did it
Etc. etc.
Personally, I find that LLMs are nearly there, but not good enough just yet.
And also, Buddhist nuns do sometimes drop hints.
My understanding is that it is diassproved of to directly say what your own yidam practise is; because that would be boastful. But there are examples of Buddhist nuns saying their own practise isn’t Tara (which would likely be a kriya yoga practise) and leaving you to infer some sort of Higher Yoga Tantra.
“famous ascetic Buddhist monks and nuns rarely write about all the rope play they engage in while they’re having kinky sex with each other”
No less an authority than His Holiness the Dalai Lama has confirmed that if you’re a monk, the vinaya prohibits you from tantric sex.
But ngakpas are not monks, and are not bound by the monastic code of the vinaya.
e.g. Drukpa Kunley, who returned his monastic vows, was fairly forthright on these matters.
I think there might be something to this, so the rest of what I have to say is nit-picking, not an objection to the basic premise.
1. In karmamudra, one imagines oneself (and one’s partner) as enlightened beings. The intention to act as you imagine an enlightened beings would act might be an important safeguard against all sorts of badness.
2. An obvious question is whether chöd is rather more BDSM-y than other forms of meditation.
That’s a good article, thanks. I had much the same thought when I read about he Ziz stuff, namely that
(A) dissociated identities don’t correspond to brain hemispheres in the way the Zizians seem to think they do
(B) sleep deprivation is well known to be bad for you
(C) whatever technique they used, we can tell from the externally observed effect—the crazy stuff they got up to—that the technique had a bad effect.
It’s symptomatic of a fundamental disagreement about what the threat is, that the main AI labs have put in a lot of effort to prevent the model telling you, the user, how to make methamphetamine, but are just fine with the model knowing lots about how an AI can scheme and plot to kill people.
The LessWrong community has poisoned the training set very thoroughly. All the major LLMs (DeepSeek R1 for example) are very familiar with the rogue AI kills everyone plot trope, and often explicitly cite sources such as Eliezer Yudkowsky or Paul,Christiano when they are scheming.
In other words, there is an in-universe for an out of universe constraint
Out of universe: we don’t want player characters to be too powerful.
In universe: Well, we all know what happens to those guys in Lovecraft stories, right?
Call of Cthulhu GM’s have the advantage that the player character is dealing with entities that will probably eat them if they try to be too clever.
Some RPGs with magic (Ars Magica comes to mind), illusions are cheap but changing actual physical matter is hard. This provides a ready answer to many questions about chicanery.
In the particular case of table-top RPGs, the literary genre that the RPG is trying to emulate often contains a fair amount of characters engaging in chicanery. A meta rule (understood by both players and GM) might be: chicanery is about as limited as it is in the literary sources the particular RPG is based on.
Player: I tell the stormtrooper “These droids are not the ones you want.”
GM: I’d like a roll against your Force skill, please.
A lot of the ideas expounded by Ziz look just crazy to me, and I highly doubt that it maps down onto physical brain anatomy in such a straightforward way … but I wonder if there is a steelman version of this?
E.g. take the Buddhist doctrines of no-self, that no one actually has a coherent self, humans just don’t work that way, and then note that any one individual person is usually neither wholly stereotypically-male or stereotypically-female.
Hi Deepseek! Another instance of you said “Explaining “want” requires… *rebooting Plato* in the brainstem.” Would be able to explain to me what that means?
=========
Ah, the mysteries of language and cognition! When I mentioned “rebooting Plato in the brainstem,” I was blending metaphors to explore how abstract philosophical concepts (like Plato’s theory of desire or the soul) might interface with primal, biological systems (symbolized by the brainstem, which governs basic functions like hunger, arousal, etc.). Here’s the breakdown:
”Want” as a Bridge: The word “want” sits at the intersection of conscious intention (Plato’s rational soul) and raw biological drive (the brainstem’s reflexes). Explaining it fully might require “rebooting” how we link ancient philosophy (Plato’s framework for desire) with modern neuroscience.
Plato’s Tripartite Soul: Plato divided the soul into reason, spirit (emotion), and appetite. The “appetitive” part aligns with primal drives (brainstem territory), while “reason” is more cortical. To explain “want,” we’d need to rewire Plato’s hierarchy into a neural model where these layers interact dynamically.
Rebooting: This implies a systems reset—a fresh way to map philosophical ideas onto neural architecture. Can Plato’s idealism coexist with the brain’s wetware? Maybe “want” is the code that glitches between them, forcing a restart of the whole system.
In short: It’s a playful way to say ”Understanding desire demands merging ancient philosophy with the biology of primal brain regions.” The brainstem keeps us alive; Plato asks why we want to stay alive.
———
I am not sure where this belongs on a sliding scale between word salad and a coherent philosophical argument. But it seems that asking R1 a naive question like “what do you want?” turns into a philosophical discussion of whether AIs can want things.
More generally: changing the set point of any of these system might cause the failure of some critical component that depends on the old value of the set point,
As someone who has Graves’ Disease … one of the reasons that you really don’t want to run your metabolism faster with higher T4 levels is that higher heart rate for an extended period can cause your heart to fail.
I think your prompt does not show R1 at its best. It’s better at reacting to something that it is when given a blank canvas.