abcd111

Karma: 2

abcd111 10 Apr 2026 3:43 UTC
1 point
0
on: Some takes on UV & cancer
Skin damage happens before you start burning. Sunscreen helps little because its a poor medium for protection (and also because sunscreen can be inconvenient). Also SPF I think measures quantity, but not depth which is as important.
For UV protection supplements are better:
- carotenoids (astaxanthin, lycopene, lutein)
- polypodium

abcd111 23 Mar 2026 14:48 UTC
1 point
0
on: The Hundred Schools of Dietary Advice
Personally I don’t see that much pluralism in dietary recommendations if you can filter out the salesmen. Does the presence of marketing oriented material make the question more plural than other? I would say no, you can find the same in medicine, fitness, education, job opportunity and other niches.
Sure some presentations might emphasize different biological pathways when talking about the same food, but the majority is usually in favor of against. You can test this empirically on that ai app that searches for publications and can give you the number of papers in favor or not.
Ofc you also have to account for who is funding the study and possible lobbies affecting public perception (e.g. big sugar, big meat,...). Again, these are not issues specific to diet, every sector has their own lobbies.
We also shouldnt assume that we know everything about human diet today that we can answer all the questions. E.g. we definitely know that trans fats are bad. Since you mentioned eggs, it is true that today medical opinion isn’t definitive because of cholesterol. How much should you care about cholesterol? Hdl, ldl, particle count. The answer would be like eggs are healthy unless you are seriously trying to lower your cholesterol which you have verified has caused plaque build up in your circulatory system.
We can also frame every health advice statistically, like in 80% cases eggs are healthy. But then the patient has to guess in which probability slice they are in.
The problem is more like, can you simplify the answer in a way that doesn’t require reading 10 books 100 articles and watching 1000 YouTube videos?

abcd111 17 Mar 2026 3:48 UTC
0 points
0
in reply to: derelict5432’s comment on: You can’t imitation-learn how to continual-learn
Its about the homogeneity of the data representation. An argument you could make is, if the hippocampus “stores” data in the neocortex, isn’t the neocortex “just a database”? Even the brain uses different forms of neural networks. At that point the only distinction is that those neural networks share the same protocol of communication (spiking, dendrites, synapses, etc) whereas an llm breaks the protocol when it switches to tool use.
Another distinction, however, is that in a continual learning system, the new data affects the previous set, in a way that “code is data and data is code”, while an llm accessing a database doesn’t affect the capabilities already learnt by the llm. But what if the llm accesses e.g. a prompt persona from the database? At that point the important question becomes: is ICL really “learning”? I would say not, it is just a preference optimization over what the llm has already learned.
An attempt to reach a more formal definition of continual learning could be that ultimately it is a system that is irreversibly updated (it has no inverse) from its learning process.

abcd111 12 Mar 2026 4:25 UTC
−1 points
0
on: Don’t Let LLMs Write For You
many people use llms to write garbage. But this was the case even before llms, e.g. article farms. I would say that llms have raised the bar in such cases.
I let the llm write for me. I prompt it for the concepts to write about, read it, tell it to rephrase something I don’t like. The initial prompt (before the article key ideas) includes information about me, such that the output is closer to what I would express things with. The output is blatantly still LLM sounding.
Could I just put the prompt in my blog and let others ask their own llms? That wouldn’t convey what I want. There is still value in the right prompt as the possible questions are uncountable. There is value in tuning the output.
To retain some human touch I prefer to manually write the footnotes in places where I don’t like the phrasing of the llm that much but changing it would break the flow.
English standardized international language in the last half century, through the internet. LLMs are standardizing the English form. Yes, its kinda Orwellian, but compared to news papers and televisions, they are not worse, but better. You can say that old media does a worse job at plasmating human minds with repetitive and often times subliminal messages, but there is variance in the media sources. LLMs definitely provide much more variance in their outputs, but less in their sources.
Speed reading is as important as it has ever been. LLMs generate content with proper bolding, pacing, keywords better than your average article. If you think llm content feels empty, just speed read through it, tune your attention to the hints the content gives you.

abcd111 26 Feb 2026 14:52 UTC
−2 points
0
on: UFO Aliens Are Your Gods
There are some problems with UFOs:
- History interpretation is not clear cut. There is a lot of things that can be explained with hallucinogenics, like mushrooms being a staple in the communion of early days christianity.
- Rare earth theory keeps adding more and more constraints. There is a lot that we don’t know about the cosmos and quite frequently previous assumptions about astrophysics are invalidated by new observations
- UFOs are one of the favorite psyops to hide the existence of global mind control tech and advanced VR tech
- The hidden bias: we went from anthropocentric to heliocentric, but maybe the truth is in the middle. Meaning that there might be quantum properties exploited by biology that only happen in our corner of the universe.
- There is a class of flying objects that exploits tech just for the purpose of “awing”, very useful in global cognitive war.
The problem with LessWrong and conspiracies is that LW relies on rationatily and epistemology. But when you have an adversary well equipped to drown the 1% truth in distracting narratives, corrupting evidence, planting fake testimonies, there isn’t much of a rational conversation to be had. Weaponization of plausible deniability, the failure of Occam’s razor are real issues with no solution that I know of.

abcd111 25 Feb 2026 13:31 UTC
3 points
0
on: What is a species?
Because evolution happens in discrete steps, speciation happens at particular moments in history. If you wanted the “correct” method for categorization you would pick the cause(s) of the speciation events. But this often times not possible because we only know traces of the history of evolution.
Crossbreeding is a poor choice for categorization because it only captures one chapter of the story. Its a solution used by genes to carve out their niche against other genes, but its not the only solution.

abcd111 25 Feb 2026 4:00 UTC
1 point
0
in reply to: Steven Byrnes’s comment on: [Intuitive self-models] 7. Hearing Voices, and Other Hallucinations
A thought is the event that retrieves a piece of knowledge, you can’t have one without the other.
A simpler model to how the thoughts pop in your head is based on trauma. Let’s call every past experience a trauma, be it good or bad doesn’t matter, the intensity matters instead. The ones that pop before the others are the ones that originated from a stronger trauma. You then couple this with a form of down regulation where if you have thought something recently your brain might present the second association which was generated by a trauma less intense than the first one and so on. You will see though that it is not clear cut, you might have still the first thought in mind, but it doesn’t monopolize your attention (a person that struggles to do this is affected by OCD). Although attention is a sequential process, the subconscious is the one that processes all the sensory inputs and it has to be parallel.
The other thing is better discussed in dm if you are interested. You can say there is a cloud of evidence and a precipice of logic around it.

abcd111 24 Feb 2026 14:16 UTC
1 point
0
on: [Intuitive self-models] 7. Hearing Voices, and Other Hallucinations
For example, people report that hallucinated voices say things, and they don’t have introspective access to where those words came from. But with self-generated inner speech, you also don’t have introspective access to where those words came from!!
Actually most times you do have introspective access of where your knowledge comes from. The brain is quite zealous about where it acquires a particular piece of knowledge, because memory is fundamentally based on associations, and associations are much stronger if based on emotions. Anything that you remember, if you think about it there is an emotion tied to it (but you have to traverse the full trace of the memory, up to the last node, which is not readily available).
Schizophrenia is an artificial illness popularized after ww2 (1947+). Before that you had “ideophrenia” which didn’t mention hallucinations. After that its just been like 1% complicitness, 9% trying to hopelessly frame it in a biological way, and 90% applying whats written in the DSM. Its astonishing how much damage the epistemology of just one narrow domain such as psychiatry can cause. Imagine if it had been done to physics.

abcd111 18 Feb 2026 14:37 UTC
2 points
0
on: The brain is a machine that runs an algorithm
The sensory inputs and the body are much more integral than one would think. Biology is more of a cooperation between intelligent subsystems. Moreover the quantum properties that are harnessed by biology are not at all clear. There might be fundamental ways in which reality works that enable DNA to do what it does thanks to reliable interactions between particles like photons and electrons and maybe other excitations of quantum fields. Does this mean that the brain is not a machine? Well its a machine just like society is a machine, the curse of semantics. Does it run an algorithm? Well nobody wrote it. It just so happens that every complex form of matter is subject to a number of constraints proportional to its level of complexity. And its those constraints that dictate its behavior. You can say that such “algorithm” is sculpted by the constraints of the system, its just a byproduct. So there isn’t a “well defined entity” that its running a “well defined process”.

abcd111 12 Feb 2026 13:30 UTC
7 points
1
on: My journey to the microwave alternate timeline
Velcro has clear downsides, it can’t handle dirt on what is usually the dirtiest place on your body, and nobody wants loose shoes.
Microwave requires people to follow the programs to a T or know enough about the tools and the foods they eat and the portions in order to avoid over/under heating (in other words it requires too much precision). Many people cook by taste and MW doesn’t allow you to do that. The utensils for MW is likely the last of your problems, you can use ceramic, clay, pyroglass all quite common. In the past you didn’t have pre set programs, and frankly even today they aren’t that useful. The safety standards for microwave are higher than ovens, and anyway many still come with radiating plates just like normal ovens if you want to make the surface crispy. A large microwave oven is impractical considering you need specialized cookware of the right size to cactch the waves. MW doesn’t allow you to dry foods as easy as a common oven, the box usually gets stuffed with vapor, or smoke. The tech is just not competitive enough vs radiating ovens, not by price, not by features, not by ease of use. The reason they are only good at reheating 1 portion foods is because that’s their all around niche. 2T motors are still the standard when it comes to 25/50c instead of 4T.
Air friers seem to have a brighter and crispier future.

abcd111 27 Jan 2026 13:43 UTC
2 points
0
in reply to: Dom Polsinelli’s comment on: Notable Progress Has Been Made in Whole Brain Emulation
Bioresonance. Current (public) hardware and software can get you to 1mm resolution, advancements could probably get you to cell size. You don’t exactly get molecular resolution, but exchange of energy (electrons, photons in bulk and their spin), which is what I mean by cause and effect. The technology is non invasive and is not damaging (like xrays) so you can collect lots of data and then train a NN on it. The output of the NN still needs labeling though, but labeling the outputs is a much smaller problem. Its “probably” been done before because the technology is old but not really popular in US as it is in Russia (and east Europe), but still not public because not all science is open.

abcd111 26 Jan 2026 6:09 UTC
−1 points
−1
on: Notable Progress Has Been Made in Whole Brain Emulation
If you want to scan the whole brain you don’t want contrasting agents, you just scan the biofield, which gives you accuracy down to photons. There is a lot of noise in the biofield, but you can reduce it by focusing on fewer parts at a time. You get molecular fingerprints, which would be unlabeled, and the big effort is effectively labeling which is which. Nonetheless, if your goal is just WBE you don’t need to know “what each part is”, but just the cause and effect of a sequence of variables. This has already been done...but I digress.