LSD is dropped onto paper in solution though. So to control dose is easy since you can easy halve a dose by doubling the volume of solution. Dosing a powder/crystal is much more difficult, especially if you need to get it back out of solution.
LSD is a powder/crystal (a salt). People just don't consume opioids orally, usually. There's something similar though: skin patches, since (other than LSD) fentanyl can be absorbed through the skin.
In context, we're talking about pills cut with fentanyl, in which case it is often consumed orally, mixed in at a very small concentration compared to the other ingredients.
Powedered drugs like cocaine mixed with fentanyl are even more horrible, since there is absolutely nothing to keep the concentration of fentanyl homogeneous throughout as it is handled.
As a student at a top worldwide university, I can tell you we order a lot more stuff off Amazon and eBay than you'd think. There's an awkward middle ground where you either buy something cheap or make it yourself because labour is basically free in academia thanks to the large amount of students and staff but grant money is not.
yes ! but it also assumes you have: a good optical breadboard + bench + dampeners, a beautiful set of lenses, all sorts of nice lasers and kinematic mounts and linear stages etc etc
so yes, we _also_ (back in my phd lab) built equipment in that sense, but there was a pretty good foundation of Fairly Fancy stuff already sitting around !
All of those parts can also be acquired through alibaba for a stiff discount off the thorlabs pieces though. Whilst some labs have fancy stuff going around, a significant amount don't and there isn't very good equipment sharing between labs at most institutions.
If education is welfare then so is everything. Defence is welfare becuase before you might have to hire private security. Police and fire serviecs are welfare because they used to be private. etc....
Ok so can you name me a single piece of government spending that isn't welfare Or are you advocating for governments to just cease existing all together.
There isn't much difference between this and an OS. It's just that in an OS you switch between processes, whereas here they switch between async tasks. One could argue this is pretty semantic and that you could easily just call this an OS.
There are design patents specifically for looks[1], in other countries such as the UK where I am from this is known as registering a design rather than using the word 'patent'[2].
A lot of people are commenting on the conclusion but I'm surprised no one is commenting on the methodology? The distributions given by the models seem weird. The LLM's enough so that I would just discount those and focus on the BERT models, but even then roBERTa for instance seems to suggest there is NO positive sentiment, with only scores of 0.5 and above given. Then there is the axis which is "ai_sentiment" against the classification, but it's not clear what "ai_sentiment" is, and it's not expanded upon in the paper. It seems to basically just map to the DistilBERT score apart from a few outliers?
Given that, it seems that there is basically zero agreement between DistilBERT and the other models..... In fact even worse they disagree to the extreme with some saying the most positive score is the most negative score.... (even acounting for the inverted scale in results 2-6).
Fair point, `ai_sentiment` should have been defined explicitly. It's the production score from DistilBERT-base-uncased-finetuned-sst-2-english, the same model family as Cloudflare's sentiment classifier. That explains the r=0.98 correlation you noticed. And you're right that the models disagree. This isn't measurement error though. They learned different definitions of "sentiment" from their training data. DistilBERT was trained on movie reviews (SST-2), so it asks "is this evaluating something as good or bad?" BERT Multilingual averages tone across 104 languages, which dilutes sharp English critique. RoBERTa Twitter was trained on social media where positivity bias runs strong, hence the μ=0.76 you see.
For HN titles, which tend to be evaluative and critical, I assumed DistilBERT's framing fits better than the alternatives. But the disagreement between models actually shows that "sentiment" is task-dependent rather than some universal measure. I'll add a methodology section in the revision to clarify why this model was chosen.
Thanks for clearing that all up for me, look forward to seeing the revision!
It would be interesting to see some of the comments that seem to be polar oposites in sentiments between the models. So ones where they are the most positive sentiment by one model but the most negative by another to analyse the cases where they disagree the most on their definition of sentiment.
reply