@JhanicManifold's banner p

JhanicManifold


				

				

				
6 followers   follows 0 users  
joined 2022 September 04 20:29:00 UTC

				

User ID: 135

JhanicManifold


				
				
				

				
6 followers   follows 0 users   joined 2022 September 04 20:29:00 UTC

					

No bio...


					

User ID: 135

What GPT does is predict the next token. That's a simple statement with a great deal of complexity underlying it.

At least, that's the Outer Objective, it's the equivalent of saying that humans are maximising inclusive-genetic-fitness, which is false if you look at the inner planning process of most humans. And just like evolution has endowed us with motivations and goals which get close enough at maximising its objective in the ancestral environment, so is GPT-4 endowed with unknown goals and cognition which are pretty good at maximising the log probability it assigns to the next word, but not perfect.

GPT-4 is almost certainly not doing reasoning like "What is the most likely next word among the documents on the internet pre-2021 that the filtering process of the OpenAI team would have included in my dataset?", it probably has a bunch of heuristic "goals" that get close enough to maximising the objective, just like humans have heuristic goals like sex, power, social status that get close enough for the ancestral environment, but no explicit planning for lots of kids, and certainly no explicit planning for paying protein-synthesis labs to produce their DNA by the buckets.

This part never usually pans out lol.

Oh it did for me, I still remember her reaction when she saw me for the first time in like 6 months: * looks at me, does a double take, eyes widen, face becomes fully red, furtive looks the whole evening *. I actually feel a little bit guilty about just how good revenge feels.

Yeah the pulsing patterns seem very specific, and are probably the entire technical moat of the company. The way it works is that there are a variety of "programs" on the app, so you have a "stress program" that lasts 15 min, which starts with short, intense pulses that get quicker and quicker, then you have a "calm program", a "sleep program", etc. The device modulates both the intensity and the frequency of the haptics over time depending on the program you chose.

Get a meat thermometer and grill to 170F internal temperature. I mix 4 eggs, 2 cups panko bread crumbs, 1.3kg of extra lean beef, 3 tbsp smoked paprika, 2 tbsp garlic salt and 2 tbsp crushed oregano together into the burger paste, and they turn out great every time if grilled at around 450F. Spices don't burn if you mix them into the meat itself.

Short answer: it depends on how much cardio you're doing already. Cardio contributes to systemic fatigue, and too much of it will reduce your NEAT (Non-Exercise-Activity-Thermogenesis), basically you'll fidget less and be less inclined to take the stairs instead of the elevator, which will have a net-negative effect on caloric balance.

Very relevant video: Does More Cardio Equal More Weight Lost?

(and I'm very surprised by your 100cal per 2 miles walking number, I use this calculator for estimating walking calories, which gives me much higher numbers)

I have, perhaps surprisingly, managed to make new close friends as a mid twenties guy. Here's the trick, you first say "hey bro (or girl I want as a friend), I really enjoyed talking to you at that thing that time, we should get coffee sometime, how's next thursday?". Next, you need to talk one-on-one, it's hard to make a new close friend when you're always in a group with them, and you need to (gradually) tell them some of your secrets, vulnerability is required for close friendship. Virtually everyone to whom you offer one of your small secrets will reciprocate with one of their own, and thus the friendship builds.

Then, once you've hung out a few times one on one, you hit them with a heartfelt "hey man, great hanging out today, this is kinda weird to say, but I want you to know that I'm really grateful to have you as a friend". Think of yourself, would you allow yourself to grow closer in friendship with someone you weren't sure wanted the same thing? Realising that someone you considered a close friend doesn't feel the same probably hurts at the same level as romantic rejection. That's why the earnest declaration that you consider them a close friend works so well, it alleviates their fear that their friendship will not be reciprocated. People do want close friends, they're just afraid to hope.

Ben Shapiro also makes the point that they drop "knock bombs" before the real bombs, the only purpose of those bombs is to shake the building to tell civillians to evacuate.

Buying it on indiamart.com from India pharmacies, I haven't really had problems with importing it in Canada

I might have fucked up one of the easier ones, but gotten avulse correctly. That would explain things if difficult questions are worth more.

hmm, top 0.2% for me, I'm kind of surprised, I never made any explicit effort in all my english classes, and I'm not even a native speaker.

/images/1685626469587057.webp

Athens was filthy. Almost every building is completely covered in bad graffiti up to a certain height.

And don't forget the brothels... they're called "studios" there, and there was a truly surprising number of them when I went last summer.

The recommended dosing has you slowly ramp up the dose over like 2 months until you get to the truly effective amounts. I suspect that you would get very large gastro-intestinal side-effects if you just went from 0 to the effective dose. Semaglutide also has a very long half-life of a week, so if you're injecting 1.0mg/week steadily, to get the same blood concentration with a single injection you'd need 2.0mg. I think it would be a fairly bad idea, there's a good chance that the diarrhea and vomiting you'd get would wipe out any mental benefits from the lack of hunger.

Hmm, somehow I doubt that. It's not like all schoolgirls I pass by giggle, but it's happened like 3 times over the past month (I probably pass by more schoolgirls than you because I take the subway). But then I'm curious, what things did you notice in your case?

yeah I just use insulin needles from amazon, with alcohol prep pads. The only thing you can't get from amazon is bacteriostatic water, which is still over-the-counter, just not from amazon (I got it from here). Peptides in general are best stored dry in a freezer or fridge, but I keep them at room temperature away from the light, the tests I've seen don't really show any meaningful degradation in a few months. Though they start to degrade much faster once you add in water.

Another thing, you probably won't be able to just buy a month's supply, the minimum order quantity is 10 vials of 2mg, which comes in a prepackaged little box disguised as a chinese beauty mask. I don't think the supplier is set up to ship orders which aren't multiples of 10 vials.

Surprisingly, it kind of does! It felt to me like it helped me not think of work or other things while I'm going to sleep, the vibrations on your skin have a way of capturing attention very effectively. I tried it out for a few weeks after seeing it recommended here, but I'm now returning it, the difference just isn't that big for me, nowhere near the magnitude that the bedjet is making.

Yeah I wouldn't really trust the tests on that page too much, I took their IQ test and got 158 on it, which a ridiculous overestimation based on the previous tests I took, where I got like 135-140.

I'd be super happy to be convinced of the contrary! (Given that the existence of mesa-optimisers are a big reason for my fears of existential risk) But do you mean to imply that gpt-4 is explicitly optimising for next-word prediction internally? And what about a gpt-4 variant that was only trained for 20% of the time that the real gpt-4 was? To the degree that LLMs have anything like "internal goals", they should change over the course of training, and no LLM is trained anywhere close to completion, so I find it hard to believe that the outer objective is being faithfully transfered.

I mean, if you want large costs for the same benefits, there are plenty of effective weight loss drugs with a shit ton of unhealthy side effects, DNP and trenbolone will make you lose weight, they just might also kill you lol, their side effects are not subtle at all. Free lunches are rare in the world, but there's certainly lunches that are more expensive than others.

I'm barely half into the first chapter, but man, I already fucking love this! I can sense that this story is going to scratch an ungodly number of literary itches for me...

Why do you think that? This combination of features would be selected against in evolutionary terms, so it's not like we evidence from either evolution or humans attempting to make such a virus, and failing at it. As far as I can see no optimization process has ever attempted to make such a virus.

In the drug realm:

Biggest by far: semaglutide for weight loss. Works so well that Walmart is noticing sales drop... if that isn't an amazing endorsement I don't know what is.

Second biggest: occasional moderate dose (1.5g) phenibut taken 8 hours before a stressful social situation.

It's a close cousin to benzodiazepines (though much easier to acquire), so the withdrawal symptoms are massive as fuck, there's a reddit community dedicated to people who've fucked up their lives taking phenibut everyday, though I can't seem to find it right now. I also notice increased anxiety on the day after I take a dose. It works very well for my use case, but I periodically remember not to treat it lightly.

For regular consumption, creatine is the king, there's no other supplement with as clear and massive of a benefit, it makes you stronger, helps cognition and 30 years of intense research hasn't found a single negative effect (maybe apart from slight intestinal distress in some people).

I also use phenibut and modafinil on special occasions. Phenibut is amazing at lowering social anxiety in particular while leaving your reasoning capacities essentially untouched, and modafinil is good at boosting concentration and making you stay awake. You shouldn't take these daily, phenibut in particular will fuck up your life if you take large doses daily, the best is to use it for occasional job interviews or presentations, for which it works amazingly well.

Hmm, I would say that if the secret is like "AI will kill everyone and there's nothing you can do to stop it", don't tell her. If the secret is like "your father was a murderer" or "you have terminal cancer", then do tell her, because it's "her business" in some sense. Another factor is how much knowing the secret will eat at you over time, if the person is a close friend of yours, keeping this secret forever will be a great burden and you should tell them, if it's just an acquaintance, then not so much.

If you think you're good at acting and deception, you could even indirectly ask for their opinion on the matter, all you have to do is invent a new secret with all the relevant characteristic about some distant friend, then ask them whether you should tell your distant friend.

Hmm, basically all the libraries I listed except maybe for pytorch haven't changed all that much since 2021, gpt-4 should really still be very useful with all of them. What it will have trouble with is a library like "Transformer" by huggingface, which lets you automatically download and use pretrained deep learning models. But to even use a super-high-abstraction library like that one you still need a bunch of "glue skills" like knowing how to load a .png image from your computer into a format that the high-level functions can understand, and how to interpret and visualise the output of those high-level functions. GPT-4 would be amazing for all of that.