Erowid is great, but the trip reports were universally fanciful, typically having no discernible connection to the drug (often drugs) being described. They became a literary form of their own, and something of an inside joke among psychedelics users.
The general idea of training LLMs on forum archives seems promising though!
I was just playing a board game with a friend, and we spent a while looking through the (fairly complicated) rules. At some point, I had the bright idea to feed the manual to Claude, so we could ask questions about it.
It all went great, Claude even gave references, until it hallucinated an innocuous but significantly game-breaking rule ("rest lets you heal one health for two stamina").
This just reinforced my lack of trust in LLMs, and I would sure as hell not ask an LLM what drugs were safe or not.
I dont think erowid content would be good to tell safety of a drug, probably more what type of experience you'd get using a drug at certain doses.
What would be interesting would be training using the data from PsychonautWiki[0] which has all sorts of good data about drug information & safety (and experiences)
I'd think the application would be used more for understanding the psychedelic experience. Not about safety. There is such a wide tome of experiences - there are going to emerge similarities to a knowledgeable AI that can load it all up at once. I'm very curious what patterns can emerge with up to date language models.
> Erowid is great, but the trip reports were universally fanciful
I can assure you that my lengthy description of waves of orgasmic bliss that descended into a hospitalised bad trip† while on 2C-I and cannabis is written as I experienced it, nothing fanciful about that. But I also don't understand how do you claim stories about people literally on drugs to be fanciful. What do you expect people to see and experience if they've taken strong psychedelics? I have no basis to doubt one has really spoken to the Alien God while on salvia.
† thankfully it was only a panic attack, but they're a little bit scarier to have when the walls are melting.
I had to look up 2C-I. Phenethylamines in general seem pretty interesting, encouraging even.
I wish we were studying them with competency, in environments that were fully untethered from absolute profitism, LEO fear, politics and other non-medical agendas. We could use some better tools to address the negative effects of our increasingly complex (and decreasingly prepared) living.
2C-I was pretty fun. I ordered it online from a Canadian research lab, paid with credit card, all above board, lol. The panic attack was because of cannabis; over the years I've had half a dozen cannabis-related panic attacks, 2c-I on the other hand was pretty mellow, with lovely visual patterns. Felt less "spiritual"/introspective than shrooms, so to speak.
Also, after I got hospitalised, the ER paramedics returned the small bag of white powder to me. I don't know if that's common practice , but I appreciated the gesture.
I think anything with iodine in it in a non dietary form should be avoided because your thyroid really likes grabbing onto iodine things and there is no telling if that could, for instance, cause hypothyroidism.
> Erowid is great, but the trip reports were universally fanciful
IIRC they have moderators vet them and only a relatively small percentage get published. The founders and staff actually take their mission statement seriously.
Agreed, trip reports have become their own art form, and I suspect that a good percentage is actually made up to fit the emerging art form. My favourite example for this is the following:
The style of writing and the amount of stereotypes makes me convinced this is a rather hilarious piece of writing, not even false, but likely never happened.
> The general idea of training LLMS on forum archives seems promising though!
I chuckled. Reddit data is THE gold standard for training something such as an RHCF model, properly up voted/downvoted data through an API. FWIW I'm sure a lot of ChatGPT data is trained there.
Yes, if we were homo rationalus (or if we are enforced to vote properly), voting systems work great.
"tweakers.net" does something very interesting. They allow you to vote 'freely', but if you don't adhere to their voting rules repeatedly, you will have your voting rights locked away.
Their current rules are:
-1 = flamebait, troll, etc.
+0 = irrelevant, inaccurate
+1 = relevant
+2 = informative, helpful
+3 = must read comment
If you constantly give +3's to mediocre comments because they agree with your bias, you will get a warning. Similar with -1 to dissenting opinions. After one warning, your vote rights are locked away and you will have to write a good justification on getting them back.
To give an example on why voting sucks here for example (yes, I know) and it will sound a bit like sour grapes:
I made a comment on how Sonos is great, because their voice assistant processes your voice completely locally. I get berated and told it is not private because you need an account (the EU would fine Sonos severely if they claim local processing but do not do so). I get downvoted to -2 (!), whereas the completely incorrect comment gets sent to +3.
Similarly, another commentor says Sonos has horrible software lifetimes. However, they often update their products for ~10 years, which is a better support timeline than even Apple. I'm sent to 0 twice, incorrect commenter goes to +2.
With the Tweakers system, Dang would swoop in and annihilate the voting rights of anyone who upvoted these people. Alas, HN loves being a bandwagony echo chamber, so the pain continues :)
Hmm, it appears that they relied on the quality control of forum moderators but didn’t do any of their own QC?
From the methods section:
> The database of drug narratives from Erowid’s “experience vault” is entirely anonymous. Rigorous quality control is conducted by trained experts with careful consideration given to “quality, credibility, and focus on effects or outcomes” (www.erowid.org). By means of that mechanism, all reports that provided the basis for the present investigation have successfully exceeded a standard quality filter.
I don't think they trained an LLM. This looks like TFIDF on a bag of words to me, possibly getting informative features from a prediction model.
Did something similar on the same data at school many years back, but I used the category labels to try and predict good and bad trips. The model worked abysmally, but the features made for great word clouds.
It's been a long time since I looked at the site, but I was pretty active on it during the 90s. Arguably the prose could be quite interesting, but trip reports that were posted seemed quite congruent with the spirit of Alexander Shulgin's entries on which they were based?
Let's hope that lots of old forums in other areas of interest (programming, engineering, etc.) are preserved so that the same sort of work can be done. A current forum which would be most interesting to see results from would be the Ray Peat Forum, for better or for worse.
This reminds me: some years ago, someone noticed that some translation engine, maybe Google's (vague enough for you?) advertised a whole lot of obscure languages but didn't work all that well for some of them: if you translated from English to whatever and back, you'd get mostly gibberish with a lot of oddly Biblical turns of phrase embedded in it.
I got to thinking about that, and I came up with a theory: Machine translation of course requires a corpus of texts that exist in at least two languages, a Rosetta Stone. The Bible is the most translated book on Earth. For tiny or dying languages, the kind with some thousands of speakers at most, the Bible may be the only written work that's ever been translated into or out of that language. So the Bible is their entire training corpus, and it's not enough.
I spent a summer in Papua New Guinea in 1999 where I had the opportunity to meet one of these folks. He was a british man, from Yorkshire, who through force of believe made first contact with the tribe in the 1970s. His first job was to chop down the forest to make a runway, from which he could bring food from the sky which secured his position with the tribe. From there he developed a writing system for their language so they could read the bible, he'd recently finished the New Testement and was starting on the Old Testement. Sometime after the runway he'd brought over his wife who'd essentially built a british farm house with crockery in the jungle and they'd had 4 or 5 kids who mostly flew for the Missionary Airforce. Talk about a lifelong dedication!
The most fascinating thing about the place was that later the whole area had been designated a special scientific survey protection region. They protected the environment by scientists bringing money to do research to pay the locals (kind of like eco tourism protection for areas that aren't very safe for tourists). These folks were mostly hardcore evolutionary biologists rather than creationists like the missionaries, so there were older folks living in the village who in their life had gone through the history of western philosophy: they grew up with animast beliefs, converted to christianity as teenagers and now are experts on evolution. Conversations with them about the nature of existence remain some of the most inspiring and interesting in my life, I recorded a bunch and keep meaning to transcribe off cassette tape sometime.
Just a massive pity that they rarely thought of writing down any of the original stories or histories from the culture's they were pushing Christianity onto.
Ironically leeches are used to this day. At the most specialized hospital in Sweden they are sometimes used to treat wounds that have a hard time healing. IIRC they somehow promote revascularizing of the tissue.
if it was useful, bigPharma would provide it for us! /s
to an extent, I understand the concept of studying the old plant remedies for what is actually useful from the plant and which remedies really do not harm but do no actual good either. dosing is important and using raw plants makes that really hard. creating clean extracts of known qualities makes a helluva lot of sense. i just think bigPharma took the concept too far.
It seems likely that things with no possible way to make decent profit out of them wouldn't have much research interest in them. Of course it doesn't mean that any such thing is useful, but it definitely makes effects of something unclear and difficult to actually prove or disprove.
When it comes to something like drugs, there's no chance in hell I'm trusting what an AI is hallucinating (heh) back at me. If you're looking for reliable information and trying to practice harm reduction, why on earth would you turn to an AI which famously makes shit up all the time?
Tangential, but people interested in trying psychedelics should read about HPPD: Hallucinogen-(Induced) Persisting Perception Disorder. It is a potentially debilitating and long-lasting side effect of using hallucinogenic drugs which can cause changes to your vision, or depersonalisation/derealisation disorder at the extreme end. It may increase your side effects from using psychiatric medications in the future.
It is often caused by using LSD for the first time. Drug mixing makes it more likely to happen.
Personally, I'm avoiding psychedelics so that I don't get it. [Edit] The downvoters are shameless and reckless, and potentially responsible for ruining lives. Stop acting like a cult.
I took LSD with a group of close friends, one of the guys didn't come "down" from the trip. It was a very off putting experience, watching someone you knew 24 hours ago never be the same again. (there was a history of other drugs, but this particular trip was day and night difference)
He was in and out of mental asylums for about 10 years, and recently has just went missing.
I still take drugs when I want to, and don't advocate against them.
My rough advice is to just wait until you are at least 25, follow the safe, sound, setting tips, and always premeditate on what you about to do before you do it.
I had a bad trip, and wasn't the same for three years afterward (in a very specific way that didn't significantly alter the trajectory of my life, thankfully.). Still have echos of it about a decade later. LSD is not a toy and should be respected. I am of the opinion that it would be legal but only when be dispensed by a licensed professional, and it would be required to have a trip sitter. Not that I think that could realistically work, or be practical with where we're at currently.
Maybe but then I took LSD at burning man and had a lovely time wandering around and looking at big dusty art, and I believe such an experience should not be illegal.
To be clear, I don't think such an experience should be illegal.
At Burning Man, there is a place called Sanctuary, which operates essentially a psychedelic ER for people on bad trips. It is run by MAPS, in conjunction with the Rangers, and is located next to Rampart/traditional medical services. The problem with it is that a) it's not well known, and b) if you're having a bad trip, and can't get there/their help, they can't help. So my fantasy solution is that to have trip sitters assigned to people/groups going on an LSD trip to more fully mitigate the problem. Both on and off playa. I do realize this is infeasible since there is no such pool of experienced trip sitters to pull from, plus LSD is still a controlled substance, unfortunately.
Here's my pro-hallucinogen argument: if you have persistent, treatment resistant anxiety, there is nothing even close to a heroic dose of your favorite hallucinogen to kill your ego and let you see through the bullshit in your life. You'll have a pretty bad time during the trip, but the clarity it provides is life changing.
> You'll have a pretty bad time during the trip, but the clarity it provides is life changing.
Or it may traumatize you for life. And that "clarity" may just be an illusion, which doesn't mean it can't have a positive effect, but you can't really count on it.
That's the problem with psychedelics, they are unpredictable. The can definitely do miracles, but they can also do great harm, though usually not by the way of toxicity or addiction.
There is a reason we don't use LSD for psychiatric treatment. This is no conspiracy theory, they tried, and failed. I mean, LSD is a product of big pharma, and as you may have guessed, Sandoz really wanted to make a real treatment out of their patented drug, for money.
Maybe with modern science and better experience with psychedelics, we could success where the others have failed, but we are not there yet. That's why I really don't recommend that type of self-medication.
Now, if you want to enjoy a good trip, go for it, there are some risks, but rather low if you do it properly, and many good things in life carry risks.
But suggesting a heroic dose to someone anxious with the implication that it may lead to a bad trip is going against every advice regarding psychedelics, and essentially replicating the very experiments that failed and caused LSD to end up as a schedule 1 drug rather than a recognized treatment. It may work, but that's a huge gamble.
There has been a research gap on LSD (and many other 'drugs') of nearly 50 years. Only in the last decade or so are we actively researching 'uses' again.
Disclaimer: While I have big problems with current legal regimes and discourses on 'drugs' (virtually always excluding ethanol and nicotine), I do NOT promote or advocate the use of psychedelics. Most of them (see above) are massively under-researched and many available substances are not at all pure, so there is always a non-trivial danger associated.
>There is a reason we don't use LSD for psychiatric treatment. This is no conspiracy theory, they tried, and failed.
This is wrong. There's a lot of misinformation around the history of psychedelics. The early "clinical trials" were extremely bad science. The short version is the scientist leading the team who first synthesized it was also the person who first attempted to popularize its use, and is pretty much a poster-child for "sex- and drug-crazed hippy". The US had already adopted a heavy anti-drug rhetoric and later LSD was rolled into the war on drugs effort. There was bad-faith-actors on both sides of the fence for a long time.
>Or it may traumatize you for life. [...] But suggesting a heroic dose to someone anxious with the implication that it may lead to a bad trip is going against every advice regarding psychedelics
Completely agreed. The potential for long-term trauma cannot be understated. They shouldn't be played with unsupervised, and the supervisor should be well versed in what happens with a good AND a bad reaction. (I don't like the term 'bad trip'.)
> The short version is the scientist leading the team who first synthesized it was also the person who first attempted to popularize its use, and is pretty much a poster-child for "sex- and drug-crazed hippy".
You seem to have missed some history.
LSD was first synthesized in 1938 by Albert Hoffman and first accidentally ingested in 1943 by the same. He was not a 'sex- and drug-crazed hippy'. You are certainly thinking of Timothy Leary, a clinical psychologist, had a mushroom experience in 1960 and spun that into a bunch of research into uses of psychedelics in psychology. He was eventually fired by Harvard in 1963. He then kinda goes off the rails, eventually he went to and escaped from prison along with many other exploits.
You are correct. That was badly edited on my part: I wrote half a sentence about the first synthesis but forgot to remove it when talking about the popularization.
We have to ban accounts that post this aggressively and/or abrasively, and we have had to ask you about this more than once before. That's not ok. I don't want to ban you, but we need this to stop. If you'd please review https://news.ycombinator.com/newsguidelines.html and take the intended spirit of the site to heart, we'd be grateful. That means curious, respectful conversation without snark, swipes, or putdowns, for starters.
I did some psychedelics, and my friends did even more. They were always telling me about their truly groundbreaking personal epiphanies, but they never really changed their lives.
Under rated comment. I've seen discussion of this in the past and it's been speculated that there might be a sizable portion of users of these drugs having HPPD to some extent. I can attest personally to effects of HPPD on myself when I was younger:
- It completely destroyed my night vision to the point that I couldn't see at all in low light. No eyes would never 'adapt' because the amount of visual snow (like a TV set to an untuned channel) would overwhelm what I could see
- Massive halos around lights. Night time became dreamy.
- After-image impressions of reading text. So it became a little harder to read.
- It's very easy for me now to experience closed eyed visuals when I'm completely sober. It usually happens when I'm tired.
These effects can happen with all hallucinogens and weed is psychoactive, too. People don't mention these nuances that visual changes may linger after the trip. You can imagine that some people might feel like they've just screwed themselves up forever. I felt that way for about a month before I got used to it. The effects above have mostly gone away now though.
I really do think that these kind of drugs permanently change how you visually perceive the world. So I think your concerns are completely valid. I have heard stories about this exact topic online from people freaking out.
but imagine the potentially debilitating side effect of living life without expanding your perception.... /s
well, kinda /s.
These are drugs we're talking about. The use of drugs comes with risk. Being able to assess risk versus reward is apart of being an adult and making adult decisions, but I'm glad you've included more information (HPPD) for perusal. I've done my share of psychs but didn't know about this. Fascinating!
You can get it from one trip. Hell, you can even get it from smoking a joint. I know this sounds like something you might see in a DARE campaign but it is real. It's a very under appreciated risk of psychedelics. If you define HPPD as some level of negative visual changes that interfere with functioning I actually think a lot of people who have used psychedelics would quality to some extent. The question is: how significant are the changes? Are we talking almost imperceptible extra visual white noise or maybe like what I experienced that even made it harder to read and see at night. HPPD does seem to fade but it can take a long time.
There was a poll on the HPPD subreddit* which showed that more than half of people there got it from less than 5 uses of a psychedelic. But people there also seem to mix drugs, like cannabis with LSD.
Oh a poll on a subreddit. Please forget my previous question - if you’re using pills on subreddits as a source there’s really no hope of having a serious conversation. You’re going on and on because of some personal bias.
Erowid is a member-supported organization providing access to reliable, non-judgmental information about psychoactive plants, chemicals, and related issues. (from mission statement)
Anyone who has benefited from Erowid's psychoactive drug info should consider donating!
Why Donate to Erowid Center? The issues surrounding psychoactive drugs are complex. Prohibition discourages the spread of accurate information in favor of "Just Say No" rhetoric. But prohibition has failed and we need to move on to education. Information sources need to prioritize accuracy rather than maintaining politically driven anti-drug messages. Erowid provides free, balanced, reliable information about psychoactive plants and drugs. We don't convince users to abstain or legislators to pass particular laws. We provide the information necessary for people to make their own rational choices.
Nah, I'm kidding, there's no way I can sleep inside a MRI machine. I just think of dancing electrons going through my body. I also think it is an extreme version of dubstep.
> Venture capitalists are pouring millions of dollars into new psychedelic-based antidepressants that move beyond the classic SSRI model, but without a precise index of the neural processes at work, pharmaceutical patents are unlikely to be approved.
Oh I see! THAT'S what they are REALLY doing. One more reason to sanitize the web of old posts. Doesn't even matter if those stories the LLM is being trained on are actually useful and valid, if they are just enough to persuade some board member to grant a patent on a molecule that others have discovered long before the patent troll.
Everyone, I just happened to see this random piece of paper flying around in the wind. I caught it and it said the silliest thing, so I felt like sharing it here:
The headline made me think either Erowid or The Hive until I read a bit more and Erowid made more sense. I had forgotten how much I missed reading through and contributing to Erowid many years ago.
"All information on Erowid.org is provided non-commercially for purposes of teaching, scholarship, research, and criticism."
The point of the website is to collect information about drugs for education and research purposes. It's a 501c3 nonprofit, which operates explicitly for public education. It's funded by donors, not by consumers of the info collected.
Research falls under "purpose and character of the use", which is only one of the four factors considered in fair use analysis. I gather a court could decide that this outweighs the other factors, but I doubt that can safely be taken for granted as a general rule.
Did you ever use the site? It was like a handshake agreement situation where you had no idea how full of shit the posts were. If people weren't outright lying, they could (very well) be experiencing wish fulfillment or placebo. Many of the "negative" experiences were similar - describing reactions that were frankly hard to believe. Posts tended towards fictional exaggeration. Further, many users were anonymous, pseudonymous, lying about identity, or posting in a time when many believed no one would be probing through the internet trying to identify folks.
tl;dr - I seriously doubt any of the users wants to be identified, much less compensated for providing "data".
That's been my experience with the site, too, which made me raise my eyebrow at the article's claim of "first-hand testimonials that have been vetted for authenticity and accuracy." If the data's such junk (which I surely won't argue with), why study it?
e: From the original paper:
> Rigorous quality control is conducted by trained experts with careful consideration given to “quality, credibility, and focus on effects or outcomes” (www.erowid.org). By means of that mechanism, all reports that provided the basis for the present investigation have successfully exceeded a standard quality filter.
I once had a disagreement with a lead dev for an in-house tool that was a glorified DAM. their forms were not kept up to date, and so users were forced to enter bad data which skewed query results. when pressed about why the fields were forced to choose from an outdated pre-populated enum, i was told "bad data is better than no data".
maybe the AI researchers are from the same school of thought. having bad/unvetted medical data/advice on psychedelics would be better than no data at all, right? /s
I'm a very experienced psychonaut and even I won't dare touch that stuff. One of my friends did and just watching him on it made me never, EVER want to do it.
Thankfully, outside of Hacker News, and particularly on Erowid, there is not as high a frequency of the “where’s MY payday” mindset. This is just American / Silicon Valley brain rot.
The general idea of training LLMs on forum archives seems promising though!