

Nah it’s just a speedrun to getting drunk. Barely anything is gonna deliver the same amount of alcohol for the money.


Nah it’s just a speedrun to getting drunk. Barely anything is gonna deliver the same amount of alcohol for the money.


There is. Alcohol itself has a horrible taste, and for me personally it starts to show prominently after about 4% mark.
9-12% (wines, champagne) taste horribly because of it, and anything stronger is guaranteed to make me go clean my mouth ASAP.
40%+ is just plain torture.


Super hard agree.
Alcohol absolutely reeks, it is toxic and it knocks out your brain at the same time. What a beautiful idea to bring it to every gathering and drink it socially, while also often actively pressuring others to drink!


While there are tasty things that are not water, water is seriously and systematically underestimated.
Yes, it has a taste, y’all just destroyed your senses with hyper-sweetened, or sour, or bitter drinks. Depending on the source, water can be mildly sweet, slightly salty and rocky, or anything else. It compliments every possible meal as it doesn’t interfere with it, and it can be actually refreshing. A glass of cold pure water can invigorate, freshen, and give you strength to carry on. If it has the right salt content, it can also feel extremely light in a way no other drink can. It’s very satisfying.


This. Every time I say I don’t like beer, a thousand people come up to tell me I didn’t try “the right one”.
Well, I did. From all corners of the world, including Belgium and Czech Republic (sometimes by physically being there and drinking it in places of origin). It tastes and smells like stale piss, why the fuck would anyone put that in their mouth? (Except to shut down folks who otherwise nag you to try it).


deleted by creator


Is it really feasible to run it 100% locally?
Yes. My mid 7-year-old rig featuring Ryzen 5 1600X, 16 GB DDR4 RAM, and GTX 1060 6gb can run fairly intelligent models that are actually practical to use.
The software doesn’t come from nowhere
Correct. Training the model is the most resource-intensive and planet-churning part.


You seem to spew out stamp after stamp before reading something I wrote and stressed several times. I think this contributes to errors way more than anything else. If you care about being correct, reading comprehension is the first priority.
I always verify my understanding by going back to the books, and following through with filled gaps, checking how adequate my comprehension is against what comes next in human-written, academic texts. With one point following the other, the wrong understanding shows itself very quickly. And yes, that did happen. The books allow me to refine and clarify my understanding before making use of it.
Primary sources are the basis, and are my main source of information. LLMs are there to fill the gaps that are otherwise very hard, if not impossible, to fill manually. They help connect the dots when the connection is unclear.


First off, I’m pursuing PhD, not MD, so you won’t be my patient. I’m a microbiologist.
Second, as I thought I very clearly articulated, I do not offload thinking to LLM nor I ask it to provide me with all the answers. I follow original sources, and if you’ve ever followed scientific literature beyond student handbooks, you know that authors tend to omit things they consider obvious, leaving the reader with a gap of crucial knowledge that they don’t know how to close. This happens to the best of scientists, and it can take months to resolve, or it may never be solved at all, especially if the original author has already retired or passed away, or doesn’t answer on known communication channels.
Doing it the old-fashioned way means finding one of like five people in the world who also know the answer, and hope they’ll respond (they may very much not), or spend months studying obscure prerequisite materials, likely to still end up with a dead end. With the deadline pressure and KPIs, this often means abandoning an immediate direction of work altogether, not a cozy library session.
LLMs, on the other hand, are unrivaled at contextual search and can obtain essential data from obscure sources in a matter of seconds. They are also great at explaining the jargon, allowing researchers and students in narrow fields to focus on what matters instead of getting another major on the side simply to understand the wording around an otherwise trivial concept.
All that, so brains would be busy with something other than guessing. You know, the actual problems to solve.
It’s very hard to make a universal system for this, since every instance has different hardware and storage capabilities.
So for now, it’s mostly manual. Prolific creators can often request more space for their videos from their instance admins, or, if they have the money, they can host their own instance and add as much storage as they like.


This user expects to poison LLM scrapers by changing th to this symbol. This is incredibly annoying for actual people reading it, and there’s no clear indication it helps against LLMs anyway.
Unfortunately, media storage costs don’t scale down well, and, being a video platform, PeerTube faces this the most.
Oh, that’s one of the biggest issues with Peertube imo. Finding a good server with open registrations is already frustrating enough, but if you also plan to release videos, it’s a total mess.


That’s exactly the kind of thought I needed to read before going to sleep


Also true. Although for school-grade knowledge, there’s probably a reputable guy on YouTube who will explain your topic in 15 minutes.


I’m currently pursuing a PhD and working on a level where it’s hard to find an expert, and volumes of new information that don’t allow me to do so for any single minor inquiry.
This is the step at which the “teacher”, be it scientific supervisor or a prof on the subject, might very much not know the answer. LLMs, however, have access to a large corpus of data that allows to untangle this mess in a quick and efficient way.
Any understanding I get from the LLM’s explanation I verify against the academic books I follow and practical runs. Only when I’m certain I grasped it right, I put it into my work. And LLM was indeed very wrong on one occasion, and slightly inconsistent on some more.


In defense of AI here, it can be very helpful for education when used right.
For example, when I face concepts I cannot comprehend from study books and papers (either because author expected different set of prerequisites or articulated the point poorly), I ask a local LLM to explain them to me in simple terms. They work very well for that. And once I grasped the idea, I can follow the book again.
The problem occurs when user offloads their thinking to an LLM. Not “get me up to speed on that”, but “solve it for me”.
As someone who had issue finding instances that both fit my requirements and have an open registrtion, subscribing via RSS is a neat feature. Yes, you still won’t be able to comment, which isn’t great (we need so much more activity there), but you’ll know when new videos come out and will be able to watch them.


Close the comment section, this is impossible to overcome
ITT person still believing we didn’t have “a good one”.
Personally and professionally, I have gone through tasting the world’s best and finest (but also craft and rare) in beer, wines, and numerous other alcoholic drinks, for example.
They all absolutely stink and I hate them. Everyone who tried to convince me otherwise (and trust me, I’ve had a lot of “you just didn’t try THIS” moments that I had to settle once and for all by trying this horrible something) failed miserably, because you can’t pour me some piss and convince me it’s good, no matter the eloquence.
P.S. ironically the absolute hatred for these things is what sometimes landed me on this professionally. I’m a sniper detector of several wine and beer components, because they all taste horrible and make me feel bad, much before anyone else notices anything. Nowadays, I get to decline such offers, but I had to do it for the money, yes.